Electronic copy available at: https://ssrn.com/abstract=2981858
Effects of an Information Sharing
System on Employee Creativity,
Engagement, and Performance
Shelley Xin Li
Tatiana Sandino
Electronic copy available at: https://ssrn.com/abstract=2981858 Working Paper 17-094
Copyright © 2017 by Shelley Xin Li and Tatiana Sandino
Working papers are in draft form. This working paper is distributed for purposes of comment and discussion only. It may not be reproduced without permission of the copyright holder. Copies of working papers are available from the author.
Effects of an Information Sharing
System on Employee Creativity,
Engagement, and Performance
Shelley Xin Li
University of Southern California
Electronic copy available at: https://ssrn.com/abstract=2981858
1 Effects of an Information Sharing System on
Employee Creativity, Engagement, and Performance
Shelley Xin Li
Leventhal School of Accounting
University of Southern California, Los Angeles, CA 90089
Tatiana Sandino* Harvard Business School
Harvard University, Boston, MA 02163
May 1, 2017
Abstract
Many service organizations empower frontline employees to experiment with different ways to meet diverse customer needs across different locations. We conducted a field experiment in a retail chain to test the effects of an information sharing system recording employees’ creative work— a control system often used to promote local experimentation—on the quality of creative work, job engagement, and financial performance. While, on average, the mere introduction of the information sharing system did not have a statistically significant effect on any of these outcomes, we find that the system had a significantly positive effect on the quality of creative work when it was more frequently accessed. It also had a positive effect on the quality of creative work in stores with fewer same-company nearby stores (i.e., with less natural exposure to peers’ creative work) and on the value of creative work and the attendance of salespeople working for stores in divergent markets where customers had distinctive needs requiring customized service. We also find weak evidence that the system led to better financial results when salespeople had lower rather than higher creative talent prior to introducing the system. The findings of our study shed light on when information sharing systems can affect the quality and performance consequences of employees’ creative work.
* Corresponding author’s contact information:
Morgan Hall 367, Harvard Business School, Boston MA 02163 E-mail: tsandino@hbs.edu
Phone: (617) 495-0625
2 1. Introduction
Service and retail organizations in diverse markets often need to find ways to understand and
satisfy customer needs on a timely basis. To do so, these organizations tend to allocate greater
decision authority to their frontline employees, who have better access to relevant local
information than headquarters, so that they can gather information about customer needs and
generate creative and timely solutions to fulfill those needs (Baiman, Larcker, and Rajan [1995],
Fladmoe-Lindquist and Jacque [1995], Aghion and Tirole [1997], Dessein [2002], Campbell,
Datar, and Sandino [2009]). However, despite the perceived value in employees’ local creativity
and the formally granted permission to experiment, not all employees use discretion to generate
creative ideas or solutions (Campbell [2012]). Prior studies in economics, accounting, and
management have examined several control systems that facilitate or inhibit such local creative
experimentation. Collectively, these studies highlight the importance of providing long-term
incentives, tolerating short-term failures, and selecting the “right” type of employees
(Holthausen, Larcker, and Sloan [1995], Campbell [2012], Manso [2012]). However, these
studies have mostly overlooked the role information sharing plays in ongoing local
experimentation and, more specifically, in employees’ creative efforts.
In this study, we exogenously increased access to information on peers’ creative work through a
field experiment that introduced an information sharing system recording employees’ creative
work (hereafter ISSC). Our goal was to examine the impact of introducing an ISSC on the
quality of creative work, job engagement, and financial performance.1
1 For the purposes of this study, we assess the quality of creative work based on Hennessey and Amabile’s [2009] definition of what
3 There are various economic and behavioral forces that make it unclear whether an ISSC will, in
practice, promote experimentation and yield positive performance outcomes. This kind of system
could yield significant benefits by (a) providing employees with greater access to information,
enhancing their creative abilities through a broader and more diverse pool of ideas than their
initial set and/or (b) affecting the employees’ motivation and engagement by increasing
accountability for the quality of their creative work.
Yet, the system could also impose significant costs if it unintentionally leads employees to
reduce their experimentation by conforming to a common, unimaginative norm. Employees
could do so to minimize risk (if they fear that novel work would be judged negatively by their
peers), if they free-ride on their peers by copying their creative work, or if they felt threatened
that their peers would free-ride on their own creative work (Arrow [1962]).
We partnered with a mobile phone retail company (hereafter MPR) differentiating on customer
service and operating 42 company-owned stores to develop a field experiment testing the effects
of implementing an ISSC. Like many other customer-focused retailers, idea generation based on
the local environment is an integral part of the work at MPR’s stores. MPR operates in an
emerging market where customization of the sales process is essential to compete with local
mom and pop shops. The salespeople at the retail stores create hand-made sales posters to attract
local customers’ attention and to explain the store’s promotion plans, which are updated by the
retailer and/or its suppliers on a weekly basis. Figure 1 presents examples of the posters
salespeople create and display at their stores.2
4 MPR collaborated with us on a field experiment in which the company pilot-tested an ISSC (a
web app) showcasing the sales posters developed by the salespeople at the treated stores (i.e., the
stores where the system was tested). The site was accessible via mobile phone by the store
managers and the salespeople working at the treated stores, who could both upload posters and
browse other people’s posters by brand or by favorites.3 Although the salespeople produced these
posters on an ongoing basis, prior to this experiment, the sales posters had never been shared
across stores. In the field experiment, we randomly assigned stores to a “treatment” group, where
we introduced the ISSC, and to a “control” group.
We examined the effects of the ISSC on three outcomes of interest: quality of creative work, job
engagement, and financial performance. Following prior creativity studies (e.g. Amabile [1988]),
we measured the first outcome based on two dimensions: the value of the creative work
(measured on a 1-5 scale capturing the poster’s ability to communicate the products and deals
offered) and the novelty of the creative work (measured on a 1-5 scale, capturing the posters’
ability to distinctively grab the attention of customers). We used the brand promoter’s weekly
attendance as a proxy for job engagement, and the store-brand weekly sales to measure financial
performance. Our findings suggest that, on average, the ISSC did not have a statistically
significant impact on any of these outcomes.4 However, further tests uncover circumstances
where the system was associated with a number of favorable outcomes. Our analyses show that
the ISSC had a positive effect on the quality of creative work (and to a more limited extent, the
stores’ financial performance and attendance) when it was more frequently accessed, even after
3 In the field experiment, the exposure of the posters on the website was mandatory. Store managers and managers at the headquarters followed up with all the salespeople in the treatment stores during the monthly “uploading” period to make sure that every salesperson’s poster was in the system.
4 Based on our power analyses, we are 80% confident that, on average, the ISSC did not change the quality of creative work
5 controlling for the types of stores more prone to accessing the system. In stores where the system
was used most often, the effect of the system was associated with an increase in the value
(novelty) of creative work of 0.40 (0.23) points on a 1-5 point scale.
We also explore three conditions that could have affected the impact of the ISSC: the users’
natural exposure to others’ ideas, the ex-ante creative talent of the individuals, and the type of
market served (mainstream vs. divergent). On the first condition, we find that the ISSC was
associated with greater quality of creative work in stores with fewer same-company nearby
stores (i.e., an increase of 0.38 points in the value of creative work on a 1- 5 point scale and 0.29
points in the novelty of creative work on a 1 - 5 point scale). This suggests that the system was
more beneficial where salespeople were naturally less exposed to others’ creative work.
Interestingly, the system was also associated with a 0.30 decrease in the novelty of creative work
on a 1-5 scale in stores with more same-company nearby stores, suggesting that salespeople at
those stores may have been more concerned about their peers’ free-riding. On the second
condition, we find that the system led to better financial results when salespeople had lower
creative talent than when they had higher creative talent prior to introducing the system. A 25%
increase in sales following the adoption of the ISSC was observed for store-brands with
promoters whose ex-ante creativity was below the median, suggesting that they may have had
more to learn from others’ work. Finally, on the third condition we find that the ISSC was
associated with a 0.28 point increase in the value of creative work on a 1-5 scale and a 0.24 days
6
markets,5 consistent with the idea that stores requiring greater customization could benefit more
from the system.
In summary, although the introduction of the ISSC at our site did not strongly lead to significant
improvement in average outcomes, the system was associated with improvement in the quality of
creative output when accessed frequently and/or where information was most needed (i.e., where
the salespeople were less exposed to others’ ideas or where they needed to tailor their efforts to
specific customers’ needs). Furthermore, the system was associated with increases in financial
performance for store-brands whose promoters were initially less creative and to greater job
engagement in divergent markets where customers demanded greater customization.
Our study contributes to two streams of literature. First, we contribute to the accounting literature
that examines the effects of management control systems on the quality and performance
consequences of employees’ creative work. This literature has largely focused on examining the
extent to which executive and employee incentive pay can enhance creativity and innovation. For
instance, Holthausen, Larcker, and Sloan [1995] find modest evidence that business units where
executive pay was tied to long-term performance were linked to greater future innovation. Prior
laboratory experiments rewarding subjects for either the quantity or the creativity-weighted
quantity of their creative output show that the former type of rewards leads to greater
creativity-weighted productivity than the latter (Kachelmeier, Reichert, and Williamson [2008],
Kachelmeier and Williamson [2010]). In a similar experiment, Chen, Williamson, and Zhou
[2012] find that, when rewarding individuals for the creativity of their output, group-based
5 The manager identified as “divergent stores” those where the salespeople had to customize the service more due to distinctive
7 tournament incentives are more effective at promoting group creativity than both
individual-based tournament incentives and group-individual-based piece-rate incentives. In contrast with the focus on
incentive systems in these prior studies, our study, to the best of our knowledge, is the first to
focus on a management control system that gives employees access to information—specifically,
an ISSC—that could not only affect the employees’ motivation to be creative and engage more
closely with their jobs but also enhance the knowledge and skills needed to be creative in their
work. Our study finds that these positive effects materialize only when users engage with the
system, and/or when users have greater potential to learn from the information in the system
given either their lack of creativity or exposure to others’ creative work (limited supply of ideas)
and/or their need to customize their creative work (greater demand for ideas).
Second, this study contributes to the literature in accounting and management that studies the
effect of increased access to information on decision outcomes and new knowledge creation.
Prior studies in accounting have found that the implementation of information sharing systems in
organizations is associated with improvements in decision quality and financial performance
(Banker, Chang, and Kao [2002], Campbell, Erkens, and Loumioti [2014]). Moreover,
organizational knowledge creation theory suggests that interactions among diverse practitioners
lead to innovation (Nonaka [1994], Nonaka and Krough [2009]). In two laboratory experiments,
Dennis and Valacich [1993] and Girotra, Terwisch, and Ulrich [2010] apply this theory to test
the effects of information sharing systems on brainstorming sessions. Their findings reveal that,
by enabling subjects to work individually before exposing their ideas to others in brainstorming
sessions, the use of information sharing systems contributes to increasing the number and quality
of ideas generated as well as the participants’ satisfaction with the group ideation process. To our
8
work of individuals as well as their resulting job engagement and performance. And it is also the
first to examine these effects on individuals as they perform day-to-day creative tasks in the
context of their organizations. Our results suggest that an ISSC can affect creativity and
engagement (a) when users engage with the systems on a regular basis, a requirement that may
be harder to attain when individuals are performing their daily jobs than when, as in prior
experiments, they are explicitly asked to use a system to complete a task, and/or (b) when
individuals have greater need to learn about and implement creative ideas.
Beyond contributing to the academic literature, our study aims to shed light on a topic that has
gained significant attention from the business community in recent years, with companies
increasingly seeking ways to leverage the use of information networks within their firms to
increase employee interactions and exchanges of ideas.
The rest of this proposal is organized as follows. Section 2 presents our hypothesis development,
Section 3 describes our research method and setting, Section 4 provides a detailed description of
our research design, presents our analyses, and discusses our findings. Section 5 concludes.
2. Related Literature and Hypothesis Development
Prior research on service organizations has found that many so-called empowered frontline
employees fail to use their discretion to creatively meet the needs of local customers (Campbell
[2012]). An emerging literature in finance and accounting has identified management control
mechanisms that organizations could effectively use to motivate local experimentation among
employees, including long-term incentives, tolerance for early failures, and the recruitment of
9 [2012]). However, to our knowledge, this literature has ignored an important control mechanism,
ISSCs, which has been extensively employed in practice with the intent to promote creativity.6
While prior studies have shown that information sharing systems could benefit organizations
(e.g., Kulp [2002], Kulp, Lee, and Ofek [2004], Devaraj, Krajewski, and Wei [2007], Campbell,
Erkens and Loumioti [2014]), they have focused on benefits related to improvements in
coordination and decision making rather than on improvements in local creativity and
experimentation. Examining the effects that such information sharing systems could have on
creativity and the performance outcomes of creative efforts is relevant, since these effects are
unclear. On the one hand, organizational knowledge creation theory suggests that sharing ideas
enhances creativity (Nonaka [1994], Nonaka and Krough [2009]). In addition to the effect on
creative work itself, prior research suggests that the encouragement of individual creative
activities can lead to greater task engagement and learning (Conti, Amabile, and Pollack [1995]).
On the other hand, economics scholars suggest that the sharing of creative ideas can lead to free
riding and reduce the production of new ideas (Arrow [1962], Dyer and Nobeoka [2000]). To the
extent that free riding diminishes employee investment in thinking about how to do their work
more effectively and/or crowds out some of the best creative ideas across the organization, the
implementation of an ISSC could result in an overall decrease in creativity and value.
Our main objective is to examine whether an ISSC ultimately leads to better financial
performance. However, we first develop hypotheses on the intermediate outcomes that can affect
6According to a survey conducted by McKinsey Global Institute [2012, p.12], approximately 70% of companies use information
10 financial performance (i.e., quality of creative work and employee engagement) before we state
hypotheses on the financial results.
2.1. Effect on the Quality of Creative Work
Traditional monitoring and incentive systems can be useful means of setting expectations and
evaluating employees in tasks and goals that can be explicitly stated. However, customer-focused
service organizations that require employees to apply discretion and creativity to do their job are
often unable to specify in advance all the tasks and processes employees should follow and the
goals they should attain to engage and serve customers effectively (Banker et al. [1996]). Much
of the information on which empowered employees make decisions is idiosyncratic (Campbell,
Erkens, and Loumioti [2014]). To be effective, empowered employees need to rely on tacit
knowledge, which is know-how that is difficult to transmit through explicit means and which can
only be obtained through engaging in practical activities, interacting with mentors, and observing
how experienced colleagues incorporate idiosyncratic information on a regular basis into their
decisions and actions (Polanyi [1966], Tsoukas [2003]).
For two reasons, ISSCs have the potential to overcome some of the limitations that traditional
management control systems have in guiding and motivating empowered employees to use their
local expertise to experiment with new ways to conduct their job. First, ISSCs can expose
employees to a broader and more diverse pool of ideas from their peers, helping them acquire
tacit knowledge or domain-relevant skills that could enhance their jobs. Organizational
knowledge creation theory states that employees who are exposed to other people’s ideas are
encouraged to reflect on their own practices and are more likely to discover new ways to do their
11 Nonaka and Krough [2009, p.645], “by bringing together different biographies, practitioners gain
‘fresh’ ideas, insights, and experiences that allow them to reflect on events and situations.…
Practitioners’ diverse tacit knowledge, that they particularly acquired in their diverse social
practices, is a source of creativity.” Consistent with this theory, Chen, Williamson, and Zhou
[2012] find that tournament rewards for creativity are more likely to result in creative solutions
when awarded to teams rather than to individuals. The organizational knowledge creation
literature also suggests that the process by which ideas are shared in electronic ISSCs enhances
creativity more effectively than the process by which individuals usually share ideas in
face-to-face meetings. This is because electronic ISSCs enable individuals to work alone and share ideas
at different times, preventing the “production blocking” that occurs when individuals attempt to
simultaneously come up with ideas in face-to-face meetings where they need to take turns to
speak up (Dennis and Valacich [1993], Girotra, Terwiesch, and Ulrich [2010]).
Second, by exposing the employees’ creative work to their superiors and/or peers, ISSCs could
motivate employees to increase their efforts to produce high-quality creative work. The most
creative workers might feel motivated, knowing that their work could impact others and be
recognized within the organization, while the least creative workers might exert efforts to avoid
exposing work they might not feel proud of and the appearance of doing worse work than most
of their peers (since the work of all employees is exposed side by side).
Despite the arguments presented above, the implementation of ISSCs could backfire and lead to
a decrease in creativity. Individuals exposed to the creative work of others could converge on
doing work in similar ways and could decrease rather than increase the creativity of their work.
12 emulate common features of their peers’ creative work or from the employees’ decision to
neither pursue nor present their most original ideas for fear of being judged negatively (Diehl
and Stroebe [1987]). Furthermore, the exchange of creative ideas among a group of employees
could lead to free riding (Alchian and Demsetz [1972]). This can occur if individuals copy
others’ ideas in their work or decrease their efforts to produce high-quality work for fear of being
copied by their peers.
The conflicting arguments described above suggest that the introduction of an ISSC could lead to
either an increase or a decrease in the quality of creative work,7 as stated in the following
hypotheses:
Hypothesis 1a: The introduction of an information sharing system recording creative
work (ISSC) will lead to an increase in the quality of the employees’ creative work.
Hypothesis 1b: The introduction of an information sharing system recording creative
work (ISSC) will lead to a decrease in the quality of the employees’ creative work.
2.2. Effect on Employee Engagement
Prior research suggests that two key antecedents of employee engagement (i.e., investment of an
individual’s complete self into his/her work) are the individual’s sense that his/her work is
meaningful and valuable and the sense that he/she is capable of performing this work (Kahn
[1990], Rich, Lepine, and Crawford [2010]). ISSCs have the potential to enhance a worker’s
13 sense of value by increasing the outreach of the worker’s creative work, the expected impact this
work could have on others, and the worker’s potential to be recognized by superiors and peers.
These non-pecuniary benefits could increase employees’ overall level of effort in conducting
their creative work. These systems also have the potential to develop an individual’s abilities and
confidence in those abilities to produce creative work. Furthermore, creative activities tend to
require a more active and extensive search for information. An ISSC that could stimulate greater
participation in creative activities could increase employee efforts in information acquisition and
processing, regardless of the quality of their creative work.
However, ISSCs could unintentionally increase individuals’ fears of being singled out, evaluated,
and embarrassed in front of their peers, leading to an overall loss of confidence (Diehl and
Stroebe [1987], Toubia [2006]). Free riding could also lead individuals to withdraw from
investing in producing high-quality creative work, resulting in a decrease in the value of their
work. These two unintended effects would most likely reduce employee engagement. Given the
unclear effect that ISSC would have on engagement, we split our second hypothesis as follows.
Hypothesis 2a: The introduction of an information sharing system recording creative
work (ISSC) will lead to an increase in employee engagement.
Hypothesis 2b: The introduction of an information sharing system recording creative
work (ISSC) will lead to a decrease in employee engagement.
2.3. Effect on Financial Performance
Customer-focused service organizations that are able to unleash the creativity of their employees
and increase employee engagement in the workplace should expect to achieve greater results.
14 work can be expected to have greater mastery of the brands and deals offered at their stores.
Consequently, they should be better able to communicate these offerings to their customers and
relate them to their customers’ needs even if the quality of their creative work may not improve
significantly (i.e., there could be benefits in the act of creating, regardless of the outcome). Prior
studies on employee engagement demonstrate that employees who are highly engaged in their
work exhibit greater task performance, since they not only exert greater physical effort but also
focus their cognitive and emotional energies on achieving work-related goals (Kahn [1990],
Rich, Lepine, and Crawford [2010]). These arguments suggest that the direction in which the
implementation of an ISSC will affect financial performance will be contingent on the effect of
the system on the quality of creative work and employee engagement. Following the predictions
above, we state our third hypothesis as follows.
Hypothesis 3a: The introduction of an information sharing system recording creative
work (ISSC) will lead to an increase in financial performance.
Hypothesis 3b: The introduction of an information sharing system recording creative
work (ISSC) will lead to a decrease in financial performance.
3. Research Method and Setting
We test our hypotheses by running a field experiment in a mobile phone retail chain (MPR) in
India that operated 42 company-owned stores as of September 1, 2016. On that date, the
company introduced a pilot project in a random group of stores to test the effects of an ISSC.
The system consisted of a web app that displayed sales posters generated by salespeople across
different stores operating across different markets. The managing director of the company
15 experiment, where the stores adopted the system as part of their regular work and the store
teams did not know that they were part of an experiment (Floyd and List, 2016).
This method has several advantages. First, a natural field experiment allows us to test the
impact of an intervention on a randomly selected group of participants who are doing the types
of tasks they naturally perform in their working environment and who possess all the tacit
knowledge needed to perform those tasks. This point is particularly relevant, since some of the
reasons why ISSCs may work (e.g., employees exerting greater efforts to avoid exposing
low-quality work) or may not work (e.g., employees’ free riding on their peers’ work) may be more
relevant in natural settings where individuals have been engaged with their work and their
peers over a long period than in laboratory settings. Second, the natural field experiment
method we employ also allows us to avoid the selection problem that arises when subjects
choose whether to participate in an experiment. In our study, participants were automatically
enrolled in the experiment because the implementation of the ISSC was a pilot program that
they were unable to opt out of. Finally, apart from avoiding this selection bias, a natural field
experiment can mitigate concerns related to the Hawthorne effect (the effect of employees
reacting to being observed rather than reacting to the treatment itself).
MPR operates in a highly dynamic and locally idiosyncratic competitive environment. Its
primary competitors are small independent sellers of handsets and connection services who,
through their intimate knowledge of the local setting, are able to quickly respond to customer
needs and changes in customer preferences. The second most relevant set of competitors is
online retailers who often offer lower prices. To compete with these two sets of competitors,
MPR positions itself as a retailer that offers more value to customers. This is reflected by its
16 for the prices offered; its extensive assortment of products and services to better match customer
needs; and its trustworthy, high-quality customer service. It is essential to MPR’s business
strategy and commercial success that salespeople at the store level can effectively convey the
value of product or service offerings to local customers.
The MPR stores are staffed by a manager, a cashier, and multiple promoters representing various
brands (connection providers, insurance providers, credit providers, and handset manufacturers).
The managers and cashiers are employed directly by MPR and are tasked with efficiently
managing and operating the stores. However, most of the salespeople who interact with
customers and generate sales at each store are so-called brand promoters. Each promoter is hired
by, and receives compensation from, the MPR supplier whose brand s/he represents. A
promoter’s pay typically includes a salary component and/or a sales commission component.
MPR occasionally offers promoters incentives for strong sales performance as well as career
opportunities to become cashiers or store managers.8 None of the explicit incentives promoters
receive are related to the quality of their posters or of any other creative work they produce.
Both the MPR and MPR’s suppliers offer customers attractive packages for their products and
services, including bundles of handsets and connections, accessories, different credit options, and
promotional rewards (hereafter promotion deals). Promoters have significant discretion to
advertise these promotion deals by making their own sales posters to match the needs of their
local customers and displaying them in their stores.
8 The performance-based incentive structure at this setting is typical for the retail industry where frontline employees interact with
17 In a competitive market such as the mobile phone retailing industry in India, promotion deals are
updated weekly. Successful sales posters clearly communicate the essence of the latest deals
through a creative visual design that grabs people’s attention and attracts foot traffic into the
store. Even after customers are drawn into the store, promoters often explain the offering and
make a sales pitch by referring to the posters (see Figure 1 for a sample of the posters designed
by promoters).
The process of coming up with posters requires both a deep understanding of the local customer
base (e.g., what deals and what features in the deals would be most appreciated by the customers
in a specific region and what language is required to connect with them) and a great deal of
creativity (e.g., presenting the important features of the deals in a clear and visually appealing
way). The company realized that motivating promoters to generate more creative posters that
attract local customers is a very important part of achieving commercial success in this market.
However, although most if not all promoters come up with some form of posters as a part of their
sales effort, only a handful of them have consistently generated appealing sales posters that stand
out. Many promoters exert little effort or lack the skills to come up with more creative visual
designs (Figure 1).
To engage more promoters in the process of generating creative posters and to increase the
overall quality and effectiveness of these posters, the company implemented its ISSC.
This setting provides us with a valuable opportunity to test our hypotheses. First, it exemplifies a
typical setting in which ISSCs have the potential to add significant value: retail chains that
differentiate on customer service and which rely on the empowerment and creativity of their
18 interact with local customers and possess deep and real-time knowledge of their preferences and
needs. This knowledge is highly relevant input to producing creative work that aims to generate
sales from local customers. Second, creative output in the natural work environment is generally
difficult to measure. Sales posters developed at the store level enable us to reasonably measure
the quality of creative work, identifying the two main aspects that the literature highlights
regarding the quality of creative work: their value and their novelty (Hennessey and Amabile
[2009]). Furthermore, sales posters are a natural part of the frontline salespeople’s work at this
company and have a direct link to attracting local foot traffic and generating sales. Our findings
are generalizable to service organization settings in which frontline employees’ participation in
creative outputs is valuable and an important driver of financial performance and where such
employees have implicit or explicit incentives linked to financial performance.
4. Research Design, Analyses, and Results
4.1. Description and Timeline of the Experiment
4.1.1. General Description of the Experiment
We planned our study as an eight-month field research experiment from May 1, 2016 to
December 31, 2016. We extended the end of the experiment to January 31, 2017 due to an
exogenous economic shock in India significantly affecting the retail sector and, in particular, our
research site in the month of November.9Our main analyses include data from the month of
January 2017 and exclude data from the month of November 2016, maintaining the same number
9 During our experiment (in the post-intervention period), on November 8, 2016, the Indian government announced the demonetization of Rs.
19 of months that we originally planned to include in the pre- and post-experimental periods and
minimizing the impact of the economic shock on our analyses.
The participants in this experiment were all the brand promoters at MPR. As of September 1,
2016, 36 of the 42 stores had at least one brand promoter per store. On average, these 36 stores
had six brands with promoters per store and six promoters per store (the company generally
assigns a single promoter for a given brand to each store but sometimes—in less than 10% of the
cases— has two or more promoters of the same brand in the same store). We randomly assigned
the 36 stores into two groups: treatment group A (18 stores, 105 brands, and 106 promoters as of
September 1, 2016) and control group B (18 stores, 100 brands, and 104 promoters as of
September 1, 2016).10
Our analyses test the effects of MPR’s ISSC at the treatment stores relative to the control group
on three outcomes at the store–brand level: the quality of creative work, job engagement, and
financial performance. We measure these outcomes using weekly data from archival sources
(i.e., MPR’s accounting and personnel databases) and evaluations of the creative work by
customer panel participants, and validate these measures using survey data from pre- and
post-experimental surveys.
4.1.2 Timeline of the Field Experiment
Figure 2 shows the timeline of the field experiment. We explain this timeline by describing the
events related to (a) the pre-intervention period, (b) the intervention, (c) the post-intervention
period, and (d) the ongoing implementation of the ISSC.
10 In a few cases, store proximity could lead employees in the “control” stores to learn that an information sharing system was implemented in
20
Pre-intervention period. The first four months comprised the pre-intervention period. In June of
2016 (month 2), we tested some of the data collection methods that would be employed during
the experiment.11 The company’s managing director then sent out a memorandum to store
managers (“team leaders” as they are referred to in the company) and promoters in July of 2016
under both experimental conditions. The memorandum described the relevance of sales posters
to the brands’ sales, explained what a high-quality poster is, and let the salespeople know about
company resources they were able to use to prepare their posters. The director also asked store
managers to highlight the relevance of the sales posters to their store staff. The purpose of
sending the memorandum to both the treatment and control groups before our poster data
collection during the pre-intervention period was twofold: (1) to make sure the promoters and
store managers were aware of the importance of sales posters and what good posters should look
like and (2) to hold the information about the importance of these posters constant across both
experimental conditions. Appendix 1 shows the content of this memorandum (month 3
memorandum). In addition to sending an initial memorandum during this period, we conducted a
pre-experimental survey to measure the promoters’ engagement, their store managers’
assessment of the quality of the creative posters, the extent to which the promoters were
motivated to produce creative work, and the extent to which they improved their skills to
produce creative work over the prior month. Whenever possible, we employed survey
21
instruments validated by prior research to design our questions.12 Appendix 2A presents the
questions included in this survey. Finally, the headquarters collected pictures of the posters
displayed across all the stores for the future customer panel evaluations in month 3 (July 2016),
and again in month 4 (August 2016), shortly before implementing the intervention.
Intervention. At the beginning of month 5 (September 2016), the managing director conducted
two workshops to increase the promoters’ awareness of the relevance of creating posters and to
introduce the treatment stores to the ISSC.13 The workshops were completed in two consecutive
days, the first workshop targeted control stores, and the second targeted treatment stores. During
the workshops, the director described what a high quality poster is (emphasizing that posters
should be useful and attractive), and invited the promoters to work on posters for 30 minutes as
they sat in tables organized by store, which had a stack of poster-making materials they could
use (colored cardboards, scissors, pencils, markers, tape, etc.). The director asked each store
team to select their favorite poster and then asked the authors of the posters to explain how they
came up with their posters. Both workshops were identical, except that the workshop for the
treatment stores also included an introduction to, and explanation of, the ISSC.14 While the
promoters and store managers of the treatment stores were working on their posters, two
facilitators helped the promoters and store managers get an account for, and access to, the ISSC
web app. To reinforce the messages communicated during the workshop, the managing director
12 The survey questions that aim to measure employee job engagement are from Rich et al. [2010]. The researchers (us) designed the self-assessment questions. Some of the self-assessment questions on the motivation to generate creative work are adapted from the survey questions on psychological empowerment from Zhang and Bartol [2010].
13 The managing director believed that the ISSC had to be introduced and described in a workshop. Otherwise, it would be ignored.
To avoid confounding the effect of the workshop with the effect of the ISSC, the authors of this paper asked the managing director to run workshops for both the control and treatment groups, differing only in the introduction of the ISSC.
14 One of the authors of the paper monitored the workshops in person to make sure that there were no technical difficulties
22 sent a memorandum to all stores. The memorandum reminded the store teams of the relevance of
producing high quality posters, and, in the case of the treatment stores, also reminded the store
staff about how to use the ISSC web app and the next steps they would need to follow to upload
their posters into the system. Appendix 1 shows the content of the memoranda sent in month 5
(September 2016) to the treatment and control stores.
Post-intervention period. We collected data to evaluate the effects of the intervention during this
period. We gathered photos of the posters displayed across all the stores at two times: shortly
after the implementation of the ISSC (in week 1 of month 6, i.e., the beginning of October 2016)
and toward the end of the sample period (in week 4 of month 8, i.e., the end of December
2016).15 In month 9 (January 2017), we also conducted a post-experimental survey that asked all
of the store managers to assess the promoters’ engagement, their motivation, and their ability to
produce posters. See post-experimental survey questions in Appendix 2A. To capture qualitative
insights and users’ impressions of the system, we also conducted 19 in-person interviews in
January at the stores, asking store managers and promoters of 8 treatment stores to discuss their
experience with the ISSC. See Appendix 2B for the interview questions.
Implementation of the information sharing system. As the memoranda distributed in month 5
explains (see Appendix 1), the ISSC consisted of a web app where the promoters from the
treatment group were able to see the sales posters created by other promoters and the stores
creating these posters. On the back end, the web app was administered by three head office
15This allows us to test the effects of the ISSC on financial performance, engagement, and quality of creative work, within one
23 employees, the managing director, the web developer, and the authors of this paper. The website
enabled the promoters to observe all the posters created in all stores in their mobile or desktop
devices, to pick their favorite posters, and to search the posters by brand, by store, or by favorites
(see screenshots of this website in Figure 3). If despite the mandate, a promoter did not submit a
poster, a blank space with an X and the words “No poster was created” appeared above the name
of the store-brand. The first set of pictures of sales posters were taken and uploaded into the
ISSC right after the system was introduced (in September 2016). From then on, store managers
and promoters at the treatment stores were required to post their pictures in the ISSC once a
month (the first week of month 6, and the fourth week of months 7 and 8).16 The store managers
and promoters of all the treated stores received an email following the uploading and approval of
the pictures every month to let them know the ISSC had been updated.
4.2. Data Collection and Measurement of the Main Variables of Interest
Table 1 Panel A provides descriptive statistics for our main variables of interest. We include
observations for which we have complete data to run our analyses: 4,818 store-brand-week
observations (including only weeks when a promoter of the brand attended the store) and 544
store-brand-month when posters were collected for customer panel evaluations (for our analyses
assessing the quality of creative work).
4.2.1. Main Outcome Variables.
We measure financial performance using store-brand sales data (largely attributed to the brand
promoters working at the store) from the company’s accounting system. “Sales” is defined as the
16 The original plan required the managers and promoters to upload new pictures on the first week of every month after the ISSC
24
natural logarithm of the store-brand weekly sales.17 We exclude cases where the weekly sales for
the store-brand are less or equal to zero.18 The average weekly brand-sales at a store during our
sample period were 10.48 (about Rs.36,000 or US$ 535), but varied considerably from 0.69 to
14.52 (i.e., from less than Rs.5 or just a few US cents, to Rs.2 million or US$ 30,418).
To assess the quality of the creative work (i.e., sales posters), we collected photos of all the
posters two times during the pre-intervention period and two times after the intervention, as
explained in Section 4.1.2. We recorded the dates of the posters, and presented them (without the
date) to a customer panel after collecting all the pictures of the posters and completing the
experiment. We collected a total of 683 unique posters during our experiment. Prior literature
recommends measuring creativity based on the value and the novelty of the work (Amabile
[1988], Sethi, Smith and Park [2001], Hennessey and Amabile [2009]). We obtained two
measures for the quality of creative work (adapted from Sethi, Smith and Park [2001]) from a
customer panel, where we asked the customer panelists to evaluate the value of the posters by
assessing their usefulness in communicating the products and deals offered, and the novelty of
the posters by assessing their visual attractiveness and ability to grab the attention of customers.
The customer panel was convened after the experiment and included customers of two income
groups that the company served (low and high). Each poster was evaluated by 8 customer
panelists: 2 customers per income group x 2 income groups (low and high) x 2 dimensions (value
and novelty). We created a software application that customers at the stores used to consistently
rate the posters in batches of 25 posters at a time. The batches were organized by brand and the
17 One of the 15 brands with promoters at the store provided credit services (Home Credit). Sales in this case are estimated as the
dollar amount of items purchased on credit.
25 selection of posters for each batch was generated randomly, including posters collected
throughout the pre- and post- intervention periods.19 This approach allowed us to hold constant
the evaluating circumstances and to make a fair comparison of the quality of creative posters
drawn at different times. Appendix 3 explains in detail what the customer panel did and presents
screenshots of the software application that the customer panel used to evaluate the posters.
The customer panel measures include a value rating indicating how effective the poster is at
communicating the products or deals offered using a not useful (1) – very useful (5) scale, and a
novelty rating indicating how much the poster’s visual design grabs the customer’s attention
using a not attractive (1) – very attractive (5) scale.20 The intra-class correlation coefficient for
the four ratings received by each poster on the value dimension is 0.3917, and for the 4 ratings
received by each poster on the novelty dimension is 0.3916 (both correlations are significantly
different from zero, with a p-value<0.01). The ratings of the posters of a store-brand collected on
the same month were averaged to obtain an average store-brand rating for each dimension for
each time period, resulting on the main measures for “Value of creative work” and “Novelty of
19 We decided to create batches per brand after noticing, during a pilot test, that some customers were sorting the posters based on brand
preferences. To solve this issue, we decided to include posters of the same brand in each batch.
20 The exact wording used to define these constructs appears in Appendix 3. The customers read the following definitions when
rating the posters at the stores based on either attractiveness or usefulness: Attractiveness:
• Very attractive posters are those that grab your attention the most. For example, you know that a poster is attractive if it prompts you to stop, read it, and walk into the store.
• Not attractive posters are those that grab your attention the least. For example, you know a poster is not attractive if it does not stand out.
Usefulness
• Very useful posters are those that most clearly communicate the products or the deals offered.
• Not useful posters are those that are least able to communicate the products or the deals offered.
26
creative work” used in our analyses.21 As Table 1 Panel A shows, the average value rating for the
posters of a store-brand was 3.14 and the average novelty rating for a store-brand was 2.92.
Our main measure capturing job engagement is “Attendance,” a weekly measure of brand
promoter attendance.22 We adjusted this measure to exclude cases when a new promoter did not
last more than a week at the company or when a promoter reported that s/he attended a store
different from his/her home store for 2 or less days in a month.23 The latter case is because
promoters occasionally visit other stores for a few hours for training purposes. Table 1 Panel A
reports that the promoters of a brand attended the store an average of 5.33 days a week, with
attendance ranging from 1 to 7 days a week. In addition, we constructed measures of employee
engagement through pre- and post-experimental surveys (see Appendix 2A). The survey
measures on job engagement are positively associated with Attendance, our proxy for
engagement (correlation=39.1%, significant at a 1% level).
4.2.2. Explanatory Variables
The main explanatory variables in our analyses are the variables “Info sharing” (identifying
treated stores) and “Post” (identifying the period after which the ISSC was introduced). A
similarly relevant variable is “Active access,” which measures the number of times the staff at
each store accessed the system to actively engage with it (e.g. upload or browse posters) in the
post period. Table 1 Panel A shows that, although the average frequency of access to the system
21 In addition to these measures, we asked the promoters’ store managers to rate the posters in pre- and post-experimental surveys
(see Appendix 2A for the questions included in the pre- and post-experimental surveys). The store managers’ assessment of the quality of the promoters’ creative work was positively associated with the customers’ assessment of the quality of the promoters’ posters, estimated based on the average of the value and novelty ratings of the posters (correlation=10%, significant at a 10% level).
27 was low, there was wide variation in the use of the system, with some stores only accessing it
minimally (4 times) while others accessing it much more often (up to 122 times).
We also collected data from company records and surveys to construct control variables related
to store and promoter characteristics. Store characteristics include the number of promoters for
each brand at the store, store physical size, store age, sales days (i.e., number of days in the week
when the store made sales), number of nearby stores from the same retail chain, and distance to
MPR’s head office. Promoter characteristics include tenure, gender, and pre-intervention
measures for sales/attendance/quality of creative work. Appendix 4 describes how the variables
were constructed. Table 1 Panel A shows significant variation across all these variables.
Table 1 Panel B compares the pre-intervention values of the main variables used in our analyses
between the treatment and control groups. There are no statistically significant differences in
pre-intervention attendance, value of creative work and novelty of creative work between the two
experimental conditions. There was greater pre-intervention sales for the treatment group.
Treatment and control groups also show different pre-intervention values in a number of
covariates despite the random assignment of treatment conditions. In our analyses, we control for
all these covariates as well as the pre-intervention average values of the outcome variables.
4.2.3. Correlation Tables.
Table 2 Panels A and B display two correlation tables corresponding to the two samples used to
assess the effect of MPR’s ISSC on weekly store-brand sales and store-brand promoter
attendance, and the effect of the system on the quality of creative posters. Table 2, Panel A
28 dummy variable for adopting the ISSC (Info sharing), but reports a general increase in sales and
attendance across treatment and control stores after the ISSC was introduced (Post).
Furthermore, the table shows a positive and significant correlation between the salespeople’s
access to the system and sales. Many control variables are significantly associated with Sales,
Attendance, or both measures, and in many instances are also associated with our Info sharing
and Active access variables, suggesting a multivariate analysis would yield more accurate
insights than the correlation table. Similarly, Table 2, Panel B does not show any association
between the proxies for the quality of creative work and Info Sharing, but it shows a significantly
positive association between the quality of creative work measures and salespeople’s access to
the ISSC. Several statistically significant correlations between the control measures and both, the
dependent variables and the ISSC variables (Info sharing and Active access), support the use of
multivariate analyses.24
4.3. Regression Analyses Testing Our Hypotheses
We test our hypotheses by estimating the following difference-in-differences regression model at
the store (i)–brand (j) level, using robust standard errors clustered by store:
Outcomeijt = β0 + β1 Info sharingi + β2 Info sharingi*Postit + β3 Postit
+ βn Controlsijt + Brand fixed effects+ εijt (1)
where Outcomeijt is measured using (a) the natural logarithm of sales for brand j at store i in
week t, (b) the quality of the creative work displayed at store i in month 3, 4, 6, or 8 (based on
24 Matching each poster to the average weekly sales data ranging from two weeks before a given poster collection to two weeks
29 customer panel evaluations of the posters), or (c) promoter attendance for brand j at store i in
week t.25 The variable Info sharing is equal to one if the store is treated (i.e., is part of group A)
and zero otherwise. Hypotheses 1 to 3 (predicting significant effects of the ISSC on the quality
of the creative work, job engagement, and financial performance) are tested by examining the
direction and significance of β2.
The controlsijt include the store and brand promoter characteristics described in Tables 1 and 2,
as well as brand fixed effects (since the brands have different types of products spanning
different price ranges). Appendix 4 defines the variables used in our analyses.
We first evaluate the average treatment effect of the ISSC on sales, the financial outcome (Table
3, column 1). Then, we run similar regressions to evaluate the average treatment effect of the
ISSC on important intermediary outcomes, that is, the quality of the creative work, measured
based on the value and novelty of the posters (Table 3, columns 2 and 3), and job engagement,
measured based on weekly promoter attendance (Table 3, column 4). Our results in Table 3
suggest that the introduction of the ISSC did not have a statistically significant effect on any of
the outcomes of interest. Based on our power analyses, we are 80% confident (1-β=0.8, using a
10% significance level, α=0.1) that, on average, the mere introduction of the ISSC did not
change our creative measures by 0.18 points or more on a 1-5 point scale; the promoters’
attendance by 0.23 or more days per week; or sales by 13% or more.26 We conduct a series of
25 For stores with two or more promoters for a given brand (which make up less than 10% of the store-brands) in a given week, the attendance variables and promoter-level characteristics are an average value across the promoters in the same store-brand. 26 These power analyses are based on regressions where we use brand and store fixed effects as our control variables and rely on
30 tests to ensure that our results are robust to alternative specifications. We rerun our analyses
making the following changes, one at a time:
a. We substitute all of our control variables for brand and store fixed effects (the results of
these tests are presented in Table 4).
b. We winsorize our sales variable at the 1st and 99th percentiles.
c. We re-include store-brand-weeks when the sales are equal to zero, as long as a brand
promoter attended the store in those weeks.
d. We re-include in our attendance measure cases when the promoter attended a different
store than his/her home store for 2 days or less in a month or when the (new) promoter
did not last more than a week with the company.
e. We re-include the month of November into the analyses (which we had excluded from
our main specification due to disruptions caused by the demonetization in India).
All, except the first, of these five specifications yield coefficients for the Info sharing x Post
variable that are statistically insignificant and similar in magnitude to the coefficients reported in
Table 3. The only analysis that produces somewhat different results (the first in the list above) is
presented in Table 4. In this case, the Info sharing x Post coefficient for the “Value of creative
work” variable becomes statistically significant. The coefficient in this case is only slightly
bigger than that in Table 3, and suggests that the introduction of MPR’s ISSC resulted in a
marginally significant increase in the value of the poster ratings of 0.18 points (in a 1-5 scale).27
27 The size of this coefficient is roughly equal to the minimum detectable effect size that we estimated for this analysis before
31
In summary, our results suggest that the ISSC did not significantly affect financial performance,
quality of creative output, or job engagement at the treatment stores. We conjecture that the lack
of results may be due to either (a) the limited extent to which salespeople accessed the ISSC28 or
(b) the inherent nature of the ISSC, which could have simultaneously provided useful
information to produce high quality posters but also triggered free-riding or other unintended
behaviors. To test our first conjecture, in section 4.4.1, we examine whether salespeople that
accessed the system more often were significantly impacted by the introduction of the ISSC. To
test our second conjecture, in section 4.4.2, we examine factors that could have increased or
decreased the benefits of the ISSC.
4.4. Additional Analyses
4.4.1. Effect of the Information Sharing System Contingent on Active Access to the System
We examine whether the lack of impact of the ISSC could be related to the promoters’ limited
use of the system. Columns 2 and 3 of Table 5 show that the introduction of the ISSC was
associated with greater increases in the value of the creative work and the novelty of the creative
work when the store staff accessed (and used) the system more frequently. In terms of economic
significance, the ISSC was associated with an increase in the value (novelty) of creative work of
0.40 (0.23) points on a 1-5 point scale for the store(s) that used the system most often.29 These
increases in creativity are similar or larger than those resulting from providing creativity-based
28 This limited access, combined with the fact that the managing director drew significant attention to the relevance of creating high
quality posters across both treatment and control stores through the in-person workshops and the memoranda, may have diminished the power to distinguish a significant difference between the treatment and control groups.
29 We use data from Tables 1 and 5 to estimate the effect of the system for the stores that used the system most often on the value of
32 incentives to individuals (e.g. Kachelmeier et al. 2008 find increases in creativity ratings of 0.49
points on a 1-10 scale, following the implementation of such incentives).
The results in Table 5 are robust to several alternative specifications where we (a) winsorize
sales at the 1st and the 99th percentiles, (b) re-include observations where sales are equal to zero,
(c) re-include attendance measures when the promoter attended a different store than his/her
home store for 2 days in a month or less or did not last more than one week with the company,
and (d) when we re-include the month of November into the analyses. Yet the results change
slightly when we implement an additional specification: when we substitute all of our control
variables with brand and store fixed effects, the effect of the ISSC on our proxies for the quality
of creative work become insignificant. However, its effect on both sales and attendance becomes
significantly positive for stores where the staff accesses the system more frequently.
In addition to the different model specifications described above, we split the treatment stores
based on the median frequency of access of the system to control for any inherent differences
between treatment stores that used the system more frequently and those that used it less
frequently. We rerun the analyses including separate dummies for high access treatment stores
(i.e., where the frequency of access to the system was at or above the median of all the treatment
stores) and low access treatment stores (where the frequency of access was below the median),
and interacting those dummies with the Post indicator variable. Untabulated results show that,
for high access treatment stores, the introduction of the system was associated with a significant
increase in the value of creative work (Coef.=0.34, t-stat=2.38), and a positive but insignificant
increase in the novelty of creative work (Coef.=0.14, t-stat=1.33). These stores also experienced
33 Overall, our results provide evidence that the ISSC had a positive effect on the quality of creative
output (and to a limited extent, on the stores’ financial performance and job engagement) when it
was most frequently accessed.
We complement our analysis with qualitative insights on the extent to which salespeople were
accessing the system from the 19 interviews that we conducted at 8 of the treatment stores in
January. Although the salespeople found that the app was easy to use, about half of them cited
reasons why they did not access the system more often: forgetting their usernames, their
passwords, the link to the system, etc. Furthermore, our interviews suggested that several
promoters only used the system as a mechanism to upload rather than to actively browse posters
and many of them asked the head office to upload posters on their behalf. Most of them used the
system only when they were required to upload their posters.
Despite their limited use of the ISSC, all of the store managers and promoters interviewed stated
that they had found the new focus on creating posters very useful. They made several comments
praising the use of posters, such as: “There is a difference in a simple and a well-designed
handmade poster. The customer will definitely have a look at well-designed posters.” The
salespeople that more actively used the system suggested that the ISSC had changed the way
they made posters. Elaborating on this, one interviewee stated: “We take ideas from other posters
now.” Another one indicated: “There is always this thought in my mind that these posters are
going to be uploaded in the system and it changes the way I make posters.”
Based on the feedback gathered, and the results suggesting that salespeople accessing the system
34 to roll out the system to all of the stores, to promote more aggressively the use of the system, and
to make investments to create a more engaging user experience.
4.4.2. Conditions Affecting the Effect of the Information Sharing System on Outcomes
Whether a company would be better off letting ideas naturally emerge and spread throughout the
company (the current model of the company studied) or implementing an ISSC might depend on
the conditions of the stores. We explore three conditions that could lead to better or worse
economic outcomes associated with the implementation of an ISSC: the salespeople’s natural
exposure to others’ ideas, the creative talent of the salespeople, and the type of market they
served (mainstream vs. divergent).
a. Natural exposure to others’ ideas. Alternative channels to learn about others’ creative work
will likely diminish the potential benefits of an ISSC. At our site, promoters could potentially
learn about others’ posters by walking into nearby stores and looking at the posters. To
formally examine whether the system’s effect was greater in stores isolated from other MPR
stores, in Table 6 we split our samples into stores with more nearby stores (above the
median) and those with fewer nearby stores, rerun our baseline regressions, and compare the
treatment effects of the two subsamples. Although the level of exposure to other stores did
not affect the extent to which the system affected sales and attendance, our findings suggest
that the ISSC was associated with greater Value and Novelty of creative work in the
35
stores.30 The ISSC was associated with a 0.38 point increase on the 1-5 point scale measuring
the value of creative work (usefulness) and a 0.29 point increase on the 1-5 point scale
measuring novelty of creative work (attractiveness) in stores with fewer same-company
nearby stores. Interestingly, the system was also associated with a 0.30 decrease in the
novelty of creative work on the 1-5 point scale in stores with more same-company nearby
stores, where promoters may have been more aware that others could free-ride on their work.
b. Creative talent of the salespeople. Individual creative talent could affect the effectiveness of
the ISSC. On the one hand, if only a few individuals have creative talent at the stores, it may
be suboptimal to impose an ISSC that would require all individuals to spend time producing
creative work rather than focus on selling. On the other hand, even if many individuals lack
creative talent, their investment in developing posters could lead to greater engagement with,
and understanding of, their work, producing greater benefits. Furthermore, exposing the work
of creative individuals to their peers could further motivate them to develop higher-quality
sales posters. To explore the effect of creative talent on the effectiveness of the ISSC, we
compare the treatment effects between stores with higher-quality ex ante creative output
(above the median) with those with lower-quality ex ante creative output, rerun our baseline
regressions, and compare the treatment effects between the two subsamples. Table 7 suggests
that the effect of the system on the outcomes of interest was generally statistically
insignificant, regardless of the ex-ante creative talent of the salespeople. Yet, the effects of
the system were consistently positive for the subsample of stores with lower ex-ante creative
talent. A comparison of the Info sharing x Post coefficients across subsamples, suggests that
30 The difference in the Info sharing x Post coefficients between the subsamples with more vs. fewer nearby stores is statistically significant
