One forecast for merchandising, the warehouse, and finance

Most growing brands already have a forecast. It lives in a spreadsheet with a name like FINAL_v7. It does not match what the warehouse sees. Finance has a third version for cash. Every Monday someone rebuilds it from exports that do not quite agree.

The fight is rarely about the math. It is about whose number counts. Merchandising wants to buy. The warehouse wants to know what will actually arrive and leave. Finance wants a number it can put in a plan. If those three stay in separate files, stockouts and leftover stock show up together.

Make the object small

A useful forecast is by product and week, or by product, channel, and week. That is small enough to argue about. A yearly total in one cell is not. Promotions, seasonality, and past stockouts should be flags in the model, not stories in Slack.

Merchandising must be able to change a number and leave a reason. If they cannot, they will keep a private sheet. The warehouse must see inbound risk next to the plan, late purchase orders, slipped dates. If they cannot, they will ignore the plan and work from the dock.

What the model is for

A model helps when history is messy and channels compete. It fills a first-pass number so Monday does not start from a blank sheet. It does not replace the weekly meeting. There still needs to be one pack everyone uses.

Connect storefront, ads, and warehouse or 3PL data so the SKU list matches. If product IDs differ, fix that first. A pretty chart on broken keys is still a broken chart.

What to watch

Stockouts on the products that sell most. Extra weeks of stock on slow items. Forecast error where it costs money, not where it is cheap. How often someone overrides the model. If nobody ever overrides it, the model is either perfect or unused. In ecommerce it is rarely perfect.

When the meeting uses one pack, finance stops reconciling three versions of what the team thought it would sell. That is the point. Not a smarter algorithm for its own sake.