What an attribution window is
When a user clicks a recommended product but buys it several days later, something has to tie that
purchase back to the original interaction. That is exactly what the attribution window does.
Recommendation click → [Attribution window: 7 days] → Purchase on day 5 → ✓ Counted
Recommendation click → [Attribution window: 3 days] → Purchase on day 5 → ✗ Not counted
How to pick the right window
| Product category | Recommended window |
|---|---|
| FMCG, groceries | 1–3 days |
| Fashion, apparel | 7 days |
| Electronics | 14–30 days |
| Furniture, home improvement | 14–30 days |
Anchor the decision on the median time from first product view to purchase in your own category —
that figure is measurable in your analytics platform.
Impact on A/B test results
The attribution window is one of the most underrated settings in an A/B test. The usual mistakes:
- Window too short → understates the CR of recommendations in long decision cycles
- Window too long → contaminates the data with organic conversions unrelated to the test
- Different windows across tests → results that cannot be compared between experiments
View-through vs click-through
- Click-through: counts a conversion only when the element was clicked — the stricter standard
- View-through: counts a conversion on an impression without a click — common in advertising, but
it inflates the measured effect of onsite elements
For recommendation widgets, use click-through: it is the more objective and more reproducible standard.