CUSTOMER STORY | RETAIL
Bavaria knows which client will buy which product, every day, across 300,000 stores
Across 300,000 stores, 30 million transactions and 63 products, classic time series models were not enough. TIMi built 30,000 neighborhood models that give a daily purchase probability and help pre-fill orders.
Model accuracy, up to
92%
Scores per second
2.4M
Neighborhood models
30,000
OVERVIEW
The challenge.
Avoid stock-outs and make life easier for store managers.
300,000 stores
30 million transactions covering 63 products.
More than time series
Sales depend on weather, local events and social dynamics: conventional time series models are not enough.
SOLUTION
What was done.
Richer variables, and one model per neighborhood.
More variables
The models include weather, sales of nearby stores and other relevant information, with 160 million trends.
30,000 neighborhood models
Each neighborhood has its own model.
Deliverables
A daily probability, a recommendation, and upsell and cross-sell opportunities.
RESULTS
What changed.
A daily view of purchases, scored in record time.
Daily purchases
Each day, we know which client will purchase which product.
Demand and orders
The result is linked to a demand prediction model that estimates the quantity needed and pre-fills the order.
Speed and accuracy
About 2,400,000 scores per second, integrated automatically in Qlik, Tableau or other DB and BI systems. Accuracy up to 92%.
SEE FOR YOURSELF
Test it on your own data.
Run TIMi on your own data with a free trial, or talk to us about your case.
