A sliding window keeps estimation focused on outcomes from a recent fixed or adaptive interval. In changing markets, this prevents stale evidence from dominating decisions after customer tastes, campaign performance, competitors, or seasonality shift. section 5.6
A changing campaign
An advertisement’s click-through rate may look stable yesterday but change today. A sliding window emphasizes recent clicks instead of allowing the full historical record to preserve an outdated estimate. section 5.6
Use sliding windows when:
- Customer preferences, creative performance, competitive campaigns, or seasonal traffic make older outcomes less representative. section 5.6
- A decision system needs to remain agile rather than cling tightly to historical evidence. section 5.6
- Pair the approach with continued exploration, because changing contexts can create blind spots. section 5.7







