Homework 13-1
pdf
keyboard_arrow_up
School
Georgia Institute Of Technology *
*We aren’t endorsed by this school
Course
6501
Subject
Industrial Engineering
Date
Dec 6, 2023
Type
Pages
2
Uploaded by JusticeFreedomBadger13
ISYE-6501 Homework #13
1
Question 19.1
Before discussing the models we have to consider the relevant data for the problem. The following is a
list of attributes that can help make the decision:
•
Shelf Space data
–
Minimum shelf space required per product
–
Maximum shelf space required per product
–
Shelf space available
•
Total daily sales
•
Daily profit amount
•
Inventory
•
Product History
•
Sales data
•
Store data
The first thing we need to consider is this removal of seasonal items (Christmas items, Halloween items
etc) to ensure that only the products that are available year round are counted.
1.1
Step 1 - Remove Seasonal Data
Given:
{
Sales data, Inventory, Product History
}
Use:
{
Exponential Smoothing
}
To:
{
To remove variance and seasonality from the data
}
Now that we have data without variance and seasonality, we can use that to determine correlated products.
1.2
Step 2 - Find correlated items
Given:
{
Exponentially Smoothed Data, Product History
}
Use:
{
K Means Clustering
}
To:
{
To determine correlated products by identifying clusters of high value paired items.
}
Using the cluster data that we’ve found in the previous step, we can use a community finding model
to find modularity between the correlated complementary products, this ensures we include all correlated
products together, and ensure we have the best communities for each.
1
1.3
Step 3 - Determine product communities and best shelf correlations)
Given:
{
Clustered data from previous step, Shelf space data
}
Use:
{
Louvain Algorithm
}
To
{
Determine product and complementary product and their shelf correlation.
}
Finally, using the updated communities we need to run simulations to determine the best shelves for the
items to maximize sales.
1.4
Step 4 - Optimization model
Given:
{
Up-to-date cluster data, Sales data, Shelf data, Store data
}
Use:
{
Discrete Stochastic Simulation model
}
To
{
Determine optimum model with maximum sales of products and their complementary counterparts.
}
To ensure the model is a good one, we need to roll out the solution to a small subset of pilot stores
with differing constraints; namely the store size, the shelf space, and product selection. We can determine a
threshold for the sales that need to be met in order to consider the model acceptable. We can use the data
from all the pilot stores, and compare to see how the sales went when compared to the previous shelving
data. If a positive change is observed we can start rolling out the model to more stores. Furthermore, if
we see an increase in sales significantly in stores that had less sales overall, we can safely assume that the
effect of putting complementary items on shelves together does exist. Furthermore, we can check the sales
and profits from different stores with different shelf space to see if increases were similar across the board,
this again would validate the hypothesis and confirm that the effects of product placement on shelves is
real. Realistically, I think that this would be the case and that customers would be more likely to purchase
complementary and correlated items if they are near the item they want.
Sales and profits can be safely
expected to grow.
Alternatively, we could also use a Multi Armed Bandit approach rather than an Optimization approach
to generate more models, and implement them in stores faster. The Multi Armed Bandit approach would
gather data faster however, it could lead to lower sales overall while the best model is being determined. Due
to this reason I chose the Optimization model even though it takes longer to determine a feasible solution
based on the number of replications.
2
Your preview ends here
Eager to read complete document? Join bartleby learn and gain access to the full version
- Access to all documents
- Unlimited textbook solutions
- 24/7 expert homework help