ORIGINAL COMPANY-CONTEXT PRACTICE
Forecast demand when scarce items sell out
Original practice scenario: A hypothetical luxury retailer sees low recorded sales for an item that was unavailable for much of the month. How would you forecast demand?
Written for independent practice using company context; this is not a verified question from a company interview.
A REASONED APPROACH
Model answer.
Observed sales are constrained by availability and cannot directly represent unconstrained demand. A forecast should express uncertainty rather than treating every stockout day as zero interest.
Work through it
- Reconstruct sellable exposure by store, channel and day, separating stockouts, closures, launches and intentional assortment limits.
- Compare models using availability-aware features and suitable censored-demand assumptions; retain substitution and wait-list signals as imperfect evidence.
- Backtest chronologically across launches and replenishments, evaluating forecast intervals and downstream inventory decisions by item maturity.
Tradeoffs to discuss
- Pooling similar products improves sparse estimates but can miss brand or collection effects; detailed models may overfit scarce observations.
Common pitfalls
- Do not fill stockout periods with fabricated sales labels.
- Avoid using future replenishment outcomes as features available at forecast time.