Reorder sizes that move with demand, not against it.

Weighted moving average: each of the last 12 weeks gets a weight, recent weeks weighted heavier. Add a simple linear trend so a steadily-rising SKU gets bigger orders and a fading SKU gets smaller ones. Project the result over the supplier's lead-time-in-days plus a 50% safety margin.

No black-box neural nets, no probability bands. Honest math you can verify by hand. Cold-start (no shipment history) falls back to the flat reorder-point × 1.5 default, so new shops don't get garbage forecasts.

The reorder-alert cron already drafts a PO every time on-hand drops below the reorder point. With the forecast in place, the drafted quantity is whatever you'll actually need across the supplier's lead time, not a flat number that ignores the last three months of customer demand.

You still review and Issue. I just stop asking you to do that math in your head.