Price Elasticity Project
S Cebellan-Lopez & M Van Gastel
May - July 2021
About
Goal: The goal of this project is to develop a model which be able to calculate a elasticity KPI for each SKU in the assorment.
Exploratory results Below you can find four plots to have an idea of the kind of data we are managing for the project. The purchased are shown per Year-Month, per TopLevelCategory, per ClusterCountry, which are groups of countries based on similarities between them, and per B2B-B2C Customer. All plots divided by Noxion brand.
Price Elasticity KPI: This Elasticity KPI is an indicator that shows you how elastic is a price regarding which product or sort of products we are talking about. The model we ave developed during this project is a loop where four kinds of linear regression are tested for each clusterCountry-TopLevelCategory-Noxion 3-value vector. For each case, is selected the best option based on the highest R-squared. Here you can dive for more information.
Data & Requirements: The orders to the analysis are from the last 12M
Results
The success of this project is to have a KPI Price Elasticity per article, although it has been calculated per the 3-value vector ClusterCountry-TopLevelCategory-Noxion because of a considerable amount of orders are needed to have consistency in the linear regression model.
For all the assortment, the KPI Price Elasticity has de following distribution:
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## -68.5239 -0.5279 -0.3347 -0.3319 -0.1616 95.7596Draws attention that at least the 75% of the values are negative, in the following plot we can see that just the cluster UK-IE has values above 0 and also NL is the one which has more variability within the KPI calculated:
We can see also the distribution of the KPI in terms of the Noxion Brand:
Next Steps
- Apply business rules to the cases the Price elasticity KPI is not significant
- apply a neutral value of the KPI for those cases
- calculate the KPI in a higher level than the 3-value vector TopLevelCategory, Noxion, Cluster Country
- Simulate the prices & the margin with the new prices [research]
- Prove in real: Make proves test of the prices
- Think about other options to calculate the model-loop, por example:
- 2-value vector: TopLevelCategory – ClusterCountry
- 2-value vector: MidLevelCategory – ClusterCountry
- Make de average of the KPI taking into account several results
The steps to new testing may be something like this:
- Define the target/option of the test
- Develop the KPI applying the model+loop
- Calculate prices & make the simulation of prices and margins
- Run the price in real
- Compare the orders & margin with the same month of the previous year
- Decide to test again
Finally here there is the first 1K rows from the result table: