The scholarly article I have referred to is called Information Exchange in Group Decision

Making: The Hidden Profile Problem Reconsidered - ISSN 0025-1909 _ EISSN 1526-5501 _
09 _ 5504 _ 0568.

Research Question: The research deals with hidden profile issue. This means that when group
members have common information favoring an inferior choice and some members have private
information favoring a superior choice, groups typically choose an inferior alternative. The
research aims to look at previously unconsidered causes for the hidden profile issue, solutions to
overcome this issue and problems with these solutions.

Method: Two reasons for the hidden profile (HP) issue that have previously not been considered
as per the research are 1) mistakes in recalling information and 2) structure of the problem or
bias in recalling information. The data collection is done using experiments. The experiments
have been set up in such a way as to test for the above two reasons/theories put forward. The
experiment results are checked for consistencies with previous literature. The reference point
model was used to identify factors responsible for the failure to detect more HPs. Statistical
models such as probit and logit regression models were used to identify statistically significant
parameters. Tables, line charts and bar graphs have been used to visualize information.
Counterfactuals were used to check if the rate of discovering hidden profiles increased. When
looking at recalling mistakes they also look at a factor called group correction factor. The
solutions suggested for the HP issue are Devils advocate (DA) and Process Accountability (PA).
The advantages and disadvantages of these solutions are discussed. The research concludes as
identifying previously unidentified causes of HP but accepts that there are limitations in the
solutions.
The problem at work is one of the Thinking Conundrums which is called the Syndrome of
Superficiality How to get in deep when there is so much pressure to get it done. We are
looking at optimizing promotions for various products. Executing the optimized promotions
requires co-ordination between various teams the business units, the operations team, the
signing team and the modeling team which changes based on the type of product. Since this took
majority of our time we couldnt spend enough time on reflaction (reflection + action). To help
solve this, we have now hired two independent contractors. This brings the dilemma of
delegation in the picture. I need to decide how much of the work chaos should be made visible
to these contractors. I also need to layout some kind of order for them to ensure stability and
business continuity.
Plan your Project

Research Question: Establishing a methodology to optimize promotions of products

I believe that using the combined approach discussed on slide 12 of the Exploratory Projects
class will be most suitable to my research question

Problem Definition and Questions: Establish a method to optimize promotions for products at
store cluster level.

What prices work better for each type of product e.g. seasonal products, basic products?

How to establish co-ordination between the various teams so that everyone can give their inputs?

How to allocate time for reflection (reflection + action) where there is so much to get done?

Sampling: Currently the focus is on simple promotions where products are promoted at a price
point or a percent off. The products selected for optimization are those where we expect to
receive the highest impact (in terms of optimum merchandise sold for seasonal products and
margin for basics products) from our optimization efforts. So, our sample is a purposefully
biased sample.

Data Exploration and Cleaning: The data is pulled by Information Technology (IT) team every
week which is a dump of our promotion management system. This data gives us the detailed
view of the promotions planned by the business units for each week. This helps us capture the
tacit knowledge possessed by business units. Extraneous data and duplicates resulting from
process issues are removed. The products we are interested in are filtered out. The sales,
inventory data and past performance data are available in our internal systems.

Data Transformation: This is basically looking at the data at various levels category level,
merchandise group level or item level. This involves rolling up the data to our desired level as IT
provides the data at the most granular-item level. Sales and inventory data are converted to
operational ratios e.g. Percentage sold from inventory.

Model Development: There are various modeling teams which model for each type of product
and depending on the store cluster to which the prices are sent to. The data collected is used as
input for these models. The statistics and data analysis are an important part of the modeling.
Additionally, the capturing of tacit/qualitative data is done through co-ordination with the
various business units who know the competitive landscape and front-end realities in the store.

Model Evaluation: The model is evaluated by using a test design approach to implementation.
The evaluation depends on the type of product a Pre-Post Test Control approach is used in
most cases. Life of Product revenue/merchandise sold is used for seasonal products. Data
visualizations are created to understand pre-post trends in sales.

Implementation/Intervention: Lack of resources is a problem in this area which made data

Feedback & Updating: The feedback is the test results pulled using the methods mentioned in
Model Evaluation. Prices that work can be rolled out to more store clusters or nationally. Prices
that dont work give us learning to help us price differently by taking a different approach during
modeling. This helps us get empirically tested price recommendations.