ChapterPDF Available

Sales Promotions

Authors:

Abstract and Figures

This article is designed to assist managers in understanding the current state of the art of promotional modelling and its managerial applications, and to identify future issues that need to be addressed. It is organized as follows. Section 24.2 describes the major types of sales promotion offered. Section 24.3 discusses the behavioural underpinnings of sales promotions. Specifically, several theories are provided that explain why consumers respond more strongly to sales promotions than to a price decrease and shows the benefits to managers of understanding why sales promotions are used. Section 24.4 describes how promotions affect sales, and discusses the sources of incremental volume associated with a promotion. Section 24.5 discusses strategic issues associated with sales promotions, and provides insight into why firms may still offer promotions even when they are unprofitable. Section 24.6 provides an overview of empirical methods used to estimate sales promotions models. Section 24.7 discusses emerging issues related to sales promotions. The article concludes with a section focused on managerial implications and emerging managerial issues in sales promotions.
Content may be subject to copyright.
1
S
ALES
P
ROMOTIOS
R
OBERT
C.
B
LATTBERG
T
EPPER
S
CHOOL OF
B
USINESS
C
ARNEGIE
M
ELLON
U
NIVERSITY
R
ICHARD
A.
B
RIESCH
C
OX
S
CHOOL OF
B
USINESS
S
OUTHERN
M
ETHODIST
U
NIVERSITY
S
EPTEMBER
,
2010
F
ORTHCOMING IN
“O
XFORD
H
ANDBOOK OF
P
RICING
M
ANAGEMENT
The authors would like to thank William Dillon, the Editors and an anonymous reviewer for their
helpful comments and suggestions for this chapter.
2
Introduction
Over the last three decades, there have been numerous academic papers in economics,
psychology and marketing regarding the topic of sales promotions. Practitioners and consulting
firms have made significant progress in applying and, in some cases, extending models from the
academic community. Portions of this growth can be attributed to the increased managerial
importance of sales promotions, as consumer packaged goods (CPG) companies allocate more
than 50% of their marketing budgets and 13% of their revenues to sales promotions (ACNielsen,
2002; Gómez et al., 2007). Part of this growth is also driven by the wider availability of data,
particularly point-of-sale (POS) data from supermarkets and other retailers, which has allowed
academics and practitioners to model and optimize sales promotions. This chapter is designed to
assist managers in understanding the current state-of-the-art of promotional modeling, its
managerial applications and to identify future issues that need to be addressed.
Before going further, it is useful to define sales promotions. Blattberg and Neslin (1990)
define a sales promotion as “an action-focused marketing event whose purpose is to have a direct
impact on the behavior of the firm’s customer.” There is also an important distinction between
sales promotions and a permanent price reduction. Sales promotions are temporary and a “call-
to-action.” If customers do not take advantage of promotions within specified time frames, they
will lose the benefit offered by the promotions. Sales promotions are almost always combined
with some type of communication (e.g., a retailer ad) that the price is reduced and that the time
period is limited (price is reduced only up to some point in time). Long-term price reductions
may be accompanied by a price reduction signal (e.g., Walmart rollbacks in the US) but the time
period is “until further notice.” A consumer can wait to make a purchase when price is reduced
without a time limit with the only risk being that the consumer may not accurately estimate when
3
the price will increase. The distinction between sales promotions and permanent price reductions
leads to differences in consumers’ responsiveness to price changes and buyer (retailer) behavior
to trade promotions.
The remainder of this chapter is organized as follows. In the next section, we provide a
description of the major types of sales promotions offered. Section three provides a discussion
of the behavioral underpinnings of sales promotions. Specifically, several theories are provided
that explain why consumers respond more strongly to sales promotions than to a price decrease
and shows the benefits to managers of understanding why sales promotions are used. Section
four provides a description of how promotions affect sales, with a discussion of the sources of
incremental volume associated with a promotion. These sources of the incremental volume
dramatically affect the profitability of the promotion for both the manufacturer and retailer.
Section five discusses strategic issues associated with sales promotions, and provides insight into
why firms may still offer promotions even when they are unprofitable. Section six provides an
overview of empirical methods used to estimate sales promotions models. Section seven
discusses emerging issues related to sales promotions. Finally, we conclude with a section
focused on managerial implications and emerging managerial issues in sales promotions.
Description of Types of Promotions
Sales promotions are designed for different purposes and different target audiences:
retailer, trade and consumer. Retailer promotions are offered by retailers to consumers to
increase sales for the item, category or store. Trade promotions are offered to members of the
channel distribution (called the trade) and are designed to stimulate the channel members to offer
promotions to consumers (retailer promotions) or the channel member’s customers. Consumer
promotions are offered directly to consumers by manufacturers and are designed to stimulate the
4
consumer to make a purchase at some point close to the time of the receipt of the consumer
promotion. Figure 1 shows the interrelationship between the types of promotions.
===== Put Figure 1 about here =====
Because of the growth of the Internet, Internet promotions have become prevalent for
both consumer and retail promotions. We separate out these promotions because the medium is
interactive and offers new capabilities to firms through the vast information now being provided
to manufacturers and retailers.
Retailer Promotions
The key elements of a retailer promotion are price discounts, the medium in which price
discounts are offered, communications of sales promotions and objectives of the promotion.
Price discounts can take different forms ranging from straight price discounts to buy-
one-get-one free (BOGO’s), to frequent shopper card discounts, to buy A and get a discount on B
(bundled promotion). Commonly used discounts are listed in Table 1 with a brief description of
each.
Table 1. Common Types of Retailer Discounts
Type of Retail Promotion Description
Price Reduction Retailers temporarily decrease prices on product.
Retailer Coupon Retailers issue coupons for product in their advertisement or on
the shelf.
Free Goods The consumer receives free goods as the discount. It includes
buy one get one free (or buy X get Y free), as well as promotions
5
where goods in complementary categories are given away (e.g.,
salsa for tortilla chip purchase).
Sweepstakes The consumer is entered into a contest where they have the
chance of winning cash or other prizes.
Free Trial Consumers are given free samples of the product to encourage
purchase of a new product.
N-for The retailer offers a discounted price for the purchase of a set
number (N) of items purchased, e.g., three for $1.
Discount Card Consumers sign up for a card that tracks their purchases. In
return, the retailer provides discounted prices on some items in
the store for only those consumers with the card.
Rebates Consumers receive notices of a rebate at the shelf or display and
then mail in proof of purchase and the rebate form.
Bundled Promotion The retailer gives the consumer a discount for purchasing
products from complementary categories (e.g., hamburgers and
ketchup).
Different types of price discounts have different effectiveness and result in different types
of behavior. For example, three for $2 has a greater impact on sales than a price discount of $.67
because consumers tend to purchase three units. Rebates result in “breakage” which means
consumers respond to the rebate but do not send in the requisite documentation to receive the
rebate. Figure 2 reproduces a figure from Chandon, et al (2000) that measures consumers
perceptions of different types of discounts on two dimensions: utilitarian (monetary savings,
6
improved product quality, and shopping convenience) and hedonic (the opportunities for self-
expression, entertainment, and exploration). Interestingly, they find that non-monetary
promotions work better for hedonic products, whereas monetary promotions work better for
utilitarian products.
===== Put Figure 2 about here =====
The next issue is what medium the retailer uses to provide price discounts. One option is
simply to offer a discount in the store. Alternatives include frequent shopper card discounts
available at the store or redemption cards sent to the shopper’s home to be redeemed on the next
visit. The relative merit of the medium chosen depends upon the retailer’s objective. Frequent
shopper discounts reward “regular shoppers.” Mailed discount cards drive shoppers into the
store. These media vehicles are not mutually exclusive.
Communication of the promotion is very important. Retailers use in-store signage,
displays and periodic (weekly) fliers / feature advertising to communicate discounts. Obviously,
it is very important to communicate price discounts so shoppers recognize them. For large
discounts, retailers will have some form of display. For small discounts they will use in-store
signage (e.g., shelf tags). Manufacturers often fund retailer communications through co-op
advertising funds which we cover in the next section.
The retailer can have different objectives associated with promotions. One of the most
common is to generate traffic (customers visiting the store). The choice of the products / brands
offered on the promotion and the level of discount offered greatly influences the amount of
traffic generated. Products / brands that have high penetration levels and high frequency of
purchase are often very good traffic generators for grocery, drug and mass retailers. Obviously,
7
the greater the discount level offered, ceteris paribus, the more traffic that is generated by the
promotion.
Besides traffic generation, another key objective of promotions is to sell excess
merchandise that is caused by overstocking. After Christmas or after major holiday promotions
to clear unsold holiday merchandise are very common. Airlines have unsold seats during a
specific day, week or season and offer special promotions to sell those seats because they are
“perishable.” For a more complete discussion of this area, see the chapter on markdown pricing
by Ramakrishnon in this volume. . Promotions have other purposes beyond just driving
customers into the store and selling unsold, perishable inventory. They can be used to increase
category profitability by switching consumers from lower to higher margin products or they can
be used to generate trial for a new product or a product category that has low penetration levels
(For a more comprehensive list of retailer promtional goals, see Blattberg and Neslin, 1990)
In summary, retail promotions are a very powerful sales tool. Many retail promotions are
driven by trade promotions offered to the retailer by manufacturers and they are described in the
next section.
Trade Promotions
Trade promotions are used by manufacturers to achieve objectives such as generating a
price decrease by retailers to consumers or gaining distribution for a new product.
1
Manufacturers offer retailers trade promotions to stimulate them to offer retail price discounts
and communicate the discounts to consumers. Trade promotions use various price discounting
1
For trade promotions we are going to focus on manufacturer to retailer promotions. There are also trade promotions designed for wholesalers to
offer incentives to retailers.
8
mechanisms and incentives to gain price discounts, displays, in-store communications and space
in the retailers’ advertisements.
Trade promotions have specific objectives and forms of incentives offered to the retailer.
Typical trade promotion objectives are shown in Table 2. The mechanisms by which these
objectives are met are imbedded in the type of trade promotion offered and the requirements
placed on the retailer. The issue of channel coordination is covered in more detail by Kava and
Ozer in the chapter on "Pricing Contracts" in this volume.
Table 2. Typical Trade Promotion Objectives
Gain or maintain distribution
Obtain temporary price discount
Display product
Include product in retailer’s advertisements
Gain market share from competition
Increase sales
Reduce inventory
Sell off old or obsolete inventory
There are a number of types of trade promotions incentives offered by manufacturers.
We cover the most common types: off-invoice, scan-back, accrual funds and slotting fees.
The most widespread trade promotion is off-invoice promotions in which the
manufacturer offers a discount to the channel (e.g., retailers) so that the channel will offer a price
discount to the consumer. This type of promotion has two key issues: pass-through and forward
9
buying. Pass-through is defined as the amount of the discount that the channel passes along to
the consumer in the form of consumer discounts. Forward buying occurs when the retailer buys
more during the promotional period than they intend to sell to consumers. They may either
stockpile the additional inventory to sell in later periods at the regular price, or divert (i.e., resell)
the excess inventory to other retailers. Some or the entire stockpiled product is then sold at
regular prices. This leads to lower profitability of the manufacturer’s trade promotion.
To overcome some of the problems associated with off-invoice promotions,
manufacturers have developed several alternative trade promotions. One is called a scan-back
promotion in which the manufacturer offers the retailer payment for all the units sold during a
specific week at a pre-specified discount to the consumer. It is measured using the retailer’s
scanner (Point-of-Sale or POS) system. The advantage of this type of trade promotion offer is
that the manufacturer pays only for those units sold on promotion and not additional units
purchased by the channel at the discounted price. The disadvantage is that it requires obtaining
the unit movement from the retailer’s POS system and that requires the manufacturer to rely on
the accuracy of the system and the honesty of the retailer. In markets that do not have POS
systems, scan-back promotions cannot be used.
An alternative to scan-back promotions, also designed to limit forward buying and
diverting, is accrual funds which are based on the number of units sold in a prior period to
determine next period’s funding level. Accrual funds generally are allocated jointly by the
retailer and manufacturer to meet certain sales and profit objectives in the relevant spending
period.
For most trade promotions that offer price discounts to the retailers, allowances are
offered for feature advertising (ads by the retailer identifying the price discounts) and display.
10
These funds are generally called cooperative funds and are used to entice the retailer to
communicate the price discount on the manufacturer’s products and to put up special displays in
the retailer’s stores to communicate that the product is available at a discount. Retailers often
have “price lists” in which they indicate to manufacturers exactly how much it costs for a
specific size ad and for a specific display type.
A variant of cooperative funds is for manufacturers and retailers to market and jointly
promote the product. A retailer may decide to run a Fourth of July day special and contacts
manufacturers to receive funds to partially pay for their ad and displays linked to the holiday.
Slotting fees are another form of trade promotion with a very different objective – gain
distribution for a new product or additional SKU’s. The retailer charges the manufacturer a
slotting fee to cover the cost of stocking and managing a manufacturer’s new product
introduction. The original purpose of slotting fees was to cover the retailers’ new product
introductory costs such as removing products from the shelf and replacing them with the new
items, reclamation costs of the old items that were removed, the inventory risk of carrying new
items if they do not sell and the warehousing and administrative costs of adding new items.
Slotting fees are very controversial and have been studied by regulatory agencies who argue in
some cases that slotting fees limits competition from smaller manufacturers and for smaller
brands being introduced. Table 3 provides a more comprehensive list of trade promotion types.
Table 3. Trade Promotion Vehicles
Off-invoice Discounts offered from the invoice price on all units sold to the
retailer over a specified time period.
Accrual Funds Funds paid by the manufacturer to the retailer based on the prior
11
period’s unit sales movement.
Scan-back Discounts offered based on units sold through the point-of-sale
register rather than on units purchased by the retailer.
Count-Recount System used to pay retailers based only on units sold. Similar to scan-
back promotions except the manufacturer does the counting in the
retailer’s (or wholesaler’s) warehouse. Used when POS data are not
readily available.
Co-op Advertising Manufacturer funds to support retailer advertising for the
manufacturer’s product(s).
Display Allowances Manufacturer funds to support displays put up by the retailer.
Bill Backs Similar to off-invoice funds except the retailer must provide proof that
specific actions have taken place such as an ad was run by the retailer
containing the manufacturer’s products. Used to ensure compliance
by the retailer.
Slotting Fees Manufacturer funds to cover the retailer’s costs of new product
introductions.
Free Goods Extra cases offered to the retailer by the manufacturer often for new
products to induce the retailer to stock the items for which the free
goods are offered.
Floor Plan Financing offered by the manufacturer to retailers or dealers to cover
the financial inventory carrying cost of the retailer or dealer.
12
Financial Terms Terms to provide incentives to retailers to stock items and not have to
carry the financial cost of inventory.
Consumer Promotions
Consumer promotions are promotions from a manufacturer directly to consumers. In the
US the most common forms are rebates and coupons. In other countries, contests and
sweepstakes are more prevalent. More and more consumer promotions are being offered on the
Internet.
The purpose of consumer promotions is for a manufacturer to communicate a discount
directly to the consumer and avoid intermediaries (e.g., retailers) who may not provide the
discount the manufacturer wants. In the design of consumer promotions, the critical decisions
are medium, redemption system, restrictions and breakage. The medium used varies by type of
consumer promotion. Print and the Internet are very common media used for consumer
promotion because the promotion can be printed. Handouts, on-pack and in-pack promotions are
also used.
The redemption system is very important to avoid fraud. Coupons in the US are managed
through a complex clearing house system. Rebates are often sent to clearing houses as well. The
difficulty is that retailers can misredeem the discount delivery vehicle (e.g., coupon) without a
redemption, i.e., submit coupons for redemption that were not submitted by consumers with
purchases. In the US, misredemption of coupons is a serious problem. Misredemption and fraud
are reasons that consumer promotions can be ineffective. Many consumer promotions have
restrictions to limit long-term liability. Statements such as “good until June 30, 2010” restrict
13
the term of the promotion. Restrictions can also take the form of “only one per customer”
though those are very difficult to enforce.
One of the advantages of consumer promotions is breakage, which is the difference
between the number used to induce a purchase and the number redeemed. Breakage is very
common for product rebates. Consumers need to send in proof-of-purchase (receipt) and the
rebate document to a specific address. Many consumers will purchase based on the rebated price
(retail price minus the rebate) but fail to send in the rebate because of the time required or the
loss of the purchase receipt or rebate form. Breakage can increase the profitability of consumer
promotions because it results in a higher net price.
Consumer promotions have significant advantages and disadvantages relative to
trade/retailer promotions. The primary advantage is that the manufacturer has control over the
offer received by the consumer. When a promotion is offered through the retailer, generally the
manufacturer does not have control over the price being offered to the consumer, though this
varies by country. The primary disadvantages are the low redemption rates (free-standing insert
coupons in the US average about 1.5 to 2.5% redemption rates) and the cost of distributing the
offer.
Internet Promotions
Retailers and manufacturers can use the Internet as a vehicle for targeting and reaching
customers with promotions. Unlike direct mail, the Internet is a virtually zero-cost
communication vehicle. If a customer is willing to provide his or her e-mail address, then the
firm selling the goods or services can reach the customer at a low-cost. Offering highly targeted
promotions which were very expensive using mail or other distribution systems becomes almost
14
costless when using the Internet. The other important method of the distribution of discounts
using the Internet is websites. Many manufacturers or third-party sites offer consumers
discounts on purchases of products. Consumers can print coupons, use codes or other
mechanisms to obtain discounts.
The implication of a low-cost communication vehicle for offering targeted promotions
combined with a wealth of consumer information available on the internet is that the types of
promotions will be much more selective. This means a critical research area is to determine the
value of targeted promotions versus mass promotions (see e.g., Rossi et al., 1995). Models are
also needed to determine what types of offers should be provided to different segments of the
market. Whether firms will ever be able to offer one-on-one promotions is an open question
because of the cost and sophistication required to provide the relevant analytics.
One issue that emerges with promotional targeting is “fairness.” If one consumer
receives a different promotion than another and each learns about the others promotional offer,
the consumer receiving the less favorable promotion is likely to be very upset and may stop
purchasing from the company. Thus, the type of promotional offers used to target individual
consumers is very important. For example, with frequent flyer programs, the flyer has to earn
the benefits received. For frequent flyer and similar programs (e.g., hotels) different customers
receive different offers based on “earning” the reward. For further discussion of the consumers'
perception of fairness, see the chapter by Ozer and Zheng (in this volume)..
15
Behavioral Underpinnings of Sales Promotions
There is a stream of research that is important in understanding why promotions exist –
the psychological marketing literature. Three foundational articles provide the basis for this
stream of research and provide relevant managerial implications.
Smart Shopper
Schindler (1998) observes that price promotions are able to generate consumer responses
that are far greater than the economic value of the money saved. In other words, one can look at
the difference between a promotional price reduction and a regular price reduction and the
response to the promotion, beyond just the temporal nature of the promotion, is far greater than
the strict price reduction effect. Schindler posits the effect as consumers’ perceiving themselves
as “efficient, effective and smart shoppers.” He argues that because consumers feel that their
actions are perceived to be responsible for obtaining discounts, this will increase the
noneconomic component of the discount.
Transactional Utility
A related explanation is offered by Thaler (1985) who introduced the concept of
transaction utility. Transaction utility is defined as the gain (loss) of utility when the consumer
pays less (more) than the reference price of the product. The reference price is not the same as
the actual price of the product but the price the consumer believes the good is worth or the price
the consumer would expect to pay for the good. Positive transaction utility occurs when the
reference price is above the actual price paid. If promotions produce transactional utility, then
the firm is better off using promotions than simply lowering regular price. The outcome will be
a higher response to the promotion than to an equivalent reduction in everyday regular price.
16
One of the key differences between Schindler’s and Thaler’s theories is the notion of reference
price, or the perceived value of the good.
Reference Price
One of the earliest studies of reference price was Winer (1986) who defined reference
price as a function of past brand and category prices and estimated an empirical choice model
that demonstrated that the reference price effect could be measured. Reference price effects have
been reliably found in experimental data (Niedrich et al., 2001; Jainiszeqski and Lichtenstein,
1999; Van Ittersum et al., 2005; Chernev, 2006). Some research suggests that consumers use
both internal (memory-based) as well as external (stimulus-based) reference prices (Mazumdar
and Papatla, 2000). Howard and Kerin’s (2006) findings suggest that reference prices may be
context specific, so different internal reference prices may be invoked when a product is
advertised in the retailer’s ad compared to when it is not.
These foundational articles also explain the finding that more frequent promotions reduce
reference price (Blattberg et al., 1995). This change in reference price, in turn, can reduce the
incremental sales associated with the promotion (Nijs et al., 2001; Krishna et al., 2002).
However, Srinivasan et al (2004) find some evidence against this generalization. They find that
manufacturer revenue elasticities are higher for frequently-promoted products and retailer
revenue elasticities are higher for brands with frequent and shallow promotion, for impulse
products, and in categories with a low degree of brand proliferation. This latter finding may be
consistent with the results of Alba et al. (1999) who find that the depth of promotion dominates
the frequency of promotion in the formation of reference prices when prices have a simple
bimodal distribution.
17
The key insight from this discussion of reference price is that heavily promoted products
tend to lose “brand equity” through a decrease in the amount a consumer is willing to pay for the
product (Winer, 2005). One interesting qualification is that when the variance of the deal
discount increases, the reference price also increases, even though the average deal discount
remained constant (Krishna and Johar, 1996). A managerial implication is that brands should
have multiple discount levels and sellers should vary the levels of the discount so consumers are
less able to predict future savings.
When we combine the various behavioral theories, we begin to understand that
promotions behave differently than long-term price reductions. Further, the frequency of
promotions can affect reference price (lower it) which can make promotions less effective.
Managers must consider the frequency and timing of their promotions and their impact on
promotional effectiveness.
How Promotions Affect Sales
There has never been debate about the foundational finding in the literature – sales
promotions are associated with large increases in consumer sales. The question this raises is
“what are the sources of incremental volume?”
In general, the sources of volume from a sales promotion come from one or more of the
following sources: 1) customers switching their purchases from other brands (brand switching),
2) current consumers purchasing more quantity of the brand for inventory (stockpiling), 3)
current consumers accelerating their purchase of the good (purchase acceleration), and 4) new
consumers entering the market (primary demand expansion – also called category expansion).
18
Brand Switching
Earlier studies show a very high percentage of incremental promotional volume comes
from switching. For example, Gupta (1988) showed that 84 percent of incremental sales was
from switching. Other estimates have ranged from 43.8% to 93.9% (Bell et al., 1999) .
More recently in a study across many categories, Van Heerde et al. (2003) found that the
percentage of the incremental volume attributable to brand switchers was only about one-third of
the total incremental volume. They show that the difference in the percentage in their study
versus previous studies arises from the fact that previous studies had not accounted for category
growth in their calculations and hence overestimated the percentage attributed to brand
switching. Van Heerde et al’s findings have been supported by another study using store-level
data (Pauwels et al., 2002).
For manufacturers, incremental volume coming from brand switching can be highly
profitable because it is volume that the brand would not otherwise have. However, for retailers,
incremental promotional volume from brand switching may or may not be profitable depending
upon the brands it comes from and their profitability. Because brand does not expand category
volume, it is far less advantageous to a retailer than to a manufacturer.
Purchase Acceleration and Stockpiling
When a promotion is run, consumers can react by changing their purchase timing by
purchasing earlier than they normally would (purchase acceleration) and/or by purchasing more
units than they would normally purchase (stockpiling). Managers generally believe that
purchase acceleration and stockpiling do not expand demand and are detrimental to both retailers
and manufacturers. However, in categories in which consumption can expand due to product
19
availability (e.g., candy), increasing household’s inventory of the product or moving their
purchase forward can increase consumption and long-run sales. Purchase acceleration for
services such as oil changes can also increase long-term sales. Studies have found that
promotions can increase consumption in categories where the consumption rate is related to the
amount of the product consumers have in their pantry (Ailawadi and Neslin, 1998; Nijs et al.,
2001), especially when the promotions involve strong brands (Sun, 2005). For instance, two
categories that exhibit this pantry effect are carbonated beverages and ice cream. Bell et al.
(1999) quantified the proportion of the increase in sales due to a promotion that is attributable to
purchase acceleration and increases in purchase quantity and found it ranged from 0.7 to 42.3%,
with an average of 10.6% across 13 categories.
Because purchase acceleration and stockpiling shift consumer purchases, we would
expect to see a post-promotion dip after a promotion. Figure 3 shows the expected pattern with a
promotion in week three. The ‘trough’ or dip after the promotion occurs in weeks four and five,
and is indicated by the arrows below the normal or base-line sales. During this trough, the
additional units purchased during the promotion are consumed.
==== put figure 3 about here =====
It is also possible to have a pre-promotion dip. In Figure 3, the sales in week two would
then be less than the average due to consumers anticipating the promotion. Van Heerde et al
(2000) found evidence of pre-promotion dips as well as post-promotion dips.
Blattberg, et al (1995) summarized the literature and stated that post-promotional dips are
rarely seen in store-level data. This seems paradoxical, as consumers should be consuming their
inventory after a promotion instead of purchasing additional quantities. Subsequently, several
papers investigated this issue and found either strong evidence of post-promotional dips (Van
20
Heerde et al., 2000) or qualified support of post-promotional dips (Mace and Neslin, 2004). In
the latter case, Mace and Neslin (2004) find that the promotional dips are related both to
brand/UPC characteristics (high-priced, high-share, frequently promoted) and store trading area
demographics (older customers, larger households). One clear implication of this finding is that
inter-temporal substitution (i.e., quantity acceleration) may be overstated in the deal
decomposition studies and promotions can be more profitable if this effect is taken into account.
Category Expansion
The issue of market-level category expansion due to promotions is extremely important
because it benefits both retailers and manufacturers. The problem is that it is very difficult to
measure, partly because of the data requirements and partly because of the complexity of the
factors that need to be controlled such as store switching, brand switching and purchase timing
effects.
In general the literature has found no long-run effect of promotions on category volume,
although short-term effects do exist (Ailawadi and Neslin, 1998; Nijs et al., 2001). In a recent
study of the effects of promotions at CVS Drug Stores, Ailawadi et al. (2007) find that 45% of
the increase in sales due to promotions is attributable to category expansion effects. However,
this may just be store switching rather than category expansion. While the literature does not
find long-term category expansion effects, much more work needs to be conducted on this topic.
It is a fundamental issue in determining the economic return from promotions.
How Different Sources of Volume Payout for Retailers and Manufacturers
There has been limited research into the profitability of promotions, in part, due to the
lack of available cost data. However, several papers have studied the impact of promotions on
21
revenues and profits. Srinivasan et al (2004) find that promotions generally are profitable for
manufacturers, but are not beneficial to retailers even when cross-category and store-traffic
effects are considered. Similarly, in a study of promotions at CVS, Ailawadi et al. (2007) find
that more than 50% of the promotions are not profitable. Further, in a field test, they determined
that CVS could increase their profitability by $52 million if they eliminated promotions in the 15
worst performing categories. A way to summarize the results above is to see how different
sources of volume affect manufacturers' and retailers' profitability as shown in Table 4.
Table 4 – Effects of Source of Incremental Volume on Profitability of Promotions
Source Manufacturer Retailer
Brand Switching Highly profitable if increased
quantity covers cost of
promotion.
Profitable if consumers
purchase higher margin item
Purchase Acceleration
and/or Stockpiling
Mostly unprofitable. Can be
profitable if stockpiling
increases consumption or takes
purchases from competitors.
Mostly unprofitable. Can be
profitable if increase future
demand or consumption.
Category Expansion Profitable if incremental
volume covers cost of
promotion.
Generally profitable.
Manchanda (1999) found large effects in complementary categories, and one study
quantified these effects as 23-67% of profits and 33-40% of quantity in complementary
categories are due to promotions (Niraj et al., 2008). However, these effects appear to be
22
related to the categories and retailer brands, but not to national brands. Specifically, these effects
do not appear to persist across categories for national brands (Russell and Kamakura, 1997).
As noted above, one key objective of promotions is to drive incremental customers to the
store, which, if true, would have strong impact on the profitability of promotions. Practitioners
believe that a large percentage of store traffic is due to feature advertising. Feature advertising is
defined as print media advertising by the retailer that highlights a set of products offered on
promotion during the current week. While recent results do seem to indicate that some
consumers respond to feature advertising (Briesch et al., 2008a; Srinivasan and Bodapati, 2006),
they may be “cherry-picking” the featured items (i.e., only buying these featured items) which
can reduce the profitability of the promotions (Fox and Hoch, 2005; Rhee and Bell, 2002). An
intriguing finding is that assortment and distance are more important factors than price and
promotions in store-choice decisions (Briesch et al., 2008a).
In general there is limited empirical research showing what causes promotions to be
profitable and when promotions are profitable. The presence of retailer and manufacturer
promotions would suggest they are profitable; otherwise they would stop using them. However,
there may forces other than profitability driving the use of promotions. The next section
discusses some of these issues and the strategy of promotions.
Promotional Strategies
This section looks at how retailers and manufacturers design promotional strategies. We
begin with retailers and then turn to manufacturers.
Retailer Promotional Strategies
23
Many retailers use a merchandising system called "category management" designed to
assist them develop pricing, promotions, item selection, space allocation, displays and retailer
advertising tactics. A category is defined as a set of products which consumers perceive to be
close consumption substitutes. For example, peanut butter and jelly are in separate categories
because they are substitutes, but not close substitutes. However, Jif and Peter Pan peanut butter
are within the same category.
Some retailers use tactical guidelines based on the category roles to set their promotional
strategies. Examples of category roles are "traffic" or "flagship." When the category role is to
generate traffic (i.e., bring customers into the retailer’s store), then promotions are likely to be
deep, frequent and less profitable for the category. The retailer is not setting promotional
profitability as its objective but rather the power of the category to draw customers into the store.
An example of a traffic generating category in grocery stores is carbonated beverages.
For categories classified as flagship (large and profitable categories), the retailer will still
promote but the depth of the promotional discount is lower and promotions will be evaluated
based on how profitable they are. An example of a flagship category win grocery stores is
cookies and crackers.
How retailers promote categories influences the manufacturer’s payout from their trade
promotions. For categories with low volume and limited importance for the retailer, trade
promotions will likely be less effective. For example, shoe polish is probably never or
infrequently promoted by the retailer. Why should a manufacturer offer a trade promotion in that
category? It is unlikely to have a positive payout. Thus, in designing a trade promotion strategy,
a manufacturer must understand the retailer’s category strategy.
24
Another strategic issue facing retailers is which brands to promote within a category.
Retailers consider the brand’s market share and consumer purchase behavior in deciding which
brands to promote. Strong brands with high brand equity are likely to be promoted aggressively
by the retailer. Weak brands, even in traffic generating categories, are likely to receive less
promotional push by the retailer. Again in assessing its trade promotion strategy, the
manufacturer must understand how the retailer is likely to promote its brand and why.
Manufacturers Promotions as a Competitive Tool
In considering how promotions are used by manufacturers as a competitive tool, we have
identified two key strategic approaches. First, when manufacturers are battling each other
through the use of promotions, an important consideration is whether the manufacturers are in a
prisoner’s dilemma. Secondly, manufacturers can use promotions to limit private label
encroachment into a category.
Prisoner’s Dilemma
Probably the most common strategic issue facing manufacturers is determining the likely
competitive response to its promotion. Generally, a manufacturer should assume that a
competitor is likely to respond and match its promotion. A simple way to capture the likely
outcome is to create a two-by-two payoff matrix as shown in Table 5. For an explanation of
such payoff matrices, see the chapter on game theory models in pricing by Kopalle and Shumsky
in this volume. The matrix can have special properties depending upon what happens in the
industry. We will concentrate on the payout when both manufacturers either promote or do not
promote.
25
Table 5 – Payoff Tables
(a) Higher Payoff for firms with Promotions
Manufacturer A
Do Not Promote Promote
Manufacturer B
Do Not Promote 8/8 12/5
Promote 5/12 10/10
Notes: Values in cells represent profits for manufacturer A/manufacturer B
(b) Lower Payoff for firms with Promotions
Manufacturer A
Do Not Promote Promote
Manufacturer B
Do Not Promote 10/10 12/5
Promote 5/12 8/8
Notes: Values in cells represent profits for manufacturer A/manufacturer B.
If the payout is higher for both firms when each competitor promotes than if they do not
promote, then promoting can benefit both competitors. This is shown in Table 5a. This is a
rarely discussed outcome in the marketing promotional literature because it is believed that
promotions do not expand long-term total demand. However, as we discussed earlier, if
promotions have a psychological advantage and cause consumers to have a greater response than
a price reduction, then promotions can expand the market and consumption. Thus, the outcome
that both firms are better off can definitely occur in the real-world.
The more common view in the marketing literature is that the outcomes on the diagonals
are as shown in Table 5b in which there is a higher payout when neither firm promotes than
when they both promote. This is the classic prisoner’s dilemma. Both firms promote because
26
they are afraid their competitor will promote. Thus, promotions actually decrease profitability in
the market and both firms are worse off.
In developing a promotional strategy it is extremely valuable to study what the likely
payout matrix is and whether promotions actually increase “industry” profits. Unfortunately,
little research has studied the actual frequency of the two cases described above in Table 5a and
5b. Procter and Gamble (P&G) tried an on off-diagonal strategy, characterized by what they
called "value pricing," designed to lower prices and reduce offering trade promotions. The
strategy failed miserably because their competitors continued to promote and the outcome was a
decrease in share for P&G. Ultimately, P&G reintroduced trade promotions on their brands.
Competitive Encroachment
A number of papers (e.g., Lal, 1990; Rao, 1991) explore the extent to which price
promotions are a way to limit competitive encroachment. Specifically, national brands (higher
quality brands) will promote to reduce encroachment by private label brands (lower quality
brands). Rao assumes that there are two firms in the market with one being a national brand and
one being a private label brand. There are two segments of consumers: segment A consists of
shoppers with no brand preference and purchase the lowest price brand; segment B shoppers rely
on both price and brand preference. Segment B shoppers only purchase if the lower-quality
brand has a large enough discount relative to the national brand. Given this setup, the
manufacturer sets a single regular price and determines the number of promotions and the depth
of the promotions. Rao shows that the strategies the two firms use are: 1) the national brand sets
its regular price directed to its natural franchise (segment A) and 2) the private label sets its
regular price to reduce the promotional encroachment by the national brand (segment B). The
27
role of promotions is to “enforce the regular prices” and to decrease the incentive of the private
label to reduce its price too much.
Lal (1990) makes a slightly different argument for price promotions. He assumes that
there are three brands competing in the market - two national brands and one private label brand.
Each national brand has a loyal segment, where loyal consumers will purchase their favorite
brand as long as the price is below their reservation price. The private label brand does not have
a loyal segment. Finally, there is a ‘switching segment’ which purchases the brands based upon
their relative prices. Lal shows that, under these assumptions, promotions are a way for the
national brands to implicitly collude against the private label brand to limit its market share. He
finds some empirical support for this theory.
Using a game-theoretic approach Agrawal (1996) examines the optimal advertising and
promotional strategies when two manufacturers sell their products through a common retailer.
Agrawal finds that advertising can be viewed as a defensive strategy to retain a loyal customer
base, whereas promotions (i.e., trade deals) are an offensive strategy to attract competitors’ loyal
customers. In this case, it is optimal for a stronger brand to take an offensive strategy and offer
more trade deals, whereas a weaker brand should adopt a defensive strategy. Additionally,
Agrawal finds that the retailer should promote the stronger brand more frequently but with
smaller discounts than the weaker brand.
All of these papers implicitly assume that cross-promotional effects are asymmetric. This
assumption – higher quality/price brands are less vulnerable to price increases and gain more in
category incidence and choice for price decreases than lower quality/price brands – has been
supported in the empirical literature (Sivakumar and Raj, 1997). However, it has been qualified
in three important ways: 1) the effect depends not only on price/quality tier, but on brand
28
positioning advantage (over/under priced relative to competition) (Bronnenberg and Wathieu,
1996); 2) promotions can diminish differentiation of high-tier brands, which would tend to
diminish the asymmetric effects (Heath et al., 2000); and 3) when absolute effects (i.e., the
change of share of the competing brand (not percentage) divided by change in price (not
percentage)) are considered instead of elasticities, the asymmetric relationship changes with
smaller brands having an advantage over the larger brands (Sethuraman et al., 1999; Sethuraman
and Srinivasan, 2002). This area has high potential for future research because it examines the
competition between national and store brands, which has grown in importance (Sethuraman,
2009).
Manufacturers/Retailers use Promotions as a Price Discrimination Mechanism
Even if we ignore the strong consumer response to promotions, several articles have
argued that promotions can exist because they are a price discrimination tool. Specifically,
promotions allow manufacturers and retailers to charge different prices different consumers. In
this section, we discuss several ways that promotions can be used to discriminate between
consumers.
Informed versus Uninformed Consumers
Under the assumption that there are two types of consumers – informed and uninformed –
it is optimal for retailers to offer promotions so that the uninformed consumers do not learn
which store offers the lowest prices (Varian, 1980). Informed consumers are assumed to know
the prices at all stores in a given period; whereas uninformed consumers select stores at random.
If the store does not randomize (promote) its prices, then the uninformed consumers would learn
which store has the lowest price after several shopping trips and only visit the store with the
29
lowest prices. Therefore, by randomizing (offering promotions), the firm can increase its profits
(and prices) so the uninformed consumers never know which store has the lowest prices.
Value of Time
A second explanation for why manufacturers offer coupons is based on heterogeneity in
the value of time across consumers (Narasimhan, 1988). The general idea is that a consumer is
maximizing his or her utility which consists of consuming the good and ‘consuming’ leisure time
(called L). The consumer must spend a certain amount of time in order to clip and use the
coupon, and saves S by using the coupon. The higher the value of S, the more of the good the
consumer can use and hence the higher the consumer’s utility.
However, the time to clip and use the coupon comes from the time devoted to leisure.
Thus the trade-off the consumer is making is reducing L, leisure time, which has some utility to
the consumer and using it to redeem coupons which provides a savings. Because different
consumers have different costs associated with L, firms can price discriminate between
consumers with different value of time using coupons.
Intertemporal Discrimination
Intertemporal discrimination has many forms and is covered more fully in the chapters on
dynamic pricing and overstock pricing in this book (Aviv and Vulcano, 2010; Ramakrishnan,
2010). However, we include a brief description and example for completeness. Assume that
there are two segments of consumers: 1) consumers who cannot stockpile a good (i.e., their
quantity purchase is limited to a single period), and 2) consumers who can stockpile for multiple
periods.
30
When a promotion is offered, both segments can take advantage of the price discount.
However, the segment which can stockpile the good, purchases enough of the good for multiple
periods; whereas the other segment only purchases for one period. Therefore, in the second
period when there is no discount, the segment which cannot stockpile pays a higher price for the
good. The same logic applies to households with different consumption rates, where the
consumption rate determines the number of periods in which the household can stockpile the
good.
Trade Deals Increase Efficiency
Many marketing managers worry about the efficiency of the distribution channels.
Generally, an efficient channel has the goal of all members aligned so that they are all working to
maximize the system profits. Implicit in this definition is the notion that the actions of the
channel members are coordinated. While the issue of channel contracts is covered in more detail
in the chapter on Pricing Contracts in this book (Kaya and Ozer, 2010), we briefly cover the
issue here for completeness. The problem can be viewed as one where the manufacturer and
retailer have different objectives within a category for a market. For instance, as a manufacturer,
Coca-Cola would like all carbonated beverage sales to be one of their products. However, the
retailer would like to sell all of the carbonated products in the market, regardless of the
manufacturer (ignoring margin considerations). Channel coordination and efficiency
mechanisms align the retailer’s goals and actions with the manufacturer’s goals, in this case
Coca-Cola.
Dreze and Bell (2003) examine scan-back promotions and compare them to off-invoice
promotions and show that retailers strictly prefer off-invoice promotions to scan-back
promotions. However, they also show that manufacturers can design scan-back promotions so
31
that the retailer is indifferent between off-invoice and scan-back promotions. The result is that
the manufacturer is better off, and the manufacturer can use scan-back promotions to align the
retailer’s actions with the manufacturer’s profit goals.
Taylor (2002) examines a more general situation under which the manufacturer enters
into a contract with the retailer where the manufacturer provides rebates to the retailer on the
units sold beyond a target level. He finds that when retailer actions do not influence consumer
demand, this contract is sufficient for channel coordination. However, when retailer actions do
influence consumer demand, this target rebate contract does not coordinate the channel. To align
the retailer's and manufacturer's objectives, the manufacturer must also allow the retailer to
return unsold units.
A second trade dealing mechanism which can impact efficiency is slotting fees. While
there has been a limited amount of analytical research into slotting fees, the key question is: “Do
slotting fees increase efficiency in the marketplace or do they hinder competition in the market
place?” (Bloom et al., 2000). The argument for hindering competition is that slotting fees
greatly increase the costs of introducing new products and may make it economically infeasible
for small and medium firms to gain widespread distribution of their products. That said, new
product introductions are inherently risky with one in three launched products unsuccessful
(Cooper, 2001). Therefore, one reason that slotting fees may increase market efficiency is that
slotting fees transfer some of the risk of new product introductions from the retailer to the
manufacturer by covering some of the potential lost opportunity costs and administrative and
operating costs associated with the failed introduction.
Aydin and Hausman (2009) analytically examine this situation. Specifically, they
assume there is some demand uncertainty which causes the retailer, but not the manufacturer, to
32
incur inventory holding costs. They show that the manufacturer may offer to pay slotting fees to
the retailer to ease the inventory costs so that the retailer will choose a larger assortment.
Therefore, slotting fees can provide channel coordination in terms of providing optimal
assortment to the consumer.
The empirical results addressing this issue are mixed. Sudhir and Rao (2006) find that
slotting fees tend to increase the efficiency in the marketplace, whereas Bloom et al (2000) find
that while slotting fees are a method to shift the risk of new product introductions to
manufacturers, they are also applied in a discriminatory manner against manufacturers. Bloom
et al.’s results tend to support the finding that slotting fees may hinder competition by making it
difficult for smaller manufacturers to obtain widespread distribution.
Empirical Models of Promotions
One of the reasons for the growth in promotional activity in the 1980’s was the ability to
measure the impact of promotions on sales. This section reviews the data available to model
promotions, provides a description of the basic models and their purpose and ends with some of
the more advanced issues, procedures and rationale for addressing them.
Data to Model Promotions
There are two types of data critical for modeling sales promotions: sales and causal
promotional data. The data sales data available for modeling promotions cover three types: 1)
point-of-sale (POS) obtained from retailer’s scanners, 2) panel data from a sample of consumers
and 3) customer purchase data. The other required data are causal data which includes
promotional discounts, presence of feature advertising and displays.
POS and Causal Data
33
POS data are currently available for most consumer packaged goods categories in the US
and other developed countries because Nielsen and IRI collect both the POS sales data and
causal data. For categories other than packaged goods (e.g., durable goods), the problem for
manufacturers is the availability of data both at the POS level and causal data because very few
third-parties collect and sell the data.
Retailers have the potential to have both POS and causal data for any type of business as
long as they have scanners. However, in order to develop promotional models, they must collect
and maintain the causal data which they often do not do. For retailers POS data provides much
of the sales information that they need with the exception being individual purchase behavior.
The disadvantage is retailers need to create causal databases and maintain them which, as stated
earlier, many do not.
For manufacturers, the advantages of POS data are accuracy, ease of modeling, and wide
availability through third-party data collection sources for some categories such as consumer
packaged-goods categories. The disadvantages are that consumer behavior cannot be modeled
directly, and with some exceptions, categories outside consumer packaged goods are not readily
available to manufacturers and causal data are not always maintained.
Panel Data
The second source of data is panel data collected from consumers who shop a category or
groups of categories. A sample of consumers is chosen to be part of the panel. All of their
purchases are recorded by the panel member and maintained in a database. The problem is the
causal data. Because each consumer shops at a different set of stores, there is no easy way to
maintain and collect all of the causal data from all of the competing stores.
34
Panel data provide advantages to manufacturers because, in many categories, panel data
can be collected when retailers are not willing to provide POS data. The problems with panel
data are causal data and sample size limitations. Many panels are too small to capture accurate
data at a local market level where promotional competition is occurring.
Purchase Data
Purchase data maintained by Internet retailers and direct-to-consumer manufacturers can
be highly accurate and provide significant insight into consumer’s purchase behavior and
sensitivity to promotions. Purchase data is by far the best source of sales data. If the firm keeps
accurate promotional history including non-responses to promotions, then the firm has
potentially a highly accurate, detailed dataset which can be used to model individual promotional
response behavior.
POS Promotional Models
POS promotional models have been utilized since the 80’s by both academics and
practitioners (see, e.g., Blattberg and Neslin, 1990 for a summary). The basic model used is:
tttttt
eDISPFADPS +++++=
4321
ln
ββββα
where S
t
= unit sales at time t, P
t
is the regular price at time t, D
t
is the promotional discount at
time t, FA
t
is feature advertising (0,1) at time t, and DISP
t
(0,1) is presence of a display at time t
and e
t
is the error term at time t. By separating price from promotional discount, the modeler is
assuming that promotions have a differential effect (usually assumed to be greater) from a price
reduction.
This basic model has been revised and restructured in many ways both in terms of inputs
such as by type of display and feature advertising, transforming it into a share model, and using
35
reference prices and their variants. One key addition is capturing the post-promotion dip
discussed earlier. By adding lagged promotion variables, dummy variables or other indicator
variables, the modeler can then estimate the total impact of promotions (Neslin, 2002; Van
Heerde et al., 2002; van Heerde and Neslin, 2008).
Using POS sales data from retailers and causal data from data suppliers such as Nielsen,
the model’s parameters are estimated. Once completed the firm can estimate sales with and
without promotions, what impact different promotional discount levels have on sales, and how
much feature ads and displays increase sales. A manufacturer can then optimize its trade
promotion offer to retailers based on the likely impact it will have on both sales and profits.
Several issues arise with this basic model in practice. First, the likelihood of
multicollinearity that results from feature advertising, displays and price discounts being offered
simultaneously. This has posed problems to modelers for years and no simple solution exists.
Only through experimentation and detailed store-level data can better estimates be obtained. For
practitioners it is critical to test for multicollinearity to ensure that the model estimates for
feature ad, display and price discount have the appropriate precision.
A second issue is determining how to incorporate the promotional discount. If one uses a
semi-log model, the effect of a promotion accelerates with the size of the discount, implying that
smaller discounts will have limited sales effect. Some attempts have been made to incorporate
S-shaped curves in modeling the promotional discount, implying a ceiling on sales after some
promotional level is reached. This may not seem plausible but at some point consumers reach a
saturation level in the quantity they can purchase. Other functional forms can be used depending
upon the expected promotional effect being modeled. Should the discount be captured using a
percentage or an absolute level? Little evidence is available to guide this decision.
36
Another issue is how to link the retail promotional model with trade promotions so that
manufacturers can maximize their trade promotions. If the trade promotion offers a specified
amount for a feature ad, does the retailer offer the manufacturer a feature ad? What is the
relationship between manufacturer spending and receiving ads and discounts from the retailer?
The requirement is to create a “linking” model which shows how various types of trade deals
lead to specific retailer promotional actions which then lead to increases in sales.
Fourth is modeling category sales to estimate how category sales increase and which
brands gain and lose sales when a given item is promoted. Many of the models in the marketing
literature concentrate on brands or SKU’s. However, retailers concentrate on brand and SKU
impact, category sales and category profit increases. Within a category certain brands are more
likely to respond to promotions, will have a greater impact on category expansion and will
cannibalize the sales of other brands including private label. Retailers need to understand the
category dynamics of promotions to design their promotional strategies.
Fifth is determining the appropriate aggregation level. There is a major question in the
real-world regarding the use of store-level data versus retailer-level or market-level data.
Generally it is believed that using store-level data is better but much of the available level data
has been aggregated to the retailer level.
In summary, the basic POS model and its variants have been a “workhorse” used by
researchers, practitioners and consultants to study the effects of promotions on sales. It has lead
to many manufacturers and retailers improving the effectiveness of their sales promotion tactics.
Promotional Models Using Panel Data
37
POS promotional models provide managers with the level of incremental sales but offer
limited insight into the sources of volume. Promotional models using individual purchase
history data can provide insights into how consumers are responding to promotions.
The types of promotional questions models can answer using individual purchase data
are: 1) what are the sources of incremental volume? 2) which types of consumers are
responding to promotions such as non-brand purchasers? 3) do promotions induce consumers to
purchase resulting in future purchase loyalty? 4) is volume coming from competitive retailers or
merely from the retailers current customers?
For almost three decades marketing academics have modeled promotions using panel
data starting with Gaudagni and Little’s (1983) path breaking article using a logit model to
capture promotional effects. Later, Gupta (1988) used various models to decompose the
promotional spike into its components.
One of the first promotional models was developed by Gupta (1988). We will not cover
his model explicitly but will show the general form of the model. Specifically, define

as
the unit sales in store s (s=1..S) in period t (t=1..T) for item j (j=1..J)
2
in category c (c=1..C).
Then Gupta's model can be written as:

= 

= 1


= 1|



= 1|

,

q


where H is the number of households, 

= 1
is the probability that store s is visited by
household h in period t (i.e., where to shop?), 

= 1|

is the probability that
2
We use the term ‘item’ here instead of brand, stock-keeping unit (SKU) or even brand-size as these different types
of items imply different aggregation levels. Aggregation of SKUs is discussed in the section on data issues below.
38
household h purchased in category c, conditioned on visiting store s (i.e., when to buy?),


= 1|

,

is the probability that item j was selected by household h, conditioned on
being in store s and purchasing in category c, and q

is the quantity that the household
purchased conditioned on selecting brand b in category c in store s in period t (i.e., how much to
buy?). Models are then created for each component of the above equation.
The probabilities in the above equations are all modeled through “utility” functions,
where the utility function consists of a “deterministic” component (i.e., observed variables like
price and feature advertising) and a “random” component (or “error term” as in the sales models
above). In these types of models, it is assumed that a consumer is most likely to select the
alternative with the highest utility. Therefore, the probabilities associated with the alternatives
are proportional to their relative utilities. The specific assumptions made about the random
component then define the form of the model. For instance, when the random component is
assumed to have a normal distribution, a probit model results; and when the random component
is assumed to have a Gumbel distribution, a logit model results. Both of these models are
estimated through Maximum Likelihood techniques.
Quantity Model
Mela, et al (1998) provide a quantity model for household purchases. Their model has a
form similar to:
(
)
sjthjthhjthhthsjthsjthsjthhjhsjt
wloyinvdDfq
β
β
β
β
β
β
β
+
+
+
+
+
+
+
=
6543210
)1(ln
where and D
sjt
(0,1) is presence of a display for brand j at time t in store s, d
sjt
is the promotional
discount percentage for brand j in period t at store s, and inv
ht
represents the household’s
inventory of the category in period t, loy
hjt
represents household h’s loyalty to brand j in period t,
39
w
hjt
is a selection correction term (similar to a tobit model) because this model is conditioned on
choice and incidence, and ε
sjt
is the error term. While Mela, et al. used a constructed inventory
variable, more recently some authors have argued that this constructed variable can introduce
bias into the model and as an alternative have suggested using three observed variables (lag
quantity, time since last purchase, and their interaction) instead (Erdem et al., 2003; Hendel and
Nevo, 2006). Finally, consumer loyalty is included in the model because the literature has
shown that consumers may purchase more of their favorite brand when it is on promotion (c.f.
Mela, et al), for which they used a smoothed brand loyalty term (Guadagni and Little, 1983).
Purchase Incidence Model
There are two basic forms of the incidence model depending upon whether or not choice
is also modeled. When choice is included, and only one category is considered in the model,
then incidence is modeled as a separate alternative (also called the outside good, or no-purchase
option). For example, if we assume that a consumer’s decision to purchase in a category is
independent of the other categories, conditioned on the consumer being in a specific store, we
can write the utility for category incidence as:
chthcthhtcht
XZU
εββ
++=
21
where Z
ht
represents household-specific factors like consumer inventory levels and/or time since
the household last purchased in the category (Briesch et al., 2008b; Manchanda et al., 1999), X
ct
represents category-specific factors like merchandising and price (Manchanda et al., 1999; Niraj
et al., 2008), and ε
cht
represents the random component or error term. This model allows
managers to determine how in-store merchandising activity affects the consumer’s purchase
timing decision, after controlling for consumer’s inventory levels and interpurchase time. As
40
noted above, as the utility increases (decreases), the consumer is more (less) likely to purchase
within the category.
While it is beyond the scope of this chapter to discuss all of the variants of incidence
models (see, Seetharaman, et al. (2005) for a discussion of this literature), we provide a model of
cross-category incidence based upon the model in Niraj, et al (2008), where they study cross-
category purchase incidence. This behavior is very important to managers, as it affects the
profitability of promotions. Specifically, managers may offer a promotion where they lose
money on pasta, but because consumers also buy pasta sauce, complementary category sales can
make the promotion profitable. Their joint category incidence model has the form:
htht
hthhththththt
hthththt
hthththt
U
XXXXU
XXU
XXU
0000
111122221112211111
0122221101
1012211110
ε
εγββββ
εββ
ε
β
β
=
+++++=
++=
+
+
=
where U
10ht
represents the indirect utility for household h in period t of purchasing in the first
category, but not the second category, U
01ht
represents the indirect utility of purchasing in the
second category but not the first category, U
11ht
represents the indirect utility of purchasing in
both categories, and U
00ht
represents the utility of purchasing in neither category. The X’s
represent predictor variables for incidence in each category (e.g., category price, display and
feature activity). When the merchandising variables are included as part of the X’s, then this
model accounts for cross-category effects as the merchandising activity in one category is
included in the indirect utility of the other category. This effect has been called purchase
complementarity /substitution, that is the categories are purchased together (complementarity) or
exclusive of each other (substitution)(Sun, 2005). γ
11h
represents the additional utility the
household receives for purchasing in both categories.
41
6.c.3 Brand Choice Models
Brand choice models normally assume that consumers have exogenously decided where
to shop and when to shop, although these models have been extended to include incidence as
well (see, Chandukala et al., 2007 for a good review of the assumptions, trends and construction
of choice models). Choice models address the important question: given the merchandising
effects in the store, which brand (or SKU) does the consumer select? Assuming that household h
is presented with J alternatives (j=1..J) in period t (t=1..T), the choice model can be written as:
hjthjhjt
hjthjthjt
Xv
vU
ββ
ε
+=
+
=
0
where U
hjt
is the indirect utility household h places on alternative j in period t, v
hjt
is called the
systematic component of utility and ε
hjt
is the random component. X
jt
then represents the
predictor variables for choice, e.g., price, display, feature advertising, etc.
Note that the competitive pricing effects are not included in the indirect utility, as they
are captured indirectly through consumers selecting one alternative from all of the alternatives.
This effect is seen more clearly when we write out the logit model as:
==
=
J
i
hjthjthjt
vvyP
1
)exp(/)exp()1(
where P(·) represents the probability that household h selects alternative j in period t. Because
choice is conditional on incidence, only variables that are alternative-specific are included in the
model. Variables common to all alternatives (e.g., consumer inventory levels) are not identified.
Individual-Level Promotional Models Using Firm Purchase History Data
42
The next frontier in promotional modeling is using actual individual-level purchase
histories rather than panel data. Two sources now exist. First, retailers offer frequent shopper
cards that record all purchases from the retailer by SKU and trip. Second, some direct-to-
consumer sellers capture all of their customers' purchase histories. These companies range from
on-line retailers such as Amazon to telecommunication companies such as Verizon or AT&T.
When a promotion is run, the firm can analyze what the impact is on the acquisition of new
customers, retention rates, spending levels and long-term purchase behavior.
New models are being developed to analyze these data such as hazard models with
explanatory variables to capture retention rates, logit models to estimate the responsiveness of
current customers to promotions for additional products, and customer equity models to compute
the long-term value of promoted versus not-promoted customers.
Hazard models study the likelihood a customer will defect given the customer has
continued to be a customer up until a promotion occurs. This allows us to address key issues
such as how does a promotion change the likelihood a customer will defect? How does the
customer acquisition method affect the hazard rate (defection rate)? (see, e.g., Gupta et al.,
2004; Jain and Singh, 2002; Thomas, 2001; Blattberg et al., 2008).
The availability of better data allows the firm to develop data-driven analytical decisions
regarding promotions. For example, one of the age-old promotion questions is does a promotion
causes customers to require lower prices in order to continue purchasing from the company if
their initial purchase is on promotion. Long-historical data series enable firms to answer this
question. Panel data, because of the limited samples sizes and problems with the accuracy and
availability of causal data, cannot easily answer many of key individual behavior questions.
Long individual purchase history data can address these questions.
43
Another area in which individual purchase histories can help a firm is in segmenting
customers based on their promotional sensitivity. Models can be used to estimate each
individual’s sensitivity to promotions and then using the results, the firm can then segment
customers so that different promotional offers can be provided based on their promotional
sensitivity.
The two potential problems with individual purchase history data are the lack of
promotional histories and the failure to model non-response to promotions. Surprisingly, many
firms do not maintain accurate promotional histories by customers. This limits the firm’s ability
to develop these models. The models built must also include all promotions offered to the
customer and then model the non-response as well as the response.
Modeling Issues
In this section we briefly discuss modeling issues associated with sales promotions.
Consumers have Different Promotional Responsiveness
Most researchers and managers believe consumer segments respond differently to
promotions. Models need to be able to capture heterogeneity in consumer response to
promotions otherwise the estimated parameters may be biased and provide nonsensical results
(Allenby and Rossi, 1999). There are two fundamental types of consumer heterogeneity –
response and structural. Response heterogeneity assumes that consumers use the same model of
choice, but differ in their parameters (e.g., the β’s in the above equations). The parameters can
be assumed to have either a continuous or discrete distribution. Research suggests that both
provide reasonable estimates (Andrews et al., 2002). From a practical standpoint, discrete
distributions (i.e., consumer segments are estimated) are more useful, but require the model to be
44
estimated assuming that the researcher knows the true number of segments. To determine the
"best" number of segments, the model must be estimated multiple times - a separate estimation
for each different number of segments in the market. These estimation results are then compared
and one final number of segments is selected based upon some information criteria (Kamakura
and Russell, 1989; Dayton and McReady, 1988).
When a continuous distribution is used, the coefficients can either be estimated using
Bayesian Methods (see, e.g. Rossi et al., 1995) or through classical econometric techniques. In
the Bayesian methods, Markov Chains are used to get simulation-based estimates of the
coefficients and their distributions (Rossi and Allenby, 2003). Classical econometrics also use
simulation-based techniques to estimate the parameters through a technique called Simulated
Maximum Likelihood Estimation (SMLE). The basic idea is that the distribution of the
parameters is assumed beforehand, and this distribution is numerically integrated during the
estimation process (Train, 2003).
Structural heterogeneity assumes that differences in consumer response among segments
are best modeled using structurally different models for each segment. If the consumer segments
are defined beforehand, estimating structural heterogeneity is straightforward. However, if the
underlying consumer segments are not pre-identified, this type of model requires discrete
heterogeneity. If the researcher is using POS data, several methods for estimating discrete
segments at the aggregate-level have been proposed (Besanko et al., 2003; Bodapati and Gupta,
2004).
Modeling Using Aggregate POS or Panel Data
Does aggregating household panel data to the retailer level produces any biases in the
estimates? Gupta et al. (1996) examined this issue and found that when similar models were
45
used for both the panel and store data, while the price elasticities were statistically different, they
would not significantly change managerial decisions, i.e., they were not managerially different.
This finding is consistent with a meta analysis of price elasticities which finds no significant
difference between POS and panel data elasticities (Bijmolt et al., 2005). The decision about the
aggregation level seems to be driven by whether or not the aggregation-level can answer the
research question and not by biases introduced by the level of aggregation.
Can Brands, SKU’s or Brand Sizes be Pruned From the Data?
In modeling panel data, it is common for researchers to eliminate SKU’s to be able to
efficiently estimate the model parameters. We note that this issue only applies to models that are
constructed as choice models as opposed to models constructed as sales models. Zanutto and
Bradlow (2006) found that the missing data introduced by deleting SKUs, brand sizes or brands
can lead to severe parameter bias in the estimation. The magnitude of the bias is related to the
pruning method selected, whether or not the model is misspecified, and the relative fit of the
model (models with better fits having lower bias).
Can SKUs Be Aggregated To A Higher Level?
Andrews and Currim (2005) found that significant biases can result from aggregating
SKU’s to the brand level in choice models. They recommend that, whenever possible, modeling
should be done at the level where the marketing mix decisions are made. For instance, in soft-
goods, such as clothing , groups of SKUs (e.g., shirts from the same manufacturer that only
differ in size) have the same prices. This bias was found to be related to linearly aggregating
(taking a simple average over the SKUs) continuous variables from different SKUs. On the
other hand, when marketing mix decisions are made at a level higher than the SKU-level (e.g.,
46
prices are set at the brand-level for multiple flavors), estimation issues are introduced into SKU-
level models because several predictors have high (or perfect) correlation.
Can Households be Removed From Panel Data Estimation?
When households are removed using a random selection technique (e.g., to generate a
hold-out sample) no biases are expected to be generated although the model loses power due to
the smaller sample size.
Gupta, et al (1996) identify two general methods used to non-randomly exclude
households from panel data: 1) household selection and 2) purchase selection. Under household
selection, all purchases from those households that only purchase the selected items are included
in the estimation. Under purchase selection, all purchases of the selected items are included
(regardless of the household purchase history). They find that household selection can create
severe biases in the parameter estimates when a large portion of the households are excluded.
Kim and Rossi (1994) show that parameters will be biased towards higher price sensitivity when
the sample is restricted to high purchase frequency or high volume households.
How Should Zero Sales of an Item be Handled?
This question has two components. First, if no observations are recorded and the analyst
does not have pricing information, what price should be used in this period (even as a
competitive price predictor)? Second, should the observation with zero sales be included in the
estimation?
Erdem, et al. (1999) investigate the first issue and find that imputing prices (and/or
coupons) can create severe biases in the estimated parameters. They recommend integrating out
the missing price data over the observed distribution. A clear drawback to this method is the
47
computational resources required to estimate the model. Some, but not all, of this problem can
be alleviated when the analyst has separate information about the retailer pricing behavior. For
instance, if the analyst knows that the retailer only changes prices once a week (say Friday), then
missing prices within that week (say Wednesday) can be imputed from the other prices within
the week.
Briesch, et al. (2008b) investigate the second issue within a choice context. They show
that even when price information is accurate and complete, the parameter estimates can be biased
when the zeros are improperly included or excluded. The general idea is that when the zeros are
due to exogenous factors like the item being out of stock, then the zero should be excluded from
the estimation. However, when the zeros are due to endogenous factors (e.g., promotion of
competing item), then the item should be included in the estimation. They develop models,
based on the literature on choice sets (Bronnenberg and Vanhonacker, 1999), that reduces the
bias from the zeros.
Emerging Issues
There are a number of issues in the promotion literature that have emerged over the last
few years that require further research.
Overstock Promotional Clearance Strategies
An emerging issue that has been previously researched but is important for soft-goods
and hard goods but less important for packaged-goods is how to price overstocked goods.
Lazear (1986) developed a theory of temporal promotions which assumed that a segment of
consumers were fashion-driven and would pay more because they could receive the goods if they
purchased early while another segment of consumers waited for lower prices and took the risk
48
that the product would be available. For firms trying to manage overstocks and perishable
inventory, an important issue becomes how many consumers are in the “wait and see” segment
and how many want to ensure the availability of the merchandise (Gallego et al., 2009). For a
more complete review of this literature, see the chapter on dynamic pricing in this book (Aviv
and Vulcano, 2010).
The emerging issue is how consumer expectations enter the process. The pricing
literature has developed models which begin to take into account consumer expectations
regarding future prices. If a firm is known to price lower in the future for the same product, then
consumers have an incentive to act strategically and delay their purchase (Soysal, 2007; Su,
2007; Aviv and Pazgal, 2008; Boyaci and Ozer, forthcoming).
An important and related research area is the accuracy of consumer expectations with
respect to promotions. Sun (2005) studies this issue and concludes that consumers can
accurately anticipate promotions. However further research into this area is needed. Obviously,
this has great significance to managers at both manufacturers and retailers.
Promotions for *on-Packaged Goods Companies
Much of the promotional research in marketing has been for high-velocity packaged-
goods products because of the availability of the data. However, promotions are very important
in many other industries such as the fashion and durable goods (e.g., electronics and
automobiles). Automobiles highlight a very important issue discussed in the last section which
is the impact of promotions on consumer expectations and the timing of purchases. Promotions
have become very common in the automotive industry and the consumer has learned to wait to
purchase based on when the promotions will occur. This raises issues about what the “reference
49
price” consumers use for a car. Because consumers have learned to purchase only when there is
a promotion, managing production, inventories and list price have become very complex (Busse
et al., 2007).
For electronics and fashion, the promotional issues are linked to seasonal product change
over holiday seasons and other peak selling seasons. An age old question is: “When is the
optimal time to offer promotions – during the peak season or the non-peak season?” The
obvious answer is to offer them in the non-peak season but there are numerous promotions made
during the peak season (Chevalier et al., 2003).
How Do Promotions Affect the Lifetime Value of a Customer within Loyalty Programs?
Loyalty programs using frequent shopper cards are now widespread in retailing. These
programs rely heavily on promotions but do these programs increase loyalty and lifetime value
of a customer? There is limited research on this topic. Lal and Bell (2003) find that frequent
shopper programs increase causal shoppers (often called cherry pickers) while subsidizing loyal
customers. Loyalty program because of the promotional strategy used by retailers may not
actually produce the desired goal of increasing loyalty.
This raises many questions that have been asked and some addressed about promotions.
In the 80’s and 90’s many anti-promotional advocates who believed that advertising was a
preferred marketing expenditure argued that promotions reduced brand equity and loyalty.
Dodson, et al. (1978) raised this issue and found in laboratory studies that promotions decreased
brand loyalty. However, others such as Johnson (1984) using real-world, (not laboratory
generated) data, found that brands that increased promotions did not exhibit decreased brand
loyalty. The empirical findings on this issue are mixed and this topic still remains highly
50
controversial. The advent of retail loyalty data provides another opportunity to assess whether
promotions can be used to increase or decrease loyalty.
Linking Psychological Theories of Promotions and Quantitative Research
There needs to be a greater link between psychological theories such as smart shopper
and transactional utility with empirical research. How do smart shoppers learn to be “smart
shoppers?” Do they simply respond to promotions that provide reference prices? How large is
transaction utility and under what conditions does it occur?
Conclusions/Summary
In this chapter we covered the important topic of sales promotions from the consumer,
strategic, and empirical perspectives. From a consumer standpoint, the behavioral theories of
transaction utility, smart shopper, and reference price provide insight into why consumers
respond strongly to promotions than to a price decrease.
It is important for the managers to know the source of the incremental volume from a
promotion, as this plays a major role in determining promotional profitability for both the
manufacturer and retailer. When the source of the volume is brand switching, the promotion can
be very profitable for the manufacturer, but is only profitable for the retailer when consumers are
switching to higher margin items. For many categories, when the source of the incremental
volume is purchase acceleration and stockpiling, then sales promotions will not be profitable for
manufacturers unless they lead to increased consumption. Consumers will simply change their
purchase timing, particularly loyal customers. When the source of the incremental volume is an
increase in the number of consumers purchasing in the category, sales promotions can be
profitable for both the retailer and manufacturer. An important and unresolved issue is when
51
promotions draw new consumers into the category (or increase consumption of the category),
can sustained promotional activity keep these new consumers purchasing in the category and/or
maintain the increased consumption rate of current consumers? If so, promotions could expand
the overall category and lead to increased manufacturer and retailer profitability.
Strategically, there are many reasons sales promotions exist. First, trade deals can help
coordinate activities within a marketing channel, ensuring that the participants are striving to
maximize profitability for all firms. Second, sales promotions can be used as a competitive tool
to temporarily steal market share from competitors, to protect one’s own market share, or to
prevent new entrants and/or store brands from gaining significant market share. Third, sales
promotions can be used to price discriminate against some consumers, raising the overall
profitability of the firm. For instance, shopper cards are a mechanism where retailers can reward
loyal consumers by providing discounts on certain merchandise. Similarly, coupons are a
mechanism where manufacturers and retailers can provide discounts to targeted consumers –
those who have the time to clip and redeem coupons.
The Internet provides new challenges and opportunities for sales promotions. The
detailed consumer information as well as accurate casual information allows researchers to
examine individual purchasing behavior. This type of data can provide better insights into
consumer behavior and decision making.
The Internet raises a host of questions about consumer behavior, and ultimately firm
profitability. For instance, given the new types of communication vehicles (e.g., email, banner
ads, partner links, keywords), which are more effective in attracting new customers?
Specifically, how do they affect traffic (clicks, unique visits, new users)? How do they affect
conversion rates? How do these different mechanisms affect brand image and loyalty? Given
52
the “ease” of information search, are consumers more or less price sensitive? Are they more or
less brand/site loyal? How large is the spike in traffic/sales associated with these promotions?
Are consumers more likely to cherry pick on promotions? How large are their basket sizes
relative to “brick and mortar” retailers? From which specific channels do internet retailers
compete or take sales? Is it mass retailers or catalogue retailers? Even if large spikes in
sales/traffic are observed due to Internet promotions, how much of this spike is due to purchase
acceleration, quantity acceleration, and brand switching?
Promotions have been one of the most researched elements of the marketing mix. There
are still many interesting and challenging issues facing managers. The growth of real-time, sales
data coupled with accurate promotional histories opens the future in which real-time on-line
promotions will become the norm. This leads to opportunities for sales promotion researchers
and practitioners to develop new methods, tactics and strategies.
53
References
ACNielsen. (2002) AC<ielsen 2002 Trade Promotion Practices Study. Available at:
http://www2.acnielsen.com/pubs/2003_q2_ci_tpp.shtml.
Agarwal D. (1996) Effect of Brand Loyalty on Advertising and Trade Promotions: A Game
Theoretic Analysis with Empirical Evidence. Marketing Science 15: 86-108.
Ailawadi KL, Harlam BA, Cesar J, et al. (2007) Quantifying and Improving Promotion
Effectiveness at CVS. Marketing Science 26: 566-575.
Ailawadi KL and Neslin SA. (1998) The Effect of Promotion on Consumption: Buying it More
and Consuming it Faster. Journal of Marketing Research 35: 390-398.
Alba JW, Mela CF, Shimp TA, et al. (1999) The Effect of Discount Frequency and Depth on
Consumer Price Judgments. Journal of Consumer Research 26: 99-114.
Allenby GM and Rossi PE. (1999) Marketing Models of Consumer Heterogeneity. Journal of
Econometrics 89: 57-78.
Andrews RL, Ainslie A and Currim IS. (2002) An Empirical Comparison of Logit Choice
Models with Discrete Versus Continuous Representations of Heterogeneity. Journal of
Marketing Research 39: 479-487.
Andrews RL and Currim IS. (2005) An Empirical Investigation of Scanner Data Preparation
Strategies for Consumer Choice Models. International Journal of Research in Marketing
22: 319-331.
Aviv Y and Pazgal A. (2008) Optimal pricing of seasonal products in the presence of forward-
looking consumers. Manufacturing & Service Operations Management 10: 339-359.
Aviv Y and Vulcano. (2010) Dynamic Pricing. In: Phillips R and Ozer O (eds) Oxford Book of
Pricing.
Aydin G and Hausman WH. (2009) The role of slotting fees in the coordination of assortment
decisions. Forthcoming in Production and Operations Management 8: 635-652.
Bell DR, Chiang J and Padmanabhan V. (1999) The Decomposition of Promotional Response:
An Empirical Generalization. Marketing Science 18: 504-526.
Besanko D, Dube J-P and Gupta S. (2003) Competitive Price Discrimination Strategies in a
Vertical Channel Using Aggregate Retail Data. Management Science 49: 1121-1138.
Bijmolt THA, Heerde HJv and Pieters RGM. (2005) New Empirical Generalizations on the
Determinants of Price Elasticity. Journal of Marketing Research 42: 141-156.
Blattberg RC, Briesch RA and Fox EJ. (1995) How Promotions Work. Marketing Science 14:
G122-G132.
Blattberg RC, Kim B-D and Neslin SA. (2008) Database Marketing: Analyzing and Managing
Customers, New York: Springer Verlag.
Blattberg RC and Neslin SA. (1990) Sales Promotions: Concepts, Methods and Strategies,
Englewood Cliffs, N.J.: Prentice Hall.
54
Bloom PN, Gundlach GT and Cannon JP. (2000) Slotting Allowances and Fees: Schools of
Thought and the Views of Practicing Managers. Journal of Marketing 64: 92-108.
Bodapati AV and Gupta S. (2004) The Recoveraility of Segmentation Structure from Store-
Level Aggregate Data. Journal of Marketing Research XLI: 351-364.
Boyaci T and Ozer O. (forthcoming) Information Acquisition for Capacity Planning Via Pricing
and Advance Selling: When to Stop and Act? Operations Research.
Briesch RA, Chintagunta PK and Fox EJ. (2008a) How Does Assortment Affect Grocery Store
Choice? Journal of Marketing Research forthcoming.
Briesch RA, Dillon WR and Blattberg RC. (2008b) Treating Zero Brand Sales Observations in
Choice Model Estimation: Consequences and Potential Remedies. Journal of Marketing
Research 45: 618-632.
Bronnenberg BJ and Vanhonacker WR. (1999) Limited Choice Sets, Local Price Response, and
Imputed Measures of Price Competition. Journal of Marketing Research 33: 163-173.
Bronnenberg BJ and Wathieu L. (1996) Asymmetric Promotion Effects and Brand Positioning.
Marketing Science 15: 379-394.
Busse MR, Simester D and Zettlemeyer F. (2007) "The Best Price You'll Ever Get" The 2005
Employee Discount Pricing Promotions in the U.S. Automobile Industry. <BER Working
Paper <o. 13140.
Chandon P, Wansink B and Laurent G. (2000) A benefit congruency framework of sales
promotion effectiveness. The Journal of marketing 64: 65-81.
Chandukala SR, Kim J, Otter T, et al. (2007) Choice Models in Marketing: Economic
Assumptions, Challenges and Trends. Foundations and Trends® in Marketing 2: 97-184.
Chernev A. (2006) Articulation Compatability in Eliciting Price Bids. Journal of Consumer
Research 33: 329-341.
Chevalier JA, Kashyap AK and Rossi PE. (2003) Why Don't Prices Rise During Periods of Peak
Demand? Evidence from Scanner Data. American Economic Review 93: 15-37.
Cooper RG. (2001) Winning at <ew Products: Accelerating the Process from Idea to Launch,
Cambridge, MA: Perseus Books Broup.
Dayton CM and McReady GB. (1988) Concomitant-Variable Latent-Class Models. Journal of
American Statistical Association 83: 173-178.
Dodson JA, Tybout AM and Sterntha B. (1978) Impact of Deals and Deal Retraction on Brand
Switching. Journal of Marketing Research 15: 72-81.
Dreze X and Bell DR. (2003) Creating Win-Win Trade Promotions: Theory and Empirical
Analysis of Scan-Back Trade Deals. Marketing Science 22: 16-39.
Erdem T, Imai S and Keane MP. (2003) Brand and Quantity Choice Dynamics Under Price
Uncertainty. Quantative Marketing and Economics 1: 5-64.
Erdem T, Keane MP and Sun B. (1999) Missing Price and Coupon Availability in Scanner
Panels: Correcting for the Self-Selection Bias in Choice Model Parameters. Journal of
Econometrics 89: 177-196.
55
Fox EJ and Hoch SJ. (2005) Cherry-Picking. Journal of Marketing 69: 46-62.
Gallego G, Phillips R and Sahin O. (2009) Strategic management of distressed inventory.
Production and Operations Management 17: 402-415.
Gómez MI, Rao VR and McLaughlin EW. (2007) Empirical analysis of budget and allocation of
trade promotions in the US supermarket industry. Journal of Marketing Research 44:
410-424.
Guadagni PM and Little JDC. (1983) A Logit Model of Brand Choice Calibrated on Scanner
Data. Marketing Science 2: 203-238.
Gupta S. (1988) Impact of Sales Promotions on When, What, and How Much to Buy. Journal of
Marketing Research 25: 342-355.
Gupta S, Chintagunta PK, Kaul A, et al. (1996) Do Household Scanner Data Provide
Representative Inferences From Brand Choices: A Comparison With Store Data. Journal
of Marketing Research XXXIII: 383-398.
Gupta S, Lehmann DR and Stuart JA. (2004) Valuing customers. Journal of Marketing Research
41: 7-18.
Heath TB, Ryu G, Chatterjee S, et al. (2000) Asymmetric Competition in Choice and the
Leveraging of Competitive Disadvantages. Journal of Consumer Research 27: 291-308.
Hendel I and Nevo A. (2006) Sales and Consumer Inventory. The RA<D Journal of Economics
37: 543-561.
Howard DJ and Kerin RA. (2006) Broadening the Scope of Reference Price Advertising
Research: A Field Study of Consumer Shopping Involvement. Journal of Marketing 70:
185-204.
Jain D and Singh SS. (2002) Customer lifetime value research in marketing: A review and future
directions. Journal of Interactive Marketing 16: 34-46.
Jainiszeqski C and Lichtenstein. (1999) A Range Theory Account of Price Perception. Journal of
Consumer Research 25: 353-368.
Johnson T. (1984) The Myth of Declining Brand Loyalty. Journal of Advertising Research 24: 9-
17.
Kamakura WA and Russell G. (1989) A Probabilistic Choice Model for Market Segmentation
and Elasticity Structure. Journal of Marketing Research 26: 379-390.
Kaya and Ozer O. (2010) Pricing Contracts. In: Phillips R and Ozer O (eds) Oxford Book of
Pricing.
Kim B-D and Rossi PE. (1994) Purchase Frequency, Sample Selection and Price Sensitivity.
Marketing Letters 5: 57-67.
Krishna A, Briesch R, Lehmann D, et al. (2002) A Meta-Analysis of the Impact of Price
Presentation on Perceived Savings. Journal of Retailing 78: 101-118.
Krishna A and Johar GV. (1996) Consumer Perceptions of Deals: Biasing Effects of Varying
Deal Prices. Journal of Experimental Psychology 2: 187-2006.
56
Lal R. (1990) Price promotions: limiting competitive encroachment. Marketing Science 9: 247-
262.
Lal R and Bell DR. (2003) The Impact of requent Shopper Programs in Grocery Retailing.
Quantative Marketing and Economics 1: 179-202.
Lazear EP. (1986) Retail Pricing and Clearance Sales. American Economic Review 76: 14-32.
Mace S and Neslin SA. (2004) The Determinants of Pre- and Postpromotion Dips in Sales of
Frequently Purchased Goods. Journal of Marketing Research 41: 339-350.
Manchanda P, Ansari A and Gupta S. (1999) The "Shopping Basket;" A Model for
Multicategory Purchase Incidence Decisions. Marketing Science 18: 95-114.
Mazumdar T and Papatla P. (2000) An Investigation of Reference Price Segments. Journal of
Marketing Research 37: 246-258.
Mela CF, Jedidi K and Bowman D. (1998) The Long-Term Impact of Promotions on Consumer
Stockpiling Behavior. Journal of Marketing Research 35: 250-262.
Narasimhan C. (1988) A Price Discrimination Theory of Coupons. Marketing Science 3: 128-
147.
Neslin SA. (2002) Sales Promotion (Relevant Knowledge Series), Cambridge, MA: Marketing
Science Institute.
Niedrich RW, Sharma S and Wedell DH. (2001) Reference Price and Price Perceptions:A
Comparison of Alternative Models. Journal of Consumer Research 28: 339-354.
Nijs VR, Dekimpe MG, Steenkamp J-BEM, et al. (2001) The Category-Demand Effects of Price
Promotions. Marketing Science 20: 1-22.
Niraj R, Padmanabhan V and Seetharaman PB. (2008) A Cross-Category Model of Households'
Incidence and Quantity Decisions. Marketing Science 27: 225-235.
Pauwels K, Hanssens DM and Siddarth S. (2002) The Long-Term Effects of Price Promotions on
Category Incidence, Brand Choice and Purchase Quantity. Journal of Marketing
Research XXXIX: 421-439.
Ramakrishnan R. (2010) Markdown Pricing. In: Phillips R and Ozer O (eds) Oxford Book of
Pricing.
Rao RC. (1991) Pricing and Promotions in Asymmetric Duopolies. Marketing Science 10: 131-
144.
Rhee H and Bell DR. (2002) The Inter-Store Mobility of Supermarket Shoppers. Journal of
Retailing 78: 225-237.
Rossi PE and Allenby GM. (2003) Bayesian statistics and marketing. Marketing Science 22:
304-328.
Rossi PE, McCulloch RE and Allenby GM. (1995) Hierarchial Modelling of Consumer
Heterogeneity: An Application to Target Marketing. In: Singpurwalla Ka (ed) Case
Studies in Bayesian Statistics. New York: Springer Verlag, 323-350.
Russell GJ and Kamakura WA. (1997) Modeling Multiple Category Brand Perference with
Household Basket Data. Journal of Retailing 73: 439-461.
57
Schindler RM. (1998) Consequence of Perceiving Oneself as Responsible for Obtaining a
Discount: Evidence for Smart-Shopper Feelings. Journal of Consumer Psychology 7:
271-392.
Seetharaman PB, Chib S, Ainslie A, et al. (2005) Models of Multi-Category Choice Behavior.
Marketing Letters 16: 239-254.
Sethuraman R. (2009) Assessing the External Validity of Analytical Results from National
Brand and Store Brand Competition Models. Marketing Science 28: 759-781.
Sethuraman R and Srinivasan V. (2002) The Asymmetric Share Effect. An Empirical
Generalization on Cross-Price Effects. Journal of Marketing Research 39: 379-386.
Sethuraman R, Srinivasan V and Kim D. (1999) Asymmetric and Neighborhood Cross-Price
Effects: Some Empirical Generalizations. Marketing Science 18: 23-41.
Sivakumar K and Raj SP. (1997) Quality Tier Competition: How Price Change Influences Brand
Choice and Category Choice. Journal of Marketing 61: 71-84.
Soysal GP. (2007) Demand Dynamics in the Seasonal Goods Industry: An Empirical Analysis.
Kellogg School of Management. Evanston: Norhwestern University.
Srinivasan S, Pauwels K, Hanssens DM, et al. (2004) Do Promotions Benefit Manufacturers,
Retailers, or Both? Management Science 50: 617-629.
Srinivasan VS and Bodapati AV. (2006) The Impact of Feature Advertising on Customer Store
Choice. Stanford University Graduate School of Business Research Paper <o. 1935.
Stanford University.
Su X. (2007) Intertemporal pricing with strategic customer behavior. Management Science 53:
726.
Sudhir K and Rao VR. (2006) Do Slotting Allowances Enhance Efficiency or Hinder
Competition. Journal of Marketing Research XLIII: 137-155.
Sun B. (2005) Promotion Effect on Endogenous Consumption. Marketing Science 24: 430-443.
Taylor TA. (2002) Supply chain coordination under channel rebates with sales effort effects.
Management Science 48: 992-1007.
Thaler R. (1985) Mental Accounting and Consumer Choice. Marketing Science 61: 427-449.
Thomas JS. (2001) A methodology for linking customer acquisition to customer retention.
Journal of Marketing Research 38: 262-268.
Train KE. (2003) Discrete Choice Methods with Simulation, Cambridge: Cambridge University
Press.
Van Heerde HJ, Gupta S and Wittink DR. (2003) Is 75% of the Sales Promotion Bump Due to
Brand Switching? No, Only 33% Is. Journal of Marketing Research 40: 481-491.
Van Heerde HJ, Leeflang PS and Wittink DR. (2000) The Estimation of Pre- and Postpromotion
Dips with Store-Level Scanner Data. Journal of Marketing Research XXXVII: 383-395.
Van Heerde HJ, Leeflang PS and Wittink DR. (2002) How Promotions Work: SCAN*PRO-
Based Evolutionary Model Building. Schmalenbach Business Review 54: 198-220.
58
van Heerde HJ and Neslin SA. (2008) Sales Promotion Models. In: Wierenga B (ed) Handbook
of Marketing Decision Models. Springer, 107-162.
Van Ittersum K, Pennings JME, Wasnink B, et al. (2005) The Effect of Primed and Framed
Reference Points on Product Attribute Importance. Advances in Consumer Research.
Varian HR. (1980) A model of sales. The American Economic Review: 651-659.
Winer RS. (1986) A reference price model of brand choice for frequently purchased products.
Journal of Consumer Research 13: 250-256.
Winer RS. (2005) Pricing (Relevant Knowledge Series), Cambridge, MA: Marketing Science
Institute.
Zanutto EL and Bradlow ET. (2006) Data Pruning in Consumer Choice Models. Quantative
Marketing and Economics forthcoming.
... See the last paragaph of the concluding section inAguirregabiria (1999), as well asBlattberg and Briesch (2010) for a more detailed description of promotional mechanisms.3 For example,Agrawal (1996) noted that smaller brands should rather focus on advertising than price promotions. ...
Conference Paper
This thesis comprises three essays addressing questions related to pricing, competition, and consumer welfare in imperfectly competitive industries characterised by the existence of demand and supply side frictions. Chapter 1 studies a dynamic pricing game amongst differentiated multiproduct oligopolists who have incentives to temporarily lower prices to attract new consumers who are more likely to purchase the same product again at a higher price due to inertia. The novel feature of the model, which allows to explain persistence in the observed patterns of retail prices, is the possibility of costly price adjustment. The magnitude of adjustment costs is estimated using scanner data on purchases of a category of dairy products by UK households. The main results suggest that adjustment costs can be substantial for manufacturers, but they are passed through to the consumers only on a very limited scale. Chapter 2 explores an alternative explanation for price dispersion by introducing a model with search frictions and private information about marginal costs. Consumers decide how many sellers to visit who in equilibrium set prices in a fashion similar to bidding in a reverse first-price auction with an unknown number of competitors. The chapter shows how the distributions of search costs and firm heterogeneity can be nonparametrically identified and analyses convergence rates of the proposed estimators. Chapter 3 uses an extension of the search model proposed earlier to study the value of information provided by mortgage brokers in the UK. Prospective borrowers can either directly search and apply for mortgages with different lenders or use a broker who finds the best rate on their behalf. The main finding suggests that, on average, the existence of brokers substantially fosters competition between lenders, leads to lower monthly payments in equilibrium and reduces the deadweight loss arising as a result of costly search.
Chapter
This paper describes research being conducted in field of promotion planning and optimization for a chain of convenience stores. The motivation for choosing this subject is an important role of promotions in retail market and availability of large amount of data that can be used to improve profitability of promotions. In addition, most of existing studies analyzed promotions in super- and hypermarkets which have a different sales characteristic than a convenience chain. Since transaction amount is typically small (in comparison to transactions in bigger stores), we want to check whether findings from previous studies can be confirmed in our testing environment. In this paper, we show how both internal and external data can be used in order to improve accuracy of forecasts and obtain more reliable performance metrics. The thesis and research goals are presented along with key results of literature review.
Article
Full-text available
This empirical study contributes towards identifying the effect of both fake and real discounts in the Indian marketing environment. A common but unsustainable practice in India is to increase the selling price and then offer a discount on the product. Increasing sales based on fake discount pricing strategy is a primary business development objective in India. The discounts, however, vary across store type and time and are based on product features. The selected databases were collected from the top five Indian e-commerce portals in terms of volume of sales, and from popular brick and motor outlets of tier 2 and tier 3 cities in India. The empirical results indicate that offers based on price in India had an impact of 2.8 times higher than the actual quality of the product. The outcomes suggest that marked price has a significant impact on consumer's behavior. The results also indicate the existence of a strong correlation between trapping fake discounts and purchase by deceiving and persuading customers in India. Research is empirical in nature and respondents have been selected based on purposive sampling. The study is limited to tier 2 and 3 cities of India for 250 days, and the results are applicable to online and offline retail stores.
Article
Full-text available
Marketing scholars commonly characterize market structure by studying the patterns of substitution implied by brand switching. Though the approach is useful, it typically ignores the destabilizing role of marketing variables (e.g., price) in switching behavior. The authors propose a flexible choice model that partitions the market into consumer segments differing in both brand preference and price sensitivity. The result is a unified description of market structure that links the pattern of brand switching to the magnitudes of own- and cross-price elasticities. The approach is applied in a study of competition between national brands and private labels in one product category.
Article
Full-text available
One of the mysteries of store-level scanner data modeling is the lack of a dip in sales in the weeks following a promotion. Researchers expect to find a postpromotion dip because analyses of household scanner panel data indicate that consumers tend to accelerate their purchases in response to a promotion - that is, they buy earlier and/or purchase larger quantities than they would in the absence of a promotion. Thus, there should also be a pronounced dip in store-level sales in the weeks following a promotion. However, researchers rarely find such dips at either the category or the brand level. Several arguments have been proposed to account for the lack of a postpromotion dip in store-level sales data and to explain why dips may be hidden. Because dips are difficult to detect by traditional models (and by a visual inspection of the data), the authors propose models that can account for a multitude of factors that together cause complex pre-and postpromotion dips. The authors use three alternative distributed lead and lag structures: an Almon model, an unrestricted dynamic effects model, and an exponential decay model. In each model, the authors include four types of price discounts: without any support, with display-only support, with feature-only support, and with feature and display support. The models are calibrated on store-level scanner data for two product categories: tuna and toilet tissue. The authors estimate the dip to be between 4 and 25% of the current sales effect, which is consistent with household-level studies.
Article
Full-text available
Studies of grocery sales show that consumers of store brands switch to (price) discounted national brands more than consumers of national brands switch to discounted store brands. Such asymmetric price competition can be explained with numerous mechanisms proposed here and elsewhere. We report a choice experiment that replicates asymmetric price competition favoring higher-quality competitors and demonstrates asymmetric quality competition favoring lower-quality competitors. Also demonstrated are multiple mechanisms contributing to competitive asymmetries, where dominance involving the otherwise preferred brand is particularly potent (e.g., when a higher-quality competitor matches the price of an otherwise preferred lower-quality brand). The findings implicate modifications to (1) theories of decision making when extended to repeat choice, (2) empirical models of secondary purchase data, and (3) strategies for positioning and attacking brands. Whereas improving competitive disadvantages often attracts consumers from competitors more than does improving competitive advantages, this benefit must be weighed against the differentiation sacrificed by improving competitive disadvantages (improving competitive advantages, in contrast, increases differentiation).
Article
Full-text available
The intensity of price discounting by retailers and manufacturers raises important questions about consumer price judgments. In the extreme, discounting can take the form of frequent but shallow discounts or deep but infrequent discounts. The research reported here explores the effects of these strategies on consumer estimation of price levels for competing stores and brands. In an initial experiment in which subjects made brand choices over time, a depth effect was observed that contrasted with the frequency effect found in previous research. Subsequent experiments identified the conditions under which depth (vs. frequency) characteristics of price data dominate consumers' price-estimation judgments. Frequency information is more influential when sets of interstore or interbrand comparative prices exhibit complex and overlapping distributions (hence creating processing difficulty); in contrast, a depth bias occurs when prices have a simpler, dichotomous distribution. These results place pragmatically meaningful limitations on the influence of frequency information and illustrate the importance of context in determining consumer price judgments in a promotional environment.
Chapter
An important aspect of marketing practice is the targeting of consumer segments for differential promotional activity. The premise of this activity is that there exist distinct segments of homogeneous consumers who can be identified by readily available demographic information. The increased availability of individual consumer panel data opens the possibility of direct targeting of individual households. Direct marketing activities hinge on the identification of distinct patterns of household behavior (such as loyalty, price sensitivity, or response to feature advertisements) from the consumer panel data for a given household. The goal of this paper is to assess the information content of various standard segmentation schemes in which variation in household behavior is linked to demographic variables versus the information content of individual data. To measure information content, we pose a couponing problem in which a blanket drop is compared to drops targeted on the basis of consumer demographics alone, and finally to a targeted drop which is based on household panel data. We exploit new econometric methods to implement a random coefficient choice model in which the heterogeneity distribution is related to observable demographics.
Article
The effectiveness of a sales promotion can be examined by decomposing the sales "bump" during the promotion period into sales increase due to brand switching, purchase time acceleration, and stockpiling. The author proposes a method for such a decomposition whereby brand sales are considered the result of consumer decisions about when, what, and how much to buy. The impact of marketing variables on these three consumer decisions is captured by an Erlang-2 interpurchase time model, a multinomial logit model of brand choice, and a cumulative logit model of purchase quantity. The models are estimated with IRI scanner panel data for regular ground coffee. The results indicate that more than 84% of the sales increase due to promotion comes from brand switching (a very small part of which may be switching between different sizes of the same brand). Purchase acceleration in time accounts for less than 14% of the sales increase, whereas stockpiling due to promotion is a negligible phenomenon accounting for less than 2% of the sales increase.