ArticlePDF Available

Collect them all! Increasing product category cross-selling using the incompleteness effect


Abstract and Figures

The familiar state of tension associated with an incomplete collection or an unfinished jigsaw puzzle is predicted by Lewin’s (1926; 1935) field theory. This feeling evokes a drive to completion—a phenomenon we label the incompleteness effect —which is useful to marketers endeavoring to cross-sell products and services. In three studies using online product configurators, we find that consumers faced with visual representations of incomplete product category collections, such as an evening drinks menu or a puzzle with its pieces representing services, are significantly more likely to complete the collection or finish the puzzle by cross-purchasing from a greater number of product or service categories as compared to those using a conventional online shopping format. We identify theoretical mechanisms through which the incompleteness effect works and potential moderators for the effect. Findings suggest that managers offering products or services across several categories can increase cross-selling by eliciting people’s drive toward completion.
Content may be subject to copyright.
Collect them all! Increasing product category cross-selling using
the incompleteness effect
Christoph Bauer
&Katie Spangenberg
&Eric R. Spangenberg
&Andreas Herrmann
Received: 25 November 2019 /Accepted: 17 December 2021
#The Author(s) 2022, corrected publication 2022
The familiar state of tension associated with an incomplete collection or an unfinished jigsaw puzzle is predicted by Lewins
(1926; 1935) field theory. This feeling evokes a drive to completiona phenomenon we label the incompleteness effectwhich
is useful to marketers endeavoring to cross-sell products and services. In three studies using online product configurators, we find
that consumers faced with visual representations of incomplete product category collections, such as an evening drinks menu or a
puzzle with its pieces representing services, are significantly more likely to complete the collection or finish the puzzle by cross-
purchasing from a greater number of product or service categories as compared to those using a conventional online shopping
format. We identify theoretical mechanisms through which the incompleteness effect works and potential moderators for the
effect. Findings suggest that managers offering products or services across several categories can increase cross-selling by
eliciting peoples drive toward completion.
Keywords Cross-selling .Completeness .Field theory .Incompleteness effect .Zeigarnik effect
Cross-selling is the act of selling different product offerings to
existing customers, thereby providing additional perceived
benefits to consumers as well as increased revenues to the
firm. The average profit associated with cross-buying cus-
tomers is greater than that of consumers not offered, or not
engaging in, cross-buying opportunities (Shah et al., 2012),
and the costs of encouraging existing customers to increase
purchases through cross-buying are much lower than those
associated with new customer acquisition (Hart et al., 1990;
Reichheld, 1996). Cross-buying customers are likely to be-
come more loyal as they purchase multiple items across prod-
uct categories (e.g., switching costs increase; Min et al., 2016),
and more loyal consumers are likely to revisit a successful
cross-selling firm as future needs arise. Relatedly, in the
non-profit sector, donors who have experienced the satisfac-
tion of making a donation that completes a project are more
likely to return to the platform and donate in the future (Argo
et al., 2020;Wash,2013). Cross-selling tactics are therefore
recognized as a key strategic challenge for marketing man-
agers. Herein, we demonstrate in two contexts (beverages
and banking services) how online configurators can create
incomplete collections of products or services in the minds
of consumers, thereby encouraging cross-buying purchases
that would not have otherwise been considered. This cross-
selling technique can benefit marketing strategists in various
organizations who desire to increase their customers
Stephanie Noble served as Area Editor for this article.
*Eric R. Spangenberg
Christoph Bauer
Katie Spangenberg
Andreas Herrmann
Simon-Kucher & Partners, Am Sandtorkai 74,
20457 Hamburg, Germany
Albers School of Business and Economics, Seattle University, 901
12th Ave, Seattle, WA 98122, USA
Center for Global Leadership, Paul Merage School of Business,
University of California Irvine, 4293 Pereira Drive, SB1 Ste 5300,
Irvine, CA 92697, USA
Institute for Mobility IMO-HSG, University of St. Gallen,
Bahnhofstraße 8, CH-9000 St. Gallen, Switzerland
/ Published online: 22 March 2022
Journal of the Academy of Marketing Science (2022) 50:713–741
propensity to engage in cross-buying of their respective goods
and services.
The current research builds from the contemporary schol-
arship of marketing and consumer behavior while drawing
from seminal theoretical work of psychologists almost a cen-
tury ago. Early in the twentieth century, Zeigarnik (1927,
1938) demonstrated that people possess better memory for
tasks they have started but not yet finished than for those
already completed. A consumption-related instantiation ofthis
Zeigarnik effectis exhibited by people collecting Panini
: a missing sticker space is much more salient in a
collectors memory than those already filled (see Fig. 1).
Perhaps due to such heightened awareness of missing prod-
ucts, it has been shown that consumers are less likely to pur-
chase a partially empty carton of eggs or a half full six-pack
beverage carrier than they are to choose completesets of a
given product, even if such incompleteproduct bundles are
correspondingly priced (e.g., an individual beverage is 1/6
the cost of a six pack; Barasz et al., 2017). This behavior is
consistent with the tendency of people to reengage with
interrupted action, complete a set, or fulfill a goal without
explicit instruction to do so (i.e., the Ovsiankina [1928]
Arguably underlying these effects is Lewins(1926,
1935) field theory, wherein the motivation to act or put
forth effort toward task completion is contingent upon a
state of internal tension caused by unmet needs or goals,
which can be eliminated by fulfilling a perceived quasi-
need through task completion (Lewin, 1926,1935). When
tasks are comprised of many steps (e.g., player stickers in
Panini albums), the goal gradient effect applies (i.e., people
accelerate behavior as they make progress toward their
goals) (Hull, 1932;Kivetzetal.,2006). As a marketing
example, research has shown that people fill car washing
stamp cards more quickly the closer they are to completion
(Nunes & Drèze, 2006).
Importantly, perceptions of incomplete sets or unfulfilled
goals can be created ad hoc and externally influenced by mar-
keters (Barsalou, 1983). For example, a bank may generate a
list of essential financial services,or a brand-related shop-
ping listmay accompany a recipe; such actions represent
marketer-generated to-dosfor consumers when such com-
binations may otherwise not have comprised peoplesconsid-
eration sets. In the current research, we test Lewins field
theory (1926,1935), which suggests a greater likelihood of
consumers making cross-buying choices to complete such
lists relative to conditions under which product or service
offerings are independently presenteda drive to completion
we refer to as the incompleteness effect.Although recent
work examines related concepts (i.e., pseudo-sets; Barasz
et al., 2017), research to date has not studied how aspects of
online product configurators influence cross-buying and set
We also build from recent work examining a critical driver
of consumer purchase behavior: customer inspiration. Among
the findings of Böttger et al. (2017), visual components of
advertising increase customer inspiration, increasing purchase
likelihood. We examine this finding in conjunction with those
of Cheema and Bagchi (2011) to show how visual (vs. textual)
cues can inspire people to consider cross-buying products they
may not otherwise.
Across three studies we demonstrate how online prod-
uct configurators can be used to present people with a
collection or puzzle they are inspired to complete, there-
by making several contributions to the literature. First,
we experimentally demonstrate that the completion im-
pulse can be elicited in online shopping contexts to in-
crease cross-selling. To our knowledge, this is the first
research to show that online product configurators can
effectively trigger the impulse to complete a set, thereby
increasing consumerslikelihood of cross-buying prod-
ucts to do so. Second, we examine drivers of the incom-
pleteness effect. Online shopping configurators reduce
the number of decisions a consumer must make, thus
increasing likelihood of set completion (Jin et al.,
nisms. Specifically, online configurators increase the
number of products considered (and potentially pur-
chased) without requiring consumers to decide whether
to continue shopping. We also show that the way the
product categories are grouped (individually vs. part of
an interrelated set) and presented (visually vs. textually)
increases perceptions of incompleteness and therefore set
completion impulse, thereby increasing the likelihood
that consumers will buy all products presented as a set.
Using a phased manipulation of product configurators,
we examine separate components of this process and find
that perceptions of incompleteness, impulse to complete
a set, and number of products considered all play a role.
Importantly, we use a phased manipulation of product
configurators to examine our complete model, thereby
avoiding exaggerating the effect of any single product
configurator manipulation.
Finally, we examine moderating effects identifying bound-
ary conditions for the incompleteness effect. While previous
research has shown some support for our hypothesized incom-
pleteness effect (Barasz et al., 2017), it has not examined
conditions under which this effect is more or less likely to
manifest, as we do. Notably, we study how perceived feasi-
bility and justification of purchase can moderate the incom-
pleteness effect. That is, when a consumer perceives no pos-
sible path to completing the full set or no ability to justify
Panini sticker collections, common in Europe, are players on sports teams
(e.g., World Cup soccer teams), much like baseball card collections in North
714 J. of the Acad. Mark. Sci. (2022) 50:713–741
purchasing all items, we see no increase in cross-buying pur-
chases or set completion. Such empirical findings advance
theory regarding goal pursuit and help managers implement
this effect in practice.
Lewins field theory set the stage for a long tradition of goal-
related work explored by scholars; he showed that setting a
goal can generate tension that can only be removed through
goal attainment and until a goal is attained, the incomplete
taskdiscomforts a person (Lewin, 1926,1935). This state of
tension associated with the desire to complete collections and
factors associated with alleviating said tension are demonstrat-
ed across multiple studies (e.g., Garland & Conlon, 1998;
Long & Schiffman, 1997). Scholars have more recently di-
rected effort to applying knowledge of goal setting and goal
completion (Locke & Latham, 1990) to questions regarding
consumption-related decisions and behaviors (e.g., Barasz
et al., 2017;Nunes&Drèze,2006). Originating from the work
motivated by Lewins(1926) field theory, our research most
closely follows a growing body of contemporary research by
marketing scholars and consumer psychologists. Table 1syn-
opsizes this marketing-relevant research, helps identify gaps
in the literature, and illustrates how our work contributes to the
theoretical and methodological understanding of this stream
of inquiry.
People are more motivated by how much is left to reach
their target rather than how far they have come (Hull, 1932).
Demonstrations of this goal gradient effect in marketing con-
texts include Kivetz and colleagues (2006) and Nunes and
Drèze (2006) who used bonus/stamp cards to show that the
time between visits to a café or car-wash facility became in-
creasingly shorter when customers neared the completion goal
of a full card. Importantly, these two papers examined how
overcoming the starting problem(i.e., whether a stamp card
was empty or already started with two stamps) changed com-
pletion likelihood. Partially addressing a gap in the literature,
we examine how grouping items together as a set can increase
completion even when all consumers start with an empty
shopping basket and items are not as closely related as repeat
café or car-wash visits. Relatedly, Cheema and Bagchi (2011)
found greater motivation to work toward goals by giving peo-
ple visual finish lines relative to their objectives (i.e., goal
persistence and effort increased as ease of goal visualization
increased). Specifically, Cheema and Bagchi (2011) study at-
chase-related) tasks are split into subgoals and/or become
more proximate with regard to level of task completion (e.g.,
commitment to savings). Incorporating their findings on visu-
al progress into our studies, we consider how the acquisition
of somewhat disparate products or services can be visually
Fig. 1 Panini sticker album for the 2018 Soccer World Cup
715J. of the Acad. Mark. Sci. (2022) 50:713–741
Table 1 Synopsis of literature on sets and goal completionrelevance and extensions of current work
Reference Independent variable(s) Dependent variable(s) Type of Studies Key Findings
Current work Product Presentation
(individual items, basic
configurator, as part of a set
or puzzle to complete,
visually linked)
Set completion,
cross-buying product
Presenting items as a set to be completed increases likelihood of cross-buying purchases to
complete the set
Visual presentation (using multiple instantiations) of an incomplete collection of product or service
categories increases impulse to complete a set, leading to increased likelihood of
cross-purchasing from a greater number of product or service categories
Feasibility and rationalization moderate the effect of set completion
NO COMPLETION REWARD How current work extends findings
Argo et al. (2020)
Whether a donation completed
Donation size Field
Donors contributed significantly more money
to reach exact fundraising target
Donors make donations that complete a
campaign significantly more frequently than
donations that reach any other given fraction
of the target
Studies completed in marketing context
Böttger et al. (2017)
inspiration, idea shopping,
ease of imagery processing
Customer loyalty,
emotional attachment,
Lab experiment,
Purchase goals can be externally induced via
marketing tactics
Customer inspiration increases when: 1) firms
present existing products in new or
unexpected combinations and 2) firms
increase ease of imagery processing in
Use of configurator to create consumer purchase
goals; extends finding regarding ease of
imagery; different dependent variables (i.e.,
actual product or service choice)
Barasz et al. (2017)
Task sets presentation
(individual items vs.
# of tasks completed,
donation likelihood, set
completion by
purchasing same items
Lab experiment,
People are more likely to complete arbitrary
pseudo-sets of tasks or purchases
This effect is mediated by perceived
Examines boundary conditions of set
completion; manipulates visualization of sets
(e.g., use of puzzlein configurator);
purchase likelihood of different (vs. similar)
Wash (2013)
Does donation complete
Donation size, repeat
Field data Donations that complete a project (on are more than twice as
large as other donations
Donors whose donations complete a project are
more likely to return to the platform to donate
Studies completed in marketing context
Cheema and Bagchi
Ease of visualization (hard vs.
Goal distance (near vs. far)
Goal framing (subgoal vs.
consolidated goal)
Exerted effort,
Lab experiment,
field data
When people are near their goal, those who can
easily visualize the goal exert more effort and
report greater commitment than those who
cannot easily visualize the goal
Ease of visualization increases effort and
commitment when people are near the goal
This effect is mediated by progress perceptions
This effect is moderated by goal distance and
goal framing
Addtotheeffectiveness of visualization
findings; set configurator establishes goal for
customers; different dependent variables (e.g.,
Garland and Conlon
Degree of project completion
(80% vs. 20%)
Sunk cost amount (80% vs.
Project fund allocation Lab experiment Subjects more willing to allocate resources to a
project when close to completion regardless of
sunk cost expenditures
Studies project completion-like effectin mar-
keting context; different dependent variables
716 J. of the Acad. Mark. Sci. (2022) 50:713–741
Table 1 (continued)
Reference Independent variable(s) Dependent variable(s) Type of Studies Key Findings
Project-completion effect stronger than
sunk-cost effect
Jin et al. (2013)JCR Goal adoption vs. goal
Goal structure (fixed sequence
vs. rigid)
Goal adoption, goal
Lab experiment,
Fixed-sequence goal-related actions discourage
goal adoption, but increase likelihood of goal
No completion bonus; different dependent
Kivetz et al. (2006)
Bonusstarter stamps vs.
empty stamp card
time, interpurchase
times, program
customer data
Intervisit purchase time decreases as consumers
near stamp card completion (reward)
Stamp cards with bonusstarter stamps
provide the illusion of progress, inducing
purchase acceleration
Stronger tendency to accelerate towards goal
increases reengagement likelihood and
No completion bonus; different dependent
Nunes and Drèze
Task framing (started vs. not
Endowment reason (given vs.
not given)
Medium (points vs. purchases)
Reward card redemption,
time to card completion
Lab experiment,
Framing a task as started but incomplete leads
to greater likelihood of completion than a
similar task that has not yet been started
This finding is moderated by providing a reason
the task has already been started and the
medium of the reward (points vs. purchases)
No completion bonus; different dependent
Long and Schiffman
Conceptual Tension builds when collections are incomplete
As items in the collection are acquired, tension
is relieved
Empirically examines effect arguably
attributable to tension reduction
Hull (1932)Psych
Distance from endpoint Section completion time Animal lab
As rats got closer to the endpoint, they ran
faster, suggesting a greater motivation to
complete the goal
Human versus animal subjects
717J. of the Acad. Mark. Sci. (2022) 50:713–741
framed such that the purchase of a combination thereof con-
stitutes a goal that consumers are more likely to pursue under
certain conditions.
Our general expectation is that a completion impulse is trig-
gered by an incomplete set of products, thereby eliciting a
desire to complete a goal, arguably eliminating the discomfort
associated with failing to do so. Consistent with this notion,
Barasz et al. (2017) showed that framing products as elements
of pseudo-sets (e.g., a six-pack of beer) leads to greater likeli-
hood of purchasing multiple items, at least in part because such
pseudo-sets influence perceptions of completion. Although the-
oretically similar, we differentiate our paper from Barasz et al.
(2017) and add to the literature in two primary ways. First, set
completion in our studies require purchasing from different
product categories (e.g., not only wine but also juices and grap-
pa, or not only daily banking services but also insurance and
pension products) as opposed to purchasing more of the same
product (e.g., beer). Second, we examine boundary conditions
wherein set completion is more or less likelycircumstances
yet to be reported in the literature.
Goal-derived categories (Barsalou, 1991) can be estab-
lished intrinsically by consumers (Böttger et al., 2017)orby
extrinsic (e.g., marketing-introduced) cues (Ratneshwar et al.,
1996). For example, a pre-configured product such as a pizza
or a car can establish a completeproduct or goal in con-
sumersminds, even though no actual purchase has been
made (cf. Levin, 2002). Such categories can be obviously
related (e.g., fruitor furniture) or created ad hoc (e.g.,
drinks for an entire eveningsdinnerparty). In either case,
both types of categories can be structured in such a way that
makes them susceptible to manifestation of Hulls(1932)goal
gradient principles. The decision to continue pursuing a goal
and decisions within a goal step require different cognitive
orientations (Brandstätter & Schuler, 2013) and transitioning
between these two types of decisions (i.e., shifting between
two mindsets) can be extremely taxing (Hamilton et al., 2011).
In addition to mindset transitions, each subtask decision also
requires cognitive effort (Vohs et al., 2008). For example,
when purchasing wines for a dinner, one must consider the
budget for each item, how the items pair (both with each other
and with the food), as well as their dinner guestspreferences.
An excessive number of decisions could delay purchase or
reduce likelihood of completion (Dhar, 1997;Iyengar&
Lepper, 2000).
To counteract the negative effects of excessive choice,
companies can restrict decision-making in certain areas of
the shopping process. The current work operationalizes an
online configurator, an approach yet untested in set comple-
tion research. Configurators automatically move consumers to
the next goal-directed action when the preceding action (i.e.,
previous product choice) is completed, thus eliminating the
need for consumers to decide at each step whether to continue
goal pursuit (i.e., Should I keep shopping?). Instead, people
remain in an implemental mindset making subtask product
purchase decisions (e.g., which grappa to choose) more effi-
ciently (Gollwitzer, 1999). Thus, a fixed sequence advancing
consumers forward can facilitate goal completion by reducing
the mindset transitions and cognitive effort that may otherwise
dissuade consumers from completion (Jin et al., 2013).
Our primary contention derived from the literature and ob-
servation of consumer behavior is that using a configurator to
represent products as sets, collections, or puzzles to be
completedthat may not have otherwise been considered as
suchcan increase consumer purchase behavior to alleviate
feelings of tension associated with incompleteness on the way
to achieving goal completion. This approach can usefully be
applied to marketing contexts, such as with easily developed
online product configurators that guide consumers through the
decision-making process. In addition to our workspractical
implications, the present research makes several theoretical
contributions. First, we extend research on the goal gradient
effect (Hull, 1932; Kivetz et al., 2006)bytestingnewfactors
(e.g., how items are grouped together) that also increase goal
completion likelihood. Further, we use a novel product
configurator manipulation and combine research on subtask
decision making (Vohs et al., 2008)withthatoftaskcomple-
tion (Jin et al., 2013). Finally, we examine unexplored bound-
ary conditions identifying situations under which set comple-
tion is less likely to occur. We turn now to specific hypotheses
regarding our work.
No research to date has tested our modelled effects using an
online configurator. We therefore manipulate an online
configurator to progress step-by-step through our model, test-
ing our predictions on the dependent variable of cross-buying
consumer choice. A conceptual model of our predictions and
where each is tested in our three studies is shown in Fig. 2.
First, the guided nature of a configurator should reduce the
effort required for consideration of any single product, there-
by increasing likelihood of tasks completed (Jin et al., 2013).
Second, a feeling of incompleteness is more likely to mani-
fest if there is something to complete (i.e., the products are
presented as related to each other in a set). And third, using
visual representation to show how something is incomplete
should magnify perceptions and feelings of incompleteness.
Thus, to more fully understand the processes driving the
incompleteness effect, we manipulate our product
configurator in three phases: (1) presenting product offerings
together using a configurator is the simplest effect; (2) pre-
senting product offerings as interrelated categoriesor a
setof offerings is the second level of creating a clear con-
sumption or purchase goal (Böttger et al., 2017); and 3)
718 J. of the Acad. Mark. Sci. (2022) 50:713–741
visual presentation of the missingelements of a category
should strengthen likelihood of a consumer desiring to com-
plete the set or finish the puzzle(Cheema & Bagchi,
2011). Our first goal was to establish these three components
of the effect. Thus, the following main hypothesis is pro-
H1 People are monotonically more likely to purchase a complete
group of products if the options are A) presented in a
configurator (vs. as individual product categories), B) pre-
sented as elements of an interrelated larger whole (vs. basic
configurator), and C) presented using visual (vs. textual)
We methodologically separate the three components to
parse the potential influence of each of the three effects rela-
tive to a baseline purchase rate, thereby avoiding
overestimating the influence of one of them in isolation.
Formal statement of our further hypotheses follows elabora-
tion on the proposed mechanisms below.
Presentation in a product configurator First, we discuss the
context of cross-selling using a configurator wherein products
are presented as distinct, unrelated productscustomers are led
through categories and/or customization options in a stepwise
manner, forced to consider and make a conscious choice regard-
ing every product offering or customization option for at least a
few seconds. Although such a seemingly simple mechanical
approach motivates further shopping, at its core a configurator
typicallypresentsproductsasbeing interrelated in some fashion
(analogous to a Panini collection of world cup soccer player
stickers), harkening back to our earlier discussion of Lewins
(1926,1935) field theory. The likelihood that customers develop
a purchase interest in any of the product offerings is thereby
higher compared to conventional online presentations, where
products are presented in a way that allows customers to avoid
exposure to one or more offerings and thereby limit their shop-
ping time. The average number of observed categories should
increase with a gradual product configuration solution forcing
consideration (Bordalo et al., 2013) as compared to more con-
ventional forms of marketing (Shih et al., 2019). Thus, we pro-
pose that more categories will be considered, leading to an in-
crease in set completion likelihood, if a configurator is used (vs.
non-configured online shopping). Formally:
H2 The guided nature of an online configurator
(vs. presentation as individual product categories) leads
consumersto consider more product categories, thereby in-
creasing likelihood of set completion.
Presentation of products as part of an interrelated set
Consider again Panini collectionssticker spaces for all team
players (e.g., on world cup soccer teams) in an album. This
combining of teams and players alone increases the probabil-
ity that a fan will engage with the teams and each player as
they scroll through their album. Loosely combining players
would not work in the same manner as presenting them as part
of a larger whole.Zeigarnik (1927,1938) and Ovsiankina
(1928) delineated between clear-cut goals (i.e., a task with a
clearly specified end/goal) and indefinite goals (i.e., the end is
not clearly predictable) as an important moderator regarding
task memory and resumption in a prematurely interrupted
task. Zeigarnik (1927,1938) found recall for interrupted tasks
Product Presentation
Individually presented
Basic configurator
As part of interrelated set
Linked visually
Perception of
Number of
A, B
B, C
C, D
H1, H5
C, D
Set Completion
Cross-buying products
Product Evaluation
Time spent evaluating
cross-buying products
Fig. 2 Conceptual model. Note. Study 1 tests H1H5;Study2testsH1;Study3testsH6H7.
superscripts denote Study1 conditions compared to
test indicated hypotheses
719J. of the Acad. Mark. Sci. (2022) 50:713–741
to be much greater for those with clear-cut goals versus those
with indefinite points of completion. Similarly, Ovsiankina
(1928) found clear-cut goals to be twice as likely to be re-
sumed versus indefinite actions. It is difficult to experience a
sense of incompleteness if one does not know what constitutes
completion. Barsalou (1983,1991) provides ample evidence
that people create categories when they might not otherwise
exist (i.e., ad hoc categories). Marketers can similarly create
ad hoc collections (sometimes across categories) in con-
sumersminds (Ratneshwar et al., 1996). From a configura-
tion standpoint, product offerings can then be portrayed as an
interrelated program with a specified number of elements,
with the goal being completion of the collection by purchasing
each element. Explicitly linking products to one another in a
product configurator will increase consumersimpulse to
complete the set, increasing attention paid to all of the offer-
ings, thereby increasing likelihood of consumers buying
moreor the full set ofproducts offered. Suggesting the
products are part of a larger whole(vs.loosely related items)
allows the consumer to see that there may be a set that needs
completing, which would increase product consideration and
set completion. Formally, we hypothesize:
H3 Presentation of products as a set of interrelated products
(vs. presentation as distinct, unrelated items) increases the
impulse to complete the set, leading to consideration of
more product offerings, resulting in increased likelihood
of set completion.
Presentation of products as visually linked in a set In the
Panini sticker albums, a placeholder with a certain number is
created for each sticker (e.g., in Fig. 1, the missing sticker is
labeled #441 [out of 690 for the FIFA 2018 World Cup]).
Empty, numbered placeholders provide obvious visual re-
minders of a collections incompleteness. Without this num-
bered placeholder, a missing sticker of the collection would
be less obvious, if not unnoticed. Zeigarnik (1927)and
Ovsiankina (1928) both found visualization of incomplete-
ness to be an important moderator to the urge to complete.
Zeigarnik (1927) notes that the aesthetic structure of a com-
pletion task (e.g., visual wholeness), and consequently, the
visual clarity of incompleteness influenced capability to
remember an uncompleted task. Relatedly, Ovsiankina
(1928) described that some trial participants reported the
eye catchingcharacter of an obviously unfinished work
as reason for task resumption. Consistent with these early
observations, Böttger et al. (2017) found people to be in-
spired to pursue consumption-related goals when induced
by marketing stimuli, while Cheema and Bagchi (2011)
found a visually presented progress bar to increase effort
and commitment toward goal completion (i.e., saving for a
trip). They showed that distance-to-goal was perceived as
lower shortly before completion with the progress bar
visualization tool motivating greater savings ratesafind-
ingalsoconsistentwithLewins(1926,1935) field theory.
Therefore, we predict visual (vs. textual) cues will magnify
perceptions of incompleteness, which increase the comple-
tion impulse, and result in more products being considered.
Formally, we hypothesize:
H4 Visual (vs. textual) depiction of missing products will
increase perceptions of incompleteness, increasing the
impulse to complete a set, leading to consideration of
more product offerings, resulting in greater likelihood
of completing a set.
Effect on cross-buying products Next, we should also see
consumers evaluating cross-buying products differently than
the initially considered products as a result of marketing efforts
of the firm (Kumar et al., 2008), such as visual representation. In
a dinner setting scenario for example, people commonly shop for
red wine, white wine, and sparkling wine, whereas juice and
grappa are cross-buying products that completethe drink of-
ferings of the party. Without clear suggestion provided/created
by marketers (e.g., drinks for every dinner course), the under-
standing of what constitutes completion of ones set of drinks for
the evening is less likely to occur in consumersminds. Thus, the
(amendable) incompleteness of a collection visually represented
by placeholders on a product configurator (akin to missing team
member stickers in a Panini album) creates an incomplete set,
making consumers more likely to purchase (i.e., complete) the
Nunes and Drèze (2006) found consumers to have an in-
creased interest in completing a stamp card as evidenced by
less time passing between purchases as peoples stamp cards
neared completion. Relatedly, we would expect consumers to
exhibit different shopping behavior for products that are pur-
chased last in a set of choices (i.e., the cross-buying products
that complete a set). As people fill their baskets with products
from primary categories (e.g., red, white, and sparkling wines)
their shopping task nears completion. A configurator effec-
tively reduces the amount of effort directed at whether con-
sumers will shop for beverages beyond the wines they have
selected; rather than expending effort deciding whether to
continue shopping or not, they are automatically advanced
to the next product, where they can decide whether or not to
purchase an item, especially one they may not otherwise have
considered. This automatic exposure to visually depicted
cross-buying product choices should lead to increased goal-
related (i.e., set completing) cross-buying choice behavior
(Cheema & Bagchi, 2011). Thus, motivation to complete the
set could manifest through increased attention to, and more
positive evaluation of, visually presented cross-buying prod-
ucts (e.g., juice and grappa) in a configurator. Formally:
720 J. of the Acad. Mark. Sci. (2022) 50:713–741
H5 Visual salience of incompleteness (vs. textual grouping)
will affect behavior towards cross-buying products such
that consumers will A) provide more positive evalua-
tions, B) spend more time evaluating, and C) be more
likely to purchase the cross-buying products.
The conceptual model in Fig. 2illustrates the relationships
among constructs including the hypothesized mediators and
moderators, which we test across three studies. In Study 1 we
examine the incompleteness effect in a vineyard product sales
scenario. Study 2 replicates Study 1 in the context of banking
services, where professional service firms stand to enormous-
ly benefit from an increase in cross-selling to existing cus-
tomers. In Study 3 we test moderating Hypotheses 6 and
7discussed in detail preceding that studyin a banking
services field study with a different visual incompleteness
manipulation than that of the first two studies.
Study 1: Vineyard product sales study
Participants and procedure
This study used a real online shopping site for vineyard prod-
ucts to test our hypotheses. A commercial market research
firm screened participants to ensure participants had wine-
related beverage needs and wants (i.e., the consumers had
invited guests for a meal to their home or restaurant in recent
years and indicated a likelihood of purchasing wine over beer
for such occasions); fewer than 15% of those contacted qual-
ified. Participants were given the following purchase scenario:
You want to celebrate your birthday and you have invit-
ed 10 friends to an evening meal/dinner to your home.
You and your friends are Italy fans. You have become
besotted with Tuscany in particular. Therefore, you
want to buy drinks for dinner from one of the most
renowned vineyards of Tuscany. You have already had
good experiences with this vineyard in the past. The
vineyard is also well known by many of your friends.
The online store of the vineyard you want to use for your
order has been (anonymously) activated for you. Please
fictitiously order drinks there for your birthday dinner.
Please assume that you have the required glasses for
each respective drink or that you can borrow them from
a good neighbor. By clicking on save and send in the
shopping cart you complete this test purchase. At the
end we briefly ask you about your test purchase.
Eight hundred participants were randomly assigned to one
of four online shopping conditions; seven were dropped for
incomplete surveys leaving a final sample size of 793. Screen
shots associated with each condition are shown in Figs. 3a to
3d. In each figure, the top picture is the start screen of the
online shop, the bottom left shows the display after selection
of first product, and the bottom right is the overview display
after purchase (or non-purchase) in each of the five product
categories. Each example screenshot in Figs. 3a to 3d shows
the primary products (i.e., sparkling wine, red wine, and white
wine) having been selected but not the cross-buying products
(i.e., juice and grappa).
Condition A served as a baseline group who viewed a
conventional online layout (see Fig. 3a). When participants
clicked Start,they began on the sparkling wines category,
where they could browse and select one of three sparkling
wines and the quantity. They could also visit other product
categories using the product menu on the left side of their
screen. Each product category had three product offerings,
of which only one product could be selected. Price and quality
always increased in order from left to right, consistently la-
beled Classico (left), Selezione (center) and Eccelenza (right).
By clicking Add to Cartparticipants were sentas is com-
mon with online storesto the shopping cart (see Fig. 3a
bottom left). There, they saw their current selections and ad-
ditional products that may be of interest to them (i.e., products
from unpurchased categories). Thus, even in the condition
without the configurator, we ensured that all participants saw
all products offered, consistent with the three experimental
conditions. Participants could click Shop more productsor
click directly on a remaining product to view the page. With
each additionally selected product, the list of products below
the shopping cart shortened until it was empty (i.e., participant
had selected a product from each of the five categories).
Importantly, however, participants did not have to select a
product from every categorywhen they were done, they
clicked Save configuration and send.
In Condition B, the five product categories were presented
as five individual items in a configurator (see Fig. 3b). As in
Condition A, participants started on the sparkling wines and
selected one of the three product options. In this condition,
once the customer clicked Select,the name of the selection
appeared next to the category name in the configurator (e.g.,
Eccelenza, bottom left of Fig. 3b). Participants could also
click No selectionor Next,and would move to the next
product category (or the overview page, if they were in the last
category). On the overview page, product categories where no
explicit decision had been made were labeled, No [missing
product category].In this respect, the incomplete nature of
the selection in the configurator was textually (vs. pictorially)
displayed for participants, presenting a relatively difficult
form of visualization (Cheema & Bagchi, 2011). Although
there was not a specific Shop more productsbutton in
Conditions B, C, or D, participants could return to any of
the five categories to change their selection or click on the
Go to Shopping Cart.Total price for each product category
was identical across all conditions.
721J. of the Acad. Mark. Sci. (2022) 50:713–741
Condition C was identical to Condition B with one modi-
fication: in Condition C we changed the title and labels of the
product categories to present the five product categories as an
interrelated shopping task (see Fig. 3c), wherein each product
choice was presented as part of a larger whole.First, the
page title was changed from Our product categories at a
glanceto Our dinner product programeverything for your
dinner.Second, the product categories were renamed to tell a
connected story around the ordinary course of a dinner as
opposed to simple descriptors (e.g., Our Sparkling Wine
was changed to A welcoming sparkling wine;see
Methodological Appendix for all changes). Everything else
(e.g., products, prices) was the same.
Although Condition C provided a completion of a larger
wholetask, a feeling of incompleteness is more likely to be
experienced when a completion goal and the (temporary) incom-
pleteness are made salient. Thus, Condition D was identical to
Condition C with the addition of a Panini column
on the right
side of the configurator (see Fig. 3d), to ease visualization of the
completion goal. Rather than representing the selection (or non-
selection) as text (such as in Conditions B and C), digital
stickerswere used as placeholders for each product choice
(i.e., Classico, Selezione, or Eccelenza) in a category. Thus, cor-
responding placeholders remained empty with non-selection,
akin to a partially filled Panini sticker album. We also incorpo-
rated a progress bar visually depicting level of completion.
Following Cheema and Bagchi (2011), we developed this visual
manipulation in consultation with expert colleagues, landing on a
combination of placeholders and a progress bar emphasizing
missing elements to ease visualization of a product program
completion goal. In all conditions, purchases and time spent were
recorded for participants upon their clicking Save configuration
and send.Participants then completed a questionnaire with mea-
sures of quality, predicted mediators, and manipulation checks
(see Table 2for all measures).
Interrelated set manipulation check In Study 1, we used
two manipulation checks designed to assess the changes be-
tween conditions because different things were manipulated in
each condition. From Condition B to C, the only difference
was the verbiage in Condition C linking products together (To
what extent doyou perceived the products as partof a product
program that is designed to be fully utilized? 1 = Not at all,
10 = Very Much). As expected, a oneway ANOVA on our
treatment Conditions B-D revealed that embedding products
in a dinner program resulted in significantly greater perception
that the products existed as part of a larger product program
(F(2, 593) = 9.23, p< 0.001). Contrasts revealed that both
Conditions C and D were perceived to exist as part of a larger
product program relative to Condition B (M
=6.29,SD =
2.41; M
=7.03,SD =2.08;M
=7.17,SD =2.11;Bvs.C:
F(2, 593) = 11.56, p= 0.007; B vs. D: F(2, 593) = 15.84, p
< 0.001), but there was no difference between Conditions C
and D (p=0.55).
Visualization manipulation check Moving from Condition C
to Condition D, we added visual emphasis (cf., Cheema &
Bagchi, 2011), predicting it would increase feelings of incom-
pleteness (How noticeable was the completeness [incomplete-
ness] of the program based on the visual presentation in the
product configurator? 1 = Not noticeable at all,10 = Very
noticeable). A oneway ANOVA confirmed that salience of
incompleteness differed between groups (F(3, 789) = 44.84,
p< 0.001). Participants in the Panini condition (Condition D)
=7.56,SD = 2.12) reported significantly higher salience
of incompleteness than those in Condition C (M
=2.53;F(3, 789) = 19.54, p< 0.001), Condition B (M
6.03, SD = 2.43; F(3, 789) = 38.69, p< 0.001), or Condition
=5.73,SD =2.64;F(3, 789) = 55.06, p<0.001).
Those in Condition C also reported significantly higher salience
of incompleteness compared to Condition A (F(3, 789) = 9.18,
p= 0.003); there were no other significant differences.
Main analysis We first tested the effects of our manipulations
on completion of the product program (i.e., purchasing a product
from each of the five categories). A chi-square test of indepen-
dence showed a significant effect of condition on completion
(3, N= 793) = 147.20; p< 0.001, φ= 0.43). Hypothesis 1
predicted a monotonically increasing percentage of participants
purchasing an item in all five product categories. Consistent with
this prediction, 20.3% of participants completed the program in
Condition A, while in Conditions B and C, 37.7% and 55.7% of
participants respectively made selections from each category.
Finally, in Condition D, 78.6% of participants completed the
programalmost four times as many as when product catego-
ries were presented individually. Further, the percent of com-
pleted programs in each condition was significantly different
from each other condition (all ps < 0.001) (see Table 3).
Next, a oneway ANOVA revealeda significant main effect
of condition on the average number of product categories
purchased (F(3, 789) = 69.72, p< 0.001, η
= 0.21).
Planned contrasts showed significant differences between
groups, supporting the monotonic increase predicted by
Hypothesis 1. Specifically, participants in Condition B (M
=3.68,SD = 1.32) purchased more products than those in
Condition A (M
=3.11,SD =1.34;F(1, 789) = 25.54; p<
0.001) and participants in Condition C (M
=4.30,SD =
1.34) purchased more products than those in Condition B
(F(1, 789) = 29.34; p< 0.001). Participants in Condition D
=4.63,SD = 1.34) purchased more products than those
in Condition C (F(1, 789) = 8.59; p<0.001).
We use Panini column as colloquial shorthand to label this manipulation as it
conceptually corresponds to the idea of an obvious, visually missing sticker in
a Panini sticker collection album.
722 J. of the Acad. Mark. Sci. (2022) 50:713–741
Fig. 3 3a. Study 1 Condition A: Individually presented product cate-
gories. 3b. Study 1 Condition B: Basic configurator. 3c. Study 1
Condition C: Configurator with products as part of an interrelated set.
3d. Study 1 Condition D: Configurator with products presented as an
interrelated set and increased visual salience
723J. of the Acad. Mark. Sci. (2022) 50:713–741
Fig. 3 (continued)
724 J. of the Acad. Mark. Sci. (2022) 50:713–741
Fig. 3 (continued)
725J. of the Acad. Mark. Sci. (2022) 50:713–741
Fig. 3 (continued)
726 J. of the Acad. Mark. Sci. (2022) 50:713–741
Product categories presented individually (Condition A)
vs. basic configurator presentation (Condition B) Recall
that our product configurator offered identical products at
identical prices but embedded the products in a configurator
guiding consumers through each product category.
Hypothesis 2 predicted the number of products purchased
would increase because a configurator (vs. conventional on-
line store) forces consumers to browse more product catego-
ries. We compare Conditions A and B to test the basic effect of
using an online product configurator. In Condition A, the
number of purchased products (M=3.11)wasalmostequal
to the number of product category pages viewed (M=3.55),
suggesting participants did little browsing and only viewed
products relevant to them. The grappa, for example, was
viewed by fewer than 50% of participants in Condition A. In
Condition B, however, participants were guided through the
products by the configurator (although it was possible to skip
a category). Indeed, the number of product categories consid-
ered was significantly higher in Condition B (M
= 0.58) than in Condition A (M
=3.55,SD =1.34;F(1,
394) = 138.32;p <0.001,η
= 0.26). Thus, it seems that
moving through product categories via a configurator under-
lies the positive effects of product configurator use on cross-
buying. Testing the causal relationship between the number of
Table 2 Study 1 Measures
Measures Anchors
Condition A vs. B
# of Categories Considered Number of category pages browsed (max. 5)
Condition B vs. C
Manipulation Check
Product Program
To what extent do you perceive the products as part of a product programthat is designed to be
fully utilized?
1=Not at all
10= Very much
Completion Impulse To what extent did you experience the impulse to buy a product in all five categories? 1 =Not at all
10= Very much
# of Categories considered Number of category pages browsed (max. 5)
Condition C vs. D
Manipulation Check
Visual Emphasis of
You have [not] completed the dinner program: How noticeable was the completeness
[incompleteness] of the program based on the visual presentation in the product
1=Not noticeable at all
10= Very noticeable
Perception of Incompleteness To what extent did the non-selectionof products leave the impression of an incomplete product
1=Not at all
10= Very much
Completion Impulse To what extent did you experience the impulse to buy a product in all five categories?
# of Categories Considered Number of category pages browsed (max. 5)
Product Evaluations
How attractive was the juice you viewed? 1= Not attractive at all
7= Very attractive
*Participants could
select not sure
How attractive was the grappa you viewed?
Potential Confounds (α=.94)
Fairness How do you evaluate the product configuration regarding FAIRNESS? 1=Very bad
7= Very goodComplexity How do you evaluate the product configuration regarding SIMPLICITY?
How do you evaluate the product configuration regarding CLARITY?
User Friendliness How do you evaluate the product configuration regarding USABILITY?
How do you evaluate the product configuration regarding LAYOUT?
Satisfaction & Quality
Quality How did the online store you viewed compare to other online stores known to you? 1= Much worse
7= Much better
Satisfaction How satisfied were you with the configured solution? 1= Very dissatisfied
7= Very satisfied
All variables were translated to English from original participant questionnaire in German
727J. of the Acad. Mark. Sci. (2022) 50:713–741
observed categories and set completion using PROCESS
(Model 4; Hayes, 2017), we found that the indirect effect of
condition on completion through number of product catego-
ries viewed was significant (b=5.34,SE=7.73,95%CI:
3.62, 23.53; see Fig. 4), confirming Hypothesis 2.
Basic configurator presentation (Condition B) vs.
Products presented as part of interrelated set (Condition
C) We next compared Condition B with Condition C to test
theimpactofofferingproductsaspartofalarger wholeor
set in a configurator. Our theoretical reasoning predicted that
when products were embedded in a larger whole, participants
would experience a greater impulse to complete the program
(To what extent did you experience an impulse to buy a prod-
uct in all five categories? 1 = Not at all,10 = Very Much).
Indeed, participants in Condition C reported a significantly
stronger impulse to complete the program (M
= 2.71) than those in Condition B (M
=5.38,SD =2.86;
F(1, 398) = 10.13, p= 0.002). Next, we tested our hypothesis
that an increased impulse to complete the program would lead
participants to consider more categories and ultimately com-
plete the program. Using PROCESS (Model 6; Hayes, 2017),
we found a significant indirect effect, as the confidence inter-
val did not contain zero (b= 5.34, SE = 1.75, 95% CI: 2.08,
8.92), supporting Hypothesis 3 (see Fig. 5). Thus, embedding
product categories in an interrelated larger wholeprogram
was more likely to activate a completion impulse that in turn
led participants to consider more product categories, and
eventually to purchase the full set of products offered.
Although participants felt a stronger completion impulse in
Condition C associated with the larger story connecting the
products, the textual depiction of a non-selected product did
not elicit a strong effect of incompleteness. Perceived sense of
incompleteness (To what extent did the non-selection of prod-
ucts leave the impression of an incomplete product program?
1=Not at all,10 = Very Much) did not significantly differ
between Conditions B and C (F(1, 399) = 0.46, p=0.50).
However, we predicted this stronger feeling of incompleteness
to manifest in the Panini condition, which we test next.
Products presented as part of interrelated set (Condition
C) vs. Products linked visually (Condition D) Recall our
prediction that visual salience of incompleteness should lead to
greater perceptions of incompleteness using the Panini column,
thereby strengthening the completion impulse, leading to
consideration of more product categories and resulting in higher
completion percentages. A PROCESS model (Model 6; Hayes,
2017) found the indirect effect of condition on completion
through: (1) perceptions of incompleteness, (2) completion im-
pulse, and (3) number of categories considered, was not signifi-
cant (b= 0.0004, SE = 0.009; 95% CI: -0.02, 0.02). Looking at
each indirect path, however, we observed that the effect of con-
dition on completion through (1) perceptions of incompleteness
and (2) completion impulse was positive and significant (b=
0.06, SE = 0.03; 95% CI: 0.02, 0.12; see Fig. 6). We attribute the
lack of a significant indirect effect through number of categories
considered to a ceiling effect. Indeed, the mean number of cate-
gories considered for both conditions C and D neared the max-
imum (4.93 and 4.97 categories, respectively). Thus, we find
partial support for Hypothesis 4.
Cross-buying products As predicted by Hypothesis 5, differ-
ences between conditions for purchases and attractiveness
should be pronounced for the cross-buying products as com-
participants in Condition D purchased grappa and juice as com-
pared to those in Condition C (Grappa: 87% vs. 77%; F(1, 395)
=7.02,p= 0.008, η
= 0.02; Juice: 91% vs. 76%; F(1, 395) =
18.39, p< 0.001, η
= 0.04). Additionally, as predicted, all
participants in Condition D spent more time looking at grappa
and juice than participants in Condition C (Grappa: M
18.89 s; M
= 15.30 s; F(1, 395) = 6.11, p= 0.01, η
0.02; Juice: M
= 24.15 s; M
=20.28s;F(1, 395) = 4.62,
p= 0.03, η
= 0.01). Finally, participants evaluated grappa
significantly more positively (1 = Not attractive,7 =
Attractive) in Condition D versus Condition C (M
=4.82;F(1, 367) = 7.99, p=0.011;η
= 0.02) and the
difference between conditions for juice was marginally signifi-
cant in the predicted direction (M
= 4.96, M
380) = 3.69, p= 0.057; η
In contrast, there were no
differences across conditions for any of the primary offerings
regarding purchase rate or attractiveness (see Table 4for all
results). Thus, Hypothesis 5 was supported.
Potential confounds Consultation with managerial and academ-
ic marketing professionals before running Study 1 suggested
some possible confounds. Specifically, the Panini column may
be perceived as: (1) an unfair representation of products, (2) not
Participants who selected Not sure(N=28forgrappa,N=15forjuice)
regarding attractiveness of each product were excluded from this analysis.
Table 3 Study 1 Results
supporting Hypothesis 1 Condition A Condition B Condition C Condition D
Participant N 195 199 201 196
Percentage of participants purchasing
complete set of all 5 products
Mean (SD) number of product categories
purchased (maximum 5)
(1.34) 3.68
(1.32) 4.30
(1.34) 4.63
Different subscripts reflect significant differences between groups at p<.01
728 J. of the Acad. Mark. Sci. (2022) 50:713–741
user friendly, or (3) more complex than typical online stores. We
tested for these confounds after participants completed their shop-
ping exercise. These variables (Fairness, Simplicity, Clarity,
Usability, Layout, 1 = Very bad,7 = Very good) loaded onto
a single factor and therefore were combined for analysis (α=
0.94). There were no significant differences across conditions
(F(3, 789) = 0.79, p> 0.50). Additionally, one question
assessing how the online store in the study compared to other
online stores (How did the online store you viewed compare to
other online stores known to you? 1 = Much worse,7 = Much
better) revealed no significant differences in perceived quality
across conditions (F(3, 789) = 2.04, p= 0.11). Another question
regarded the impact on customer satisfaction of our Condition D;
critical observers may consider the Panini-configurator a some-
what sneaky marketing manipulation that could create customer
dissatisfaction. While somewhat obliquely addressing this criti-
cism, we asked participants how satisfied they were with their
configured solution (1 = Very dissatisfied, 7=Very satisfied).
No differences across conditions (F(1, 789) = 1.04, p= 0.35)
suggested no evidence of unease with the Panini method.
Study 2: Banking study
Participants and procedure
Study 2 is a straightforward replication of Study 1 exploring
the incompleteness effect in the context ofbanking services
an industry where cross-selling is a prominent management
topic. This study was conducted through a commercialmarket
research company with 634 employed German adults aged
2550 (46.0% Female). Participants were asked to consider
the following purchase scenario:
You decided to accept a new job offer in Austria and
move there. On recommendation of your new employer,
you want to open a salary account at the BANK
(anonymized, prestigious Austrian banking institution).
The following website shows the respective product cat-
egory Account & Cards.Please choose an account
plus possible complementary products from this catego-
ry. Click on Save & Submitin the Shopping Cart to
finalize the online purchase. You will then be forwarded
to a questionnaire. There we ask you about your
Participants were randomly assigned to one of three online
shopping conditions identical to Conditions A, C, and D from
Study 1, and therefore we label our conditions as such. Seeing
Product Presentation
(Individually Presented
Categories = 0, Basic
Configurator = 1) .17*** (-.63*)
Number of
Fig. 4 Study 1 Mediating effect
of number of categories consid-
ered on completion (Condition A
vs. B). Note. All coefficients are
unstandardized, and the value in
parentheses is the coefficient
when including the mediator in
the regression. *p<.05,**p<
.01, ***p< .001
Product Presentation
(Basic Configurator = 1,
Interrelated set = 2 ) .73*** (-.11)
Number of
Fig. 5 Study 1 Serial mediation effect of completion impulse and cat-
egories considered on completion (Condition B vs. C). Note. Figure 5
depicts the indirect effect of condition on completion via 1) impulse to
complete and 2) number of categories considered. All coefficients are
unstandardized, and the value in parentheses is the coefficient when in-
cluding the mediator in the regression. *p<.05,**p<.01,***p< .001
729J. of the Acad. Mark. Sci. (2022) 50:713–741
marginal benefit of inclusion of the basic product configurator
(Study 1, Condition B), and desiring to simplify
operationalization of Study 2, we focus on the larger whole
and the visual salience of incompleteness relative to baseline
Condition A. The home screen and product configurators
followed the design of Study 1, and Fig. 7shows the banking
product configurator with visual salience (Condition D). The
top image in the figure shows the start screen, and the bottom
left image shows the credit card offerings (one of the two
optional cross-buying products). The bottom right image dis-
plays the shopping cart view after selection or non-selection of
the four product categories. The primary offerings (i.e., ac-
count tariff and debit card) and one cross-buying product
(i.e., overdraft facility) have been selected, but not the second
cross-buying product (i.e., credit card).
As in Study 1, Condition A was a conventional homepage
(see Methodological Details Appendix for all stimuli).
Participants viewed the banks promise of performance and
product categories could be seen in the menu bar on the left.
All participants were required to choose an option from each
of the two primary offerings (account tariff and debit card).
After clicking Selectand Add to Cart,participants were
navigated to the shopping cart view where they saw the text,
The following additional products could be of interest to
you,accompanying any missing products. Participants could
click a Shop morebutton to view and select the optional
cross-buying products (i.e., credit card and overdraft facility),
which served as dependent measures. Participants submitted
their purchases by clicking Save configuration and send.
In Condition C, the start page and all products were em-
bedded together in a product configurator as part of a larger
whole.The program was given a title (Our account and
cards solutionAll about your daily banking)andthe
product offerings were numbered. After the start page, partic-
ipants were guided through each page of the configurator.
Once a participant clicked Select,their chosen product
was listed using text next to the category name in the
configurator. Participants had to choose a product for each
primary offering, but they could choose No selectionfor
the optional cross-buying product categories.
Condition D (shown in Fig. 7) builds upon the idea of a
larger wholewith the intention of making incompleteness
more prominent and easier to visualize, increasing likelihood
that customers would purchase cross-buying options to com-
plete the product program. As in Study 1, the addition of a
Panini column with digital stickersenabled customers to
more easily envision incompleteness by visually representing
unselected products. Participants were again guided through
the configurator, but the selected options were presented with
pictures (vs. text), and the unselected products with visual
placeholders. After submitting their purchases,all participants
completed two questions assessing the effectiveness of our
A oneway ANOVA confirmed an effective product program
manipulation (To what extent did you perceive the offerings as
part of a product program designed to be completely utilized? 1
=Notatall;10 = Very much), (F(2, 629) = 8.83, p< 0.001, η
= 0.03). Specifically, Condition A (M
=6.31,SD =2.41)was
perceived to be significantly less a part of a product program
designed to be completely utilized than Condition C (M
7.11, SD = 2.27; F(2, 629) = 12.80, p< 0.001) or Condition
= 7.14, SD = 2.22; F(2, 629) = 13.54, p< 0.001). An
Product Presentation
(Interrelated set = 2,
Visually linked = 3)
Perception of
Number of
.23*** (.89***)
Fig. 6 Study 1 Serial mediation of feelings of incompleteness, impulse
to complete, and number of products considered on completion
(Condition C vs. D). Note. All coefficients are unstandardized, and the
value in parentheses is the coefficient when including all mediators in the
regression. *p<.05,**p<.01,***p< .001.
The indirect effect of
condition on completion via 1) feelings of incompleteness, 2) impulse to
complete, and 3) number of categories considered is not significant (b=
.0004, 95% CI: -.02, .02).
The indirect effect of condition on completion
via 1) feelings of incompleteness and2) impulse to complete issignificant
(b= .06, 95% CI: .02, .12)
730 J. of the Acad. Mark. Sci. (2022) 50:713–741
additional manipulation check (You have [not] completed our
basic banking program: How noticeable was the [in]complete-
ness of the program based on the type of visual presentation in
the product configurator? 1 = Not noticeable at all;10 = Very
noticeable), compared effectiveness of the incompleteness visu-
alization manipulation between Conditions C and D (i.e., both
conditions with product configurators). A oneway ANOVA re-
vealed that participants in Condition D (M
=7.52,SD =2.34)
reported incompleteness as more noticeable than those in
Condition C (M
=6.44,SD =2.54;F(1, 418) = 20.54, p<
0.001, η
Recall that Hypothesis 1 predicted that a product
configurator would lead to more set completions than a con-
ventional homepage, and salient visualization of an incom-
plete set would further increase completion compared to
text-only presentation of offerings. Supporting this prediction,
only 9.8% of participants in Condition A completed the set
(i.e., chose the two cross-buying products) compared to 46.4%
of participants in Condition C (χ
(1, N= 425) = 70.72, p<
0.001, φ= 0.41). Further, a significantly higher proportion of
participants in Condition D (65.6%) selected the complete set
of offered products as compared to Condition C (46.4%; χ
N= 420) = 15.55, p<0.001,φ=0.19).
These findings were supported by a oneway ANOVA com-
paring the number of products purchased by conditions. A
main effect of condition was found (F(2, 631) = 131.28, p
= 0.29) and pairwise comparisons confirmed a
monotonically increasing pattern such that participants in
Condition D (M
=3.53,SD = 0.71) purchased significantly
more products than those in Condition C (M
=3.26,SD =
0.79; D vs. C: F(2, 631) = 15.25, p< 0.001) who purchased
significantly more products than those in Condition A (M
2.44, SD =0.67;Cvs.A:F(2, 631) = 135.93, p<0.001).
Across conditions, then, we again find support for Hypothesis
1 and replicate the findings of Study 1 in a different domain.
Discussion and further hypotheses
Studies 1 and 2 supported our hypotheses regarding peoples
desire to complete an unfinished set in two different shopping
contexts. These two experiments involved situations of online
cross-selling and simultaneous purchase decisions for prod-
ucts and services. In Study 3, we turn to decision-making in
thecaseofservicescustomersbuyover time, as there are often
situations in which purchases do not occur all at once. For
example, people may buy an iPad, iPhone, iMac, and an
Apple watch, but rarely purchase them simultaneously.
Similarly, in the case of banking products or services, people
may not decide on a savings account, an insurance plan, and a
pension at a single point intime. In such situations, we believe
the incompleteness effect can be used to create a setof
products or services that customers will want to complete
over time. We also consider a different marketing stimulus for
presenting an incomplete taskthat of a visually incomplete
puzzle, the pieces of which represent chosen (and missing)
services. Importantly, in Study 3 we test our moderating hy-
potheses that the incompleteness effect depends on con-
sumersperceived feasibility of set completion and purchase
Feasibility Boundary conditions for the incompleteness effect
suggested in the literature include feasibility. The perceived
feasibility of a task has beenshown to moderate completion of
a task across a variety of situations (Locke & Latham, 1990).
As a task decreases regarding perceived feasibility, partici-
pants should become less likely to remember their prior per-
formance (Zeigarnik, 1927,1938) or resume interrupted ac-
tions (Ovsiankina, 1928). If a task is perceived to be too dif-
ficult or unattainable, peoples motivation to engage in said
task is lower (Locke, 1982; Locke & Latham, 1990). In other
words, setting goals that are too high can have a negative
effect on performance (Locke, 1982). Related research has
Table 4 Study 1 Results for Hypothesis 5
Cross-buying products
Juice Grappa
Condition C Condition D pCondition C Condition D p
Percent purchased 76% 91% < 0.001 77% 87% 0.008
Time spent browsing
in seconds (SD)
20.28 (16.84) 24.15 (19.01) 0.030 15.30 (10.55) 18.89 (17.57) 0.010
Attractiveness (SD) 4.63 (1.63) 4.96 (1.66) 0.057 4.82 (1.57) 5.24 (1.53) 0.011
Primary Offerings
Red Wine White Wine Sparkling Wine
Condition C Condition D pCondition C Condition D pCondition C Condition D p
Percent purchased 97% 98% 0.38 86% 91% 0.11 95% 96% 0.67
Time spent browsing
in seconds (SD)
20.81 (20.50) 20.30 (19.34) 0.80 23.63 (18.66) 24.90 (21.39) 0.53 31.49 (21.69) 31.68 (20.54) 0.93
Attractiveness (SD) 5.42 (1.37) 5.64 (1.32) 0.13 5.29 (1.39) 5.52 (1.38) 0.11 5.17 (1.48) 5.44 (1.45) 0.08
731J. of the Acad. Mark. Sci. (2022) 50:713–741
shown that as confidence in ability to complete a task in-
creases, so too does performance (Bandura, 1986;Bandura
&Cervone,1983; Dachler & Mobley, 1973). However, the
risk that people may abandon a task altogether does exist, in
that a task deemed already partially failed, or an all-or-noth-
ingtype of goal can kill motivation (Cochran & Tesser,
1996; Soman & Cheema, 2004). As related to the current
work, the incompleteness effect should be less likely to occur
if product program completion is perceived as too difficult to
attain. Formally:
H6 Feasibility will moderate the incompleteness effect such
that if product program completion is deemed unfeasible
by the consumer, consumers will be less likely to com-
plete the program.
Fig. 7 Study 2 Condition D: Configurator with products presented as part of an interrelated set and increased visual salience
732 J. of the Acad. Mark. Sci. (2022) 50:713–741
Rationalization Another potential moderator relates to a
persons task justification or rationalization for their
actions. The degree to which one remembers prior per-
formance (Zeigarnik, 1927,1938) and likelihood of
resuming interrupted tasks (Ovsiankina, 1928)depends
on the degree to which people were involved in their
initial task-related actions, as well as the degree to
which they were initially committed to completing a
task. People are more motivated to stay engaged (or
complete a task) when it is personally relevant. A per-
sonally relevant goal is associated with greater impor-
tance, thereby creating greater internal tension which
results in greater motivation to complete the task
(Locke & Latham, 2006).
Justification of need was shown by Zeigarnik (1927,
1938) and Ovsiankina (1928) to be associated with in-
creased tension resulting in greater probability of task
completion. Failure, or inability, to justify or rationalize
purchase of products from a single source will hamper
likelihood of increased purchase of a complete set of
offerings. Relatedly, the concept of commitment has
been found to moderate goal pursuit (Lewin, 1935;
Locke & Latham, 1990). If commitment is absent, mo-
tivation to complete a task will also be absent (Erez &
Zidon, 1984;Nayloretal.,2013; see Locke et al., 1988
for a review). For example, Kivetz et al. (2006)manip-
ulated goal commitment in a coffee stamp card experi-
ment; in their control condition, participants received
stamp cards with no incentive to complete the goal
(i.e., they received payment regardless of card comple-
tion), and no goal gradient effect was observed (i.e.,
there was no decrease in time between visits as cus-
tomers neared completion). These authors argued that
the urge to complete the stamp card was cognitively
overridden and eliminated. Following this logic, we
H7 Consumer rationalization for completion will moderate
the incompleteness effect such that it will be attenuated
as reasons for program completion decrease.
Study 3: Banking field study
Participants and procedure
Study 3 was conducted in the context of an actual cross-selling
optimization project for a German regional bank. An individ-
ualized link invited customers to participate in exchange for a
chance to win an iPad or 100 euros. Customers who opted in
were informed that the banks products had been combined
into new categories-of-need, where each category focused on
a specific financial need. The largest percentage of customers
currently had only the Liquiditycategory-of-need filled
(i.e., used only their checking account and/or credit cards at
the bank); the study was conducted using customers from this
group who had online banking access. Participants were ran-
domly assigned to one of three conditions illustrated in Fig. 8.
Of the 5,000 qualified customers, 459 usable surveys were
completed (N
= 146; 40.2% female;
15 80 years old).
In Condition A, participants saw only their current status of
product use in textual form (e.g., the Group A participant in
Fig. 8only uses the Liquiditycategory). Next, participants
were told they would be informed at regular intervals regard-
ing remaining categories-of-need. Participants were not told
how many remaining categories-of-need