Price Elasticity of Demand : Framework to determine scalable and robust price elasticity model with large variety of price range

Pricing strategy is an important facet for pricing managers in retail. An optimal price quote of items in store keeps balance between revenue and customer satisfaction level in the business. One of the critical factors in building an effective pricing strategy is to understand the responsiveness of demand to price changes, referred to as Price Elasticity of Demand.

Apart from price, additional factors like promotion, holidays, weather etc also impact the demand. Sometimes, we can observe different price-demand relationship for diverse geolocation features. Another important distinguishing driver for price elasticity is price ranges of items inside a specific item hierarchy. In such cases, it is effective to build different price elasticity model frameworks for different segments of price fluctuations.

The aim of this talk will be to show how an efficient machine learning model is built to determine item-location level price elasticity. Topics include:

  1. Regularized linear models are built on demand considering relevant external factor along with price of items.
  2. Parallel computation is applied to make the solution more time efficient.
  3. Parameters for the model are trained separately for items belonging to different price range clusters.
  4. Finally conclude with how this robust and scalable method has improved forecasting accuracy and can be used as a foundational solution for any market and domain.
 
0 favorite thumb_down thumb_up 2 comments visibility_off  Remove from Watchlist visibility  Add to Watchlist
 

Outline/Structure of the Talk

The structure is as below

  1. Brief introduction of the business problem and the solution to it.
  2. Data requirement and processing method for the model.
  3. Model details to be used to get price elasticity.
  4. How to deal with different price band items.
  5. Scalability of the solution.
  6. Validation technique of the results.

Learning Outcome

Learning outcome is as below

  1. How to include external drivers of demand along with price to make the model more robust.
  2. Distinguish between different price ranges of items and how to build separate model for those.
  3. How to scale the solution for any size of retail market.

Target Audience

Any practitioner of data science - Data Scientists, Data analysts & Data Science-Managers

Prerequisites for Attendees

A intermediate understanding of statistics and machine learning using regularized linear models.

schedule Submitted 1 month ago

Public Feedback

comment Suggest improvements to the Speaker
  • Anoop Kulkarni
    By Anoop Kulkarni  ~  2 weeks ago
    reply Reply

    Thanks for your proposal. The use-case looks interesting for ML, but other than "regularized linear model", didnt get much flavour of data science and ML therein. Did I miss something? Is it possible to elaborate the ML part of the problem statement a little better and then update your proposal accordingly?

    Thanks

    ~anoop

    • Sourit Manna
      By Sourit Manna  ~  1 week ago
      reply Reply

      Hi,

      Here are the parts where ML is used. And also I want to highlight the concept of how we used ML to solve the problem.

      1. Different price band items are clustered using Clustering algorithm where price information(quartiles, sd etc) are considered as input.
      2. For every cluster of items, regularized models are run iteratively to optimize the regularizing parameters for a specific cluster.
      3. Now, considering cluster specific regularizing parameters for different items, different models have been fitted in parallel for all item.  

      Please let me know if I should provide more clarification on the ML part.

      Thanks,

      Sourit