Time series and forecasting · insight-0214
Separate stockout zeros from true zero demand before modelling a retail series
- Evidence type
- practitioner heuristic
- Difficulty
- medium
- Roles
- Data scientist, Data analyst
- Industries
- retail, e-commerce
Summary: situation, action, reason, boundary
Situation
A store-SKU series looks intermittent, and stock records show the item was frequently unavailable.
Action
Join availability data, mark out-of-stock periods as censored or missing rather than zero, and model demand only from periods when the item could be bought.
Reason
Sales equal demand only when supply did not bind, so stockout periods tell you little about demand.
Boundary
Retail, e-commerce, pharmacy, vending, any channel with stock limits.
Symptom or risk
Zero sales during stockouts are recorded as zero demand, so the model learns intermittency the customer never expressed and keeps stock low, which causes more stockouts.
Exceptions
If availability data are absent, use heuristics such as long zero runs on normally steady items, and label them as inferred.
Trade-offs
Treating stockouts as missing reduces training data; imputing demand adds assumptions.
Pitfalls
Using substitution sales on sister items as evidence of demand without checking the substitution pattern.