A stockout is a reorder decision that came too late, not a mystery.
Almost every stockout traces back to one of two things: a reorder placed after the lead-time window had already closed, or a demand acceleration that outran the velocity average the last forecast was built on. Neither is unpredictable if the runway is being watched.
What actually causes stockouts
- Late reorder timing — the most common cause: a SKU crosses its reorder point without anyone noticing, and the replacement order goes out too close to (or after) the point where lead time makes it arrive on time.
- Demand acceleration — a promotion, a viral moment, or a seasonal ramp increases sales velocity faster than a trailing-average forecast adjusts to, so the reorder quantity calculated a week ago is already too small.
- Supplier delay — a lead time that quietly grows (a slower shipping lane, a backordered material) without the forecast being told, so the buffer that used to be enough no longer is.
Why safety stock isn't a complete answer
Safety stock is a fixed or statistically-derived buffer meant to absorb normal demand variance and minor lead-time slippage. It works well against the variance it was sized for. It does not protect against a demand spike or supplier delay large enough to exceed that buffer — no fixed number can, by definition, cover every possible deviation. The realistic goal is reducing stockout frequency, not eliminating the risk entirely.
How Foreshelf approaches this
Foreshelf tracks current stock against sales velocity and supplier lead time (see inventory forecasting) and flags SKUs approaching their reorder point with enough runway to act, rather than waiting for stock to hit zero. Where a purchase order is already in transit, that's factored in so a SKU isn't flagged as at-risk when replacement stock is already on the way — see purchase order planning.
What Foreshelf doesn't claim: prediction of sudden, unprecedented demand spikes with no basis in recent velocity. A viral moment with zero prior signal is genuinely hard to forecast for any tool working from historical sales data — that's a real limit, not a gap specific to this app.