scroll

A major wallpaper retail company selling its products through e-commerce, needs as an improved and automated demand forecast to prevent stockouts and to reduce company's carrying cost.

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

29%

Increase in Forecast Accuracy (with only 8 months of historical data).

Challenge

Industry Overview

The absence of proper demand forecast leads to the various issues for the industry. Currrent methodology of demand forecast causes either excess inventory being kept at the factory which increases the carrying cost and even causes stockouts at times.

Problem Overview

The company currently owns 5 Seller Flexes in different regions of the country and has 200+ SKUs in different sizes(3m,5m,6m,10m) which are being sold through Amazon.

The company aspires to improve their ability to anticipate demand for three months period. Currently their method of demand forecasting is heuristic in nature. They also want us to deploy inventory norms such as Safety Stock and Re-order point.

Why were we brought in?

The company asked Ganit for help in developing & deploying a solution for demand forecast with improved forecast accuracy and setting up inventory norms.

Our approach

Methodology

The 2 major problems we faced during forecasting process is insufficient historical data and inconsistencies in demand data for few SKUs.

Demand Forecasting at Day-SKU level was done using a combination of actual historical demand (POS data) & other external variables like Festival, ART Sale days etc., Various hypothesis testing has been done to check impact over demand by factors such as discount, promotions, holidays, weekends etc.

To overcome these issues, backward extrapolation was performed using three months average for newly launched SKUs which helped in stabilizing the data. Outlier treatment in demand data is done to make data consistent enough.

Forecasting

Currently AWS Deep AR+ model was chosen as the forecasting engine (SKU-Day level forecast) along with training model has been done using different algorithms as well.

Data Pipeline and Workflow orchestration

Cron job was scheduled to automate forecasting activity.

A valuable difference

Impact

Using Ganit's Solutions, the company have been able to increase forecasting accuracy from 38% to 67%. Safety stock has been calculated which helps in preventing loss of revenue due to stockouts by 13%.

Back .
Top