Cutting Stockouts in Half for a National Retail Chain
Client: Carraway Home & Goods
Replaced spreadsheet-driven replenishment with a demand forecasting system that cut stockouts by 50% across 140 stores during peak season.
-50%
Stockouts during peak season
+34%
Forecast accuracy
-67%
Manual override rate
Le Défi
Carraway's regional managers were forecasting demand store-by-store in spreadsheets, with no systematic way to factor in local weather, regional events, or the early demand signals already visible in online browsing behavior. Forecasts varied wildly in quality depending on which manager built them and how much time they had that week, and stockouts during peak periods were costing the company sales it had no way to recover — a customer who can't find an item in stock simply buys it from a competitor instead. Replacing the spreadsheet process outright risked alienating managers who had built real intuition over years on the job, so any new system needed to augment that judgment rather than override it.
La Solution
We built a forecasting pipeline that combines point-of-sale history, regional signals, and online browsing data into store-level demand predictions, feeding directly into Carraway's existing replenishment system rather than requiring a new tool. Store managers kept full ability to override any forecast, but we made the default recommendation visibly show its reasoning — which signals drove the number up or down — so managers could quickly sanity-check it against their own knowledge of the store. As trust in the system grew over the first two quarters, the override rate dropped sharply, not because managers were told to stop overriding, but because the forecasts started winning often enough that overriding became the exception rather than the habit.