ArticlePDF Available

Dynamically Allocating the Marketing Budget. How to Leverage Profits across Markets, Products and Marketing Activities

{ Case Study }
/ / / Dynamic budget optimization outperforms rules of thumb
GfK MIR / Vol. 4, No. 1, 2012, pp. 50 59 / Case Study doi 10.2478 / gfkmir-2014-0041
51Case Study / Vol. 4, No. 1, 2012, pp. 5059 / GfK MIR
Rule of Thumb Marketing Budgeting is Common
Determining the marketing budget has been of para-
mount importance to marketers for many decades.
Global players such as Procter & Gamble spend billions
of dollars on advertising per year. Nevertheless, market-
ing practitioners frequently use rules of thumb when it
comes to determining marketing budgets. By far the
most often used budget rules across regions and indus-
tries are the “percentage-of-sales”, “objective-and-
task”, and “affordability” method. In addition, budget
decisions are often based on gut instinct and the nego-
tiations skills of individual managers. Consequently, poli-
tics and individual opinions tend to shape the decision
process instead of fact-based discussions. Obviously,
these rules and practices bear the risk of results far away
from the optimal profit-maximizing budget.
Challenges of Optimal Budget Allocation
The global annual marketing budget of a company is
usually set in the previous year, i.e., it is fixed. If compa-
nies offer a broad product portfolio to customers from
various countries and use a variety of communication
channels they need to break down the fixed annual
budget into expenditures across countries, products
and communication activities. For many firms this task
requires determining individual budgets for hundreds of
allocation units. As a result, firms face a complex decision
problem: they need to allocate a fixed budget across a
multitude of allocation units by evaluating the impact of
these investment decisions on future cash flows. Since
marketing expenditure is immediately recognized as
cost on the income statement but its total impact on
sales often only unfolds fully in future periods, it needs
to be evaluated in terms of an investment decision and
based on the principle of marginal returns. Technically
speaking, management needs to solve a dynamic opti-
mization problem for an investment portfolio under a
budget constraint. This management challenge recurs
on a regular basis as marketing budgets are set annually.
Consequently, a new allocation approach was developed.
In a first step, a theoretical solution is presented that
provides important insights into how individual budgets
should be set so that they account for differences in
profit contribution, marketing effectiveness, and growth
Dynamically Allocating the Marketing Budget
Marc Fischer, Sönke Albers, Nils Wagner and Monika Frie
Marc Fischer, Professor of
Marketing and Services,
University of Cologne, Germany,
Sönke Albers, Professor of
Marketing and Innovation,
Kühne Logistics University,
Hamburg, Germany,
Nils Wagner, Ph.D. candidate,
Monika Frie, formerly Head of
Global Business Support, Bayer
AG, Berlin, Germany,
The article is an adapted version
of Fischer, Marc; Albers, Sönke;
Wagner, Nils; Frie, Monika
(2011): “Dynamic Marketing
Budget Allocation across
Countries, Products, and
Marketing Activities”, Marketing
Science, Vol. 30, No. 4, 568 585.
Marketing budget decisions are critical and should be fact based rather than intuitive.
Profit can be improved by better allocating a fixed budget across products or regions. The
Excel-based decision support model presented in this article makes it possible to deter-
mine near-optimal marketing budgets and represents an innovative and feasible solution
to the dynamic marketing allocation budget problem for multi-product, multi-country
firms. The model accounts for marketing dynamics and a product’s growth potential as
well as for trade-offs with respect to marketing effectiveness and profit contribution. It
was successfully implemented at Bayer, one of the world’s largest firms in the pharma-
ceuticals and chemicals business. The profit improvement potential in this company was
more than 50% and worth nearly EUR 500 million in incremental discounted cash flows.
BO X 1 :
Constrained Dynamic Pro t
Maximization Problem
BO X 2:
Optimal Solution
BO X 3:
Heuristic Allocation Weight
Discounted net value of product portfolio
where Unit sales = f (life cycle, marketing expense, etc.)
(1) ∑k i n Marketing expensekint = Total Budgett
(2) Marketing effects decay at a constant rate over time
(3) Boundary conditions (e.g., positive marketing budgets)
factor t
k i
Time t Countries k Products i
Activities n
saleskit Max!
Countries k Products i Activities n
Optimal budgetkint = Total budgett
Optimal allocation weight
Optimal allocation weightkint
Optimal allocation weightkint =
1 + Discount rate – Marketing carryoverkin
unit saleskit
Optimal mktg
Optimal growth
Last period’s
marketing elasticity
1 + Discount rate – Marketing carryover
(Discounted) long-term
marketing effectivenesse
Growth potential
(T = Planning horizon)
Expected revenues
in T periods
Last period’s revenues
Size of profit
margin (%)
GfK MIR / Vol. 4, No. 1, 2012, pp. 50 59 / Case Study
» The profit improvement potential
from a better allocation of a total
marketing budget is much higher than
from optimizing the total budget
. «
potential. In a second step, a near-optimal allocation
rule is derived from that solution which addresses the
demand for simple allocation rules by practitioners. It can
be used with a spreadsheet. While easy to understand
and implement, the heuristic goes beyond widespread
budgeting rules such as the “percentage-of sales”.
Developing a Dynamic Budget Allocation Approach
According to previous findings the profit improvement
potential from a better allocation of a total marketing
budget is much higher than from optimizing the total
budget. Therefore the approach does not tackle how to
determine the overall budget but how to allocate a fixed
budget that is constant over the planning horizon. The
model provides a solution for an international firm that
offers a broad portfolio of products to customers from
different countries, using different marketing tools or
activities to promote its products (e.g., classical adver-
tising, below-the-line activities, personal selling, etc.).
The portfolio is composed of products that differ in their
life-cycle stage and the firm wishes to maximize the dis-
counted total profits of its portfolio.
Specifically, the model integrates and trades off infor-
mation about
> the size of the business,
> the profit contribution margin,
> the (short-term) effectiveness of marketing
> the carryover-effect of marketing investments,
> the growth potential,
> and the time value of money.
In the model, sales are represented by a general growth
function (a product life cycle) and the response of mar-
keting investments. The growth function describes the
evolution of new product sales over time and is assumed
to be influenced by marketing investments. The effect
of marketing investments is determined by a market-
ing stock that arises from previous investments and
depreciates over time (decay factor) plus the market-
ing investments in the current period. Based on these
specifications, the discounted net value of the product
portfolio is maximized. Box 1 shows the formulation of
the maximization problem and its restrictions in math-
ematical terms.
The optimal solution considers dynamics in two differ-
ent ways. First, it incorporates the dynamic effects of
building and leveraging the marketing stock, which is
reflected in the marketing carryover coefficient. Second,
it accounts for the growth potential of a product that
is related to marketing investments as reflected in the
growth elasticity.
The growth elasticity measures the power of marketing
to shape the life cycle. Hence, based on prior evidence it
is assumed that the growth process is not predetermined
but can be influenced by the level of marketing expendi-
ture in different phases of the life of a product. In par-
ticular, marketing investments in the growth potential of
a new product have a strong impact on future cumula-
tive sales and discounted cash flows. On the basis of a
parametric growth model, it can be demonstrated in the
subsequent case study how the optimal solution favors
shifting marketing resources to young products so that
they can leverage their endogenous growth potential.
The optimal solution is based on the principle of rela-
tive attractiveness of an allocation unit to get a share of
the total marketing budget. The idea is to calculate the
optimal allocation weight for a product, as an example,
and relate this weight to all allocation weights of the
portfolio. This share is proportional to the profit contri-
bution margin, current sales, marketing budget elasticity
and growth multiplier. Box 2 explains how the optimal
allocation weights for individual countries, products and
marketing activities can be determined in detail.
Case Study / Vol. 4, No. 1, 2012, pp. 5059 / GfK MIR
Implications for Budget Allocation
The optimal solution (Box 2) provides a number of intui-
tive insights into the allocation problem.
> The optimal budget for a product relative to other
products increases with its contribution margin and
its sales base.
> Similarly, the larger a product’s long-term marketing
effectiveness for a certain activity, the higher its opti-
mal budget.
> The long-term marketing effectiveness is composed
of the short-term sales elasticity, the discount rate,
and the marketing carryover. Consequently, if long-
term marketing effectiveness is larger across all
activities of a product compared to other products,
the total budget for that product increases.
> Finally, the sales growth elasticity varies over the life
cycle. It is largest at the beginning when most of the
sales are yet to come. Hence, the potential impact of
marketing expenditure on future cash flows is great-
est at this stage, which is why young products get a
higher allocation weight and thus a larger share of
the total budget.
Because of the growth potential of a new product the
optimal marketing budget might even be higher than
revenues at the beginning of its life. Therefore the solu-
tion may suggest spending money on products that
involve a temporary loss in such a case.
Adapting the Approach for Practical Application
For managers it is more transparent and easier to
use an allocation rule instead of a numerical solution
of such a model. Therefore, an allocation heuristic is
derived directly from the theoretical solution that pro-
duces near-optimal budgets, is easy to understand for
managers and can be implemented in a simple spread-
sheet. Basically, the proposed heuristic is a simple
proportional rule that integrates relevant information
from three areas:
> the long-term effectiveness of marketing invest-
ments in the focal product,
> the profit contribution of the focal product,
> and the focal product’s growth expectations.
Box 3 shows how the allocation weights are determined
using the simplified decision rule. Data for the carryover
coefficient, sales elasticity, and the growth multiplier are
not readily available but must be estimated.
Current values of revenues are available from last year
and the contribution margin is a target figure decided
by management. The growth potential is calculated as
a multiplier that divides expected revenues in 5 years
(planning horizon) by the current revenue level. By this,
products get a greater share of the total budget as long
as they are expected to grow. In contrast, when they are
expected to turn into their decline stage their budget is
By definition, the heuristic solution is likely to differ
from the optimal solution, but it should not deviate too
much to be useful. The performance of the heuristic was
tested in an experimental simulation study and found
to provide very good results which even improve after
several planning cycles and in some cases converge to
the optimum if applied consecutively.
Although the tool was applied to prescription drugs (see
below), it is suitable for many other industries, such as
consumer durables, consumer packaged goods, etc. In all
these markets, rich information is available at the aggre-
gate product level that allows the calibration of market
response models.
» The effectiveness of detailing
and other marketing activities
varies substantially across the
different therapeutic areas. «
GfK MIR / Vol. 4, No. 1, 2012, pp. 50 59 / Case Study
Company Background and Challenge
Bayer is one of the world’s leading companies in the
pharmaceuticals and chemicals business sector. As
of 2008, the company had EUR 32.9 billion sales and
around 108,600 employees (Bayer 2009). The company
invests substantial resources in marketing and sales
activities. Total marketing and selling expenditures were
EUR 7.6 billion (~23.1 % of total sales) in 2008. Bayer
consists of three major business areas, Healthcare being
the largest area in terms of sales (contributing almost
50 %). Within Healthcare, the Primary Care Unit (EUR
3.1 billion) is the largest within the prescription drug
business (EUR 10.7 billion). The unit operates in four
separate competitive market environments or therapeu-
tic areas, respectively: diabetes, hypertension, infectious
diseases and erectile dysfunction.
The challenge for the management was to fi nd a balance
in the allocation of marketing resources that trades off the
size of the business, the growth expectations, and even-
tually the effectiveness of marketing expenditure. The
main objective was to improve the process and results of
annual budget allocation in order to maximize discounted
profi ts from the product portfolio over a planning horizon
of ve years. The implementation of the allocation tool
was targeted at the ve main European countries which
contribute the largest share to total sales. The applica-
tion was developed in the period 2005-2006 and budget
recommendations for 2007 were derived.
At that point in time the three therapeutic areas dia-
betes, hypertension and infectious diseases represented
established areas which are in their saturation stage.
Due to the aging of the population in industrialized
societies and innovative new product introductions they
are, however, expected to continue to grow at moderate
rates in the future. The biggest challenge for Bayer in
these areas is to keep its market position. Existing and
new drugs by other global players are the main competi-
tors for the Bayer drugs.
In contrast, the market for the treatment of erectile
dysfunction is a new category that was pioneered by
Pfi zer with its Viagra brand in 1998. Bayer and Eli Lilly
followed in 2003 with the introduction of their brands
Levitra and Cialis. This market is still growing and does
not face generic competitors yet.
Data and Model Estimation
To obtain relevant input information such as sales elas-
ticities and growth parameters, the authors estimated a
market response model for each product market. Quar-
terly marketing and sales gures at product level of the
previous 10 years (1996-2006) were available. The mar-
ket response model is a mathematical representation of
how sales evolve over time and react to marketing and
other investments. Estimating the parameters of this
model from the observed sales time-series provides the
data input to compute marketing elasticity and other
input data, which are not observed. Bayer management
helped to identify the relevant subcategories and com-
petitors within each therapeutic area by country. Sub-
categories range from 12 for Anti-infectives to one for
Erectile Dysfunction. Products range from 15 for the Erec-
tile Dysfunction area and 306 for the Hypertension area.
Table 1 gives an overview of the key input variables used
to calibrate the heuristic allocation tool (see Table 1).
» A product with high marketing
effectiveness but a low profit
contribution level could get a lower
budget than a product with a a
high level of profit contribution but
lower marketing effectiveness. «
Case Study / Vol. 4, No. 1, 2012, pp. 50 – 59 / GfK MIR
Each therapeutic area is specified as a double-log sales
response function that accommodates nonlinear and
interaction effects. Marketing mix data in each area was
reflected by including marketing stocks (a combination of
all marketing expenditure types) for Bayer and its com-
petitors (in total), own and competitive prices and brand/
quality effects. A double-log market response model was
used to ensure diminishing marginal returns and get esti-
mated parameter values associated with marketing-mix
variables that correspond to elasticities, which indicate
the effectiveness of the specific activities. An elasticity
is a dimensionless measure of the relative change of a
dependent variable, such as sales, divided by the relative
change of an independent variable, such as the marketing
budget. Thus, if sales increase by 5 % when the marketing
budget is increased by 20 %, the elasticity is 5 %/20 % =
0.25. It can be compared across products, countries and
marketing instruments. Further, the model incorporates
a number of control variables that have been shown to
impact sales of pharmaceuticals, like order of entry, coun-
try or seasonal effects and asymmetric life cycle func-
tions. In-sample model fit and predictive validity were
very good across all four therapeutic areas.
The effectiveness of detailing and other marketing
activities varies substantially across the different
therapeutic areas. In general, they worked best in the
Erectile Dysfunction category, which is not surprising
as this category was the youngest category and still
in its growth phase. In detailing, visiting general prac-
titioners appears to work better than detailing at spe-
cialists and pharmacists. However, considering that
specialists only account for a share of approximately
20 % in Diabetes and around 27 % in Hypertension,
segment-specific specialist detailing effectiveness
is 4 – 5 times higher. These findings are consistent
with findings from other pharmaceutical studies. Own
price effects were significant, but price changes did
not have strong effects. The impact of competitive
marketing expenditure was negative across all thera-
peutic areas although it was not always statistically
significant. An earlier market entry was favorable, as
expected. Seasonal effects were only relevant to Anti-
infectives, which experience a high season in autumn
and winter.
Overview of Input Variables
for the Heuristic Allocation
Tool for Bayer
Method Antidiabetes Hypertension Erectile dysfunction Antiinfectives
Mean SD Mean SD Mean SD Mean SD
Unit sales in thousand standard units 16,319 20,674 11,891 16,649 1,008 649 5,291 8,004
Elapsed time since launch in years 14.50 12.69 10.00 7.42 2.75 1.91 12.25 10.45
Order of entry (Median) 3 4 2 3
Price in EUR per standard unit .16 .26 .50 2.96 7.00 .48 2.01 1.97
Marketing stock variables
Detailing at general practitioners in thousand EUR 22,519 36,566 64,595 87,134 55,026 30,326 44,259 34,930
Detailing at specialists in thousand EUR 2,081 4,068 8,803 13,701 14,498 12,771 10,380 11,353
Detailing at pharmacies in thousand EUR 588 1,453 1,930 3,039 1,766 2,598
Professional journal advertising in thousand EUR 149 341 458 502 165 295
Meeting invitations in thousand EUR 730 2,030 1,361 3,062 3,884 2,481 471 837
Other marketing expenditures in thousand EUR 2,558 9,278 3,912 4,404
# of countries 5 5 5 5
# of subcategories 6 10 1 12
# of products 104 306 15 100
# of observations 2,398 7,908 233 2,916
Notes: Al l units and EUR figures are on a quarter ly basis.
GfK MIR / Vol. 4, No. 1, 2012, pp. 50 59 / Case Study
The Bayer Implementation
To ensure that management can easily use the alloca-
tion formula in everyday business life, the authors
developed an Excel-based Decision Support Tool. The
tool provides budget scenarios and their implications
for the development of market shares and profi ts over
ve years and produces a recommendation for the allo-
cation of the total marketing budget. It uses input data
at the quarterly level.
The heuristic rule requires an allocation weight for each
marketing spending category and each drug to be com-
puted. Input data have been obtained either from econo-
metric analysis or internal records. The plausibility of input
data, especially the estimated sales elasticities, has been
extensively discussed with different groups of managers
in several workshops (global marketing, market research,
product management, sales management, controlling, etc.).
Following the needs of management, the tool was
extended in two ways. First, a threshold for product bud-
gets was included because of internal setup costs that
are xed at the product and marketing-activity level.
Second, manual adjustments to budgets recommended
by the heuristic were made possible. By this feature,
management could account for exogenous restrictions
to budget setting, e.g., to counter competitive attacks
in a predetermined way. In addition, it enabled manage-
ment to investigate the effects of budget scenarios on
market share and profi t as well as on the recommended
budgets for other products and marketing activities. The
tool is easy to use and fl exible enough to adapt to vary-
ing conditions of decision making. The effort to develop
and implement the budget allocation tool had signifi -
cant impact on managerial decision making.
Impact on Managerial Decision Making
> Providing Structure to the Allocation Problem
Obviously it is a challenging task to allocate a total
budget across six spending categories for 36 drugs
that are marketed in different countries and thera-
peutic areas. The suggested allocation heuristic gives
structure to this complex decision problem. It speci-
es that information and data from three fi elds are
necessary (data on the long-term effectiveness of
marketing, information on a product’s contribution to
profi t and growth potential of the product).
> Providing Solutions to the Problem
The allocation rule suggests that these three fi elds
of information are to be combined in a multiplicative
fashion so that the budgets are proportional to these
three information pieces. Implications from this rule
are straightforward.
(1) Products that generate more incremental sales
with the same budget should get a larger slice of the
total budget. Of course, relative incremental sales
tend to decline as sales and budgets increase due to
saturation effects.
(2) Products with a higher level of profi t contribution
generate more fi nancial resources to cover their own
marketing expenditure and contribute more to over-
all profi ts.
(3) Marketing should support growing and not declin-
ing products and shift resources over the life cycle.
The rule also teaches that the drivers of a product’s
near-optimal budget share interact with each other,
i.e., synergies exist between them. Finally, it makes
the trade-offs in budget allocation transparent. For
example, a product with high marketing effectiveness
but a low profi t contribution level could get a lower
budget than a product with a high level of profi t
contribution but lower marketing effectiveness. Even
though that product’s spending is less effective it
may still contribute more to overall profi t because of
its larger sales base.
» The allocation tool adds an
independent perspective and
its recommendations are fully
fact based.
Case Study / Vol. 4, No. 1, 2012, pp. 50 – 59 / GfK MIR
Understanding the Limitations of Separate ROI Analysis.
Profit calculations with the allocation tool quickly
revealed the limitations of comparing incremental
ROIs that result from raising/decreasing marketing
expenditure for individual products and market-
ing activities. First, separate ROI analyses for indi-
vidual marketing activities do not consider synergies
between marketing activities nor do they consider
the trade-offs that exist with respect to potential
profit improvements by other products and activities.
Further they do not inform about the optimal magni-
tude of budget changes for products and activities,
given a fixed total budget. All three requirements are
met by the allocation heuristic in one step.
> Organizational Impact
Although the allocation tool is not the only source
used by Bayer to generate budget options, it has sig-
nificantly improved the efficiency and quality of the
decision process. Because of its transparency and top
down perspective, the allocation tool ameliorates the
decision process that often appears emotional and
inefficient. Since it is strictly based on a range of veri-
fiable input information it adds an independent per-
spective and its recommendations are fully fact based.
The budgeting-project contributed substantially to
an organizational transformation that eventually
resulted into the creation of a completely new mar-
keting intelligence unit called Global Business Support.
This unit supports global marketing management and
sales including the global management board with
tools, results, and recommendations for a more effi-
cient and effective use of marketing resources.
> Last and Most Important: Financial Impact
The tool enables its users to simulate the financial
impact of different budget allocation options. By
analyzing the simulation results, it provides transpar-
ency about the impact of different assumptions on
financial results. Figure 1 shows an example of the
budget-shift recommendations of the model in the
hypertension market. A budget shift between the
two products implying an overall budget reduction
can cause increased profits for both products.
Example of Budget Re-Alloca-
tion Across Products in EUR
Discounted profit
product A
product B
+4.0 m
+6.7 m
Budget before
product A
product B
4.5 m
1.5 m
Budget after
product A
product B
2.2 m 2.3 m
GfK MIR / Vol. 4, No. 1, 2012, pp. 50 59 / Case Study
Albers, Sönke, Murali K. Mantrala, and Srihari
Sridhar (2010), “Personal Selling Elasticities:
A Meta-Analysis”, Journal of Marketing Research,
47 (5), 840 – 853.
Fischer, Marc, Peter S. H. Leeflang, and Peter C.
Verhoef (2010), “Drivers of Peak Sales for
Pharmaceutical Brands”, Quantitative Marketing
and Economics, 8 (4), 429 – 460.
Hanssens, Dominique M., Leonard J. Parsons,
and Randall L. Schultz (2001), Market Response
Models: Econometric and Time Series Analysis.
2nd ed., Boston et al.: Kluwer Academic Publisher.
Tull, Donald S., Van R. Wood, Dale Duhan,
Tom Gillpatick, Kim R. Robertson, and James G.
Helgeson (1986), “‘Leveraged’ Decision Making
in Advertising: The Flat Maximum Principle and
Its Implications”, Journal of Marketing Research,
23 (1), 25 – 32.
Marketing Budget, Optimization,
Resources, Allocation, Case Study
Based on the year 2007, the simulation suggested an
increase in discounted profits of 55 % over the next five
years due to an optimized allocation. This is worth EUR
493 million. In contrast, changing the overall budget by
20 % promised a profit impact of less than 5 %. Even if
only a small portion of this increase can be realized, the
additional profit for a business unit, such as Primary Care
with EUR 3 billion worldwide sales, is substantial.
Actual profit improvements are hard to evaluate. First,
management did not completely follow the suggested
reallocation by the tool for several reasons (e.g., vary-
ing personal experiences, concerns about errors in some
data from third party data providers). Second, activities
by competitors and exogenous influences on market
dynamics impact profit results. Nevertheless, the busi-
ness area Bayer HealthCare reports an increase in EBIT
of 12 % (EUR 273 million) compared to a 4 % revenue
increase for 2008 (Bayer 2009). Although there is no
validation from a field test, these results are consistent
with prior observations that reallocation really focuses
on the bottom line.
The innovative budgeting allocation approach provides
a simple but comprehensive heuristic that accounts
for dynamics in marketing effects and product growth.
Allocating a budget proportionally to the size of the
business (sales and profit contribution margin), the
effectiveness of the marketing activities (short-term
elasticity and carryover coefficient), and the growth
potential of the product (growth multiplier account-
ing for time discounting) revealed substantial profit
improvement potential compared to a simple allocation
dominated by rules of thumbs, separate ROI analysis for
different products or subjective evaluations. It is suitable
for many other industries, including consumer durables
and consumer packaged goods, provided that rich infor-
mation is available at the aggregate product level.
Case Study / Vol. 4, No. 1, 2012, pp. 5059 / G fK MIR
... They illustrate some of the models that can be easily built with modern spreadsheet packages, including models to support what-if or scenario analysis, sensitivity analysis, goal seeking and optimization. The Excel-based decision support model presented by Fischer et al. (2012) illustrates that it is possible to determine near-optimal marketing budgets. Their model accounts for marketing dynamics and trade-offs with respect to marketing effectiveness and profit contribution. ...
... Intelligent pricing decisions require a strong understanding of consumer demand, competing strategies and effective promotional approaches. "Marketing budget decisions are critical and should be fact-based rather than intuitive" (Fischer et al., 2012). Figure 1 illustrates the spreadsheet that is provided for the students which gives instructions for the case. ...
This paper describes a unique integrative business case appropriate for use in a managerial accounting course or other business courses related to economics or marketing. The case describes a scenario in which a managerial accountant is assisting in business decisions relating to factors influencing the profitability of a small manufacturing operation. Students are asked to optimize profits for company operations given information concerning costs, quality, and marketing options.
ResearchGate has not been able to resolve any references for this publication.