scroll

Major Retailer deployed Ganit's data science solutions to forecast demand with greater accuracy that helped them in improving their demand planning for reducing the cost of over stocking to improve the sales by preventing the scenarios of under stocking.

We made a visible and measurable impact to our client's business

30%

Increase in forecast accuracy.

20%

Reduction in stockouts.

10%

Increase in revenue.

Challenge

Industry Overview

High variability in the sales pattern and cannibalization effect of constantly adding new products in the portfolio

Promotions given to distributors leads to high sales due to overstocking and less sales in the coming months.

Problem Overview

"With over-forecasting, the company must bear high inventory cost. If under-forecasted, products may be out of stock, which affects customer experience and reduce the sales.

The client has more than 500 SKUs across its portfolio with constantly adding new ones and they sell their products on 18 online platforms with major ones being (Amazon, Grofers, Nykaa, Flipkart, and Big Basket) for which we have created the forecasting solution".

Why were we brought in?

The retailer asked Ganit for help in developing & deploying a solution to help in production planning by forecasting the demand of products for the next 3 months.

Our approach

Methodology

The problem statement was broken down into three simpler problems listed below.

1. What is the expected demand ?

2. How to scale this for all the ecommerce platforms?

Demand Forecasting was done using a cobination of actual historical demand (primary sales data) & other external variables like festival, seasonality of the products, promotions given to distributors, secondary sales quantity, price, promotions on online platforms like Amazon big billion day etc.,(~100 Hypothesis tested).

Forecasting

AWS AutoMl model (with RTS) was chosen as the forecasting engine (Online platform-SKU-Month level forecast).

Data Pipeline and Workflow orchestration

"All the codes were built on sagemaker, so that end-to-end process is carried over AWS.
1. Data pulling from amazon athena and S3.
2. Datasets creation.
3. Running Amazon forecast.
4. Fetching forecast from Amazon S3 and dissaggregating the forecast at required level"

How did we enable consumption?

The solution was piloted in for the online chain of Amazon. Later on Scaled up for the remaining 4 channels ( Big basket, Grofers, Nykaa, and Flipkart).

A valuable difference

Impact

Using Ganit's Solutions, retailer have been able to increase our forecasting accuracy from 20% to 50%.

Back .
Top