RetailGoal: cut operational costs and time6-week pilot + 3 months
Bestsellers on the shelf during the spring peak
Store managers of a home and garden chain ordered stock on gut feeling and an Excel file. Bestsellers ran out in spring; slow items filled the backroom.
- Client
- Home & garden retail chain, Poland, 120 stores
- Their team
- Supply chain director (owner), 2 category managers, 10 pilot store managers, BI analyst
- Our team
- Engagement lead, data scientist, 2 engineers
- Timeline
- 6-week pilot in 10 stores · rollout in 3 months
−31%
out-of-stock on the top 500 items
45 → 10 min
for a store’s daily order
−12%
write-offs
−8%
stock value across the chain
The business goal
Keep bestsellers on the shelf during the spring peak, when 40% of the year’s revenue happens — while holding less stock overall.
Where it was
Till and warehouse data existed but nobody used it for ordering. Each store manager spent about 45 minutes a day on orders, and quality depended on who was on shift.
Till + warehouse data
collected, but unused
Store manager
Excel and gut feeling
Order
45 min a day per store
What we did
- 1Outcome brief · 2 weeksBaseline from last spring, pilot and control stores chosen together with category managers.
- 2Pilot · 6 weeksSuggested orders in 10 stores; the manager adjusts and approves. The gap to control stores: 2.8 points — continue.
- 3Rollout · 3 monthsAll stores, a simple app for managers, weather and promotion calendar added to forecasts.
- 4Run & improve · ongoingWeekly forecast accuracy check and a big review after each spring.
What did not work
Garden plants: demand depends so much on weather and local events that forecasts stayed weak. Managers keep full control there; the tool only shows them the history.
Who did what
The client’s team
Category managers set rules (minimum displays, promotions); store managers gave weekly feedback.
Our team
Forecasting, data pipeline and the store app; measured pilot vs control weekly.
Together
A weekly pilot-vs-control comparison — the only honest proof.
Before and after
Under the hood
Daily till (POS) and warehouse data pipeline; demand forecasts per store and item with weather and promotion features; replenishment rules on top; web app for store managers; accuracy monitoring and drift alerts.
This spring the complaints were about queues at the till, not about empty shelves.