Cashback and Optimal GMV prediction for an e-commerce company in India

Enabled a e-commerce company to increase sales by predicting the right cashback/discount to increase GMV%

Cashback-optimal-GMV-prediction-casestudy

About the Client

The client is an e-commerce and financial technology company based in India. The client offers B2C services for consumers and has around 17 fulfillment centers across India partnering with more than 40 courier companies.

The Business Challenge

The client wanted to increase the sales figure for a particular category (footwear) by improving the conversion rate. The client also wanted to increase the GMV (Gross Merchandise Value) by figuring out the right amount of cashback or discount to be given.

What Aptus Data Labs Did

We built an AI-based analytics model to process existing data and predict optimal GMV and cashback to solve the business challenge.

The Impact Aptus Data Labs Made

Aptus Data Labs has to work on two main objectives that were:

  1. Increase category user visibility by getting more users from the home page to the category page to place orders.
  2. Improve conversion by offering the right cashback or discount.

Aptus Data Labs used the following process to prepare data and build the ML model to meet the above two objectives. Aptus Data Labs:

  • Prepared data by generating SALE TYPE attribute to identity the sale period
  • Removed outliers and non-correlation variables to build an efficient model
  • Sampled data to improve the distribution and variance of the data
  • Applied the correlated matrix to consider only correlated attributes for processing
  • Built a deep learning model to predict cashback for a given GMV target using historical data
  • Used the model to predict idea racking to give maximum conversion using historical data of click rates and past conversion lists
  • Utilized the model to predict the banner position of current offers to maximize clicks and reduce bounce offs
  • Implemented another deep learning model to predict optimal GMV based on optimal cashback results
  • Optimized both the models for efficiency and accuracy

Work Flow

Available on Request

Tools used

Rapid Miner

The Outcome

        • The deep learning analytics model was able to predict the optimal GMV and cashback with an accuracy 
      of 90%. The model enabled the client to execute this solution for every footwear brand and subcategory with sufficient data points. The client could also plan and improve the conversion rate increasing sales. The client was able to predict banner position and idea racking to get the maximum conversion.

Related Case Studies

Download Case study​

Explore more of our client stories

Download Case study​