Why it matters
Performance teams make budget and creative decisions inside platform dashboards. Those dashboards apply fixed attribution windows: Meta commonly uses 1- or 7-day click plus view-through options, Google Ads defaults to a 30-day click-through window (extendable to 90 days), and mobile MMPs often use 7-day install attribution. A conversion that happens on day 31 may not receive click credit in platform reporting even when the ad influenced the customer.
That is a reporting and credit problem, separate from optimization impact. Churney internal observation: signal potency for campaign learning drops sharply ~36–48 hours after the anchor conversion, even when later lifecycle events still fall inside a longer attribution lookback. A renewal or repeat purchase may show in reporting but arrive too late to materially re-steer live acquisition. See canonical-attribution-vs-optimization-window.md.
Finance and cohort analytics measure value over months. When attribution windows, optimization potency, and business maturity use different clocks, you can see strong platform ROAS alongside weak blended economics, or vice versa.
Attribution windows also shape experiment readouts. A holdout evaluated only inside a 7-day credit window may miss delayed converters, subscription renewals, or returns that arrive later. Teams that align window definitions across platform reporting, data warehouse exports, and pilot success criteria avoid arguing over numbers that use different rules.
Attribution window
pLTV activation bridges short platform reporting windows and long business value:
- Data warehouse (input): Store attribution data (GCLID, fbc/fbp, campaign metadata) alongside orders and lifecycle events at user grain, independent of any single network credit rule.
- Model (Churney): Train user-level pLTV on mature cohorts whose outcomes exceed default reporting cutoffs.
- Signal design: Time anchor events and value magnitudes so predictions arrive while optimization potency is still high, not only while credit would still apply in reporting.
- Activation (output): Churney sends value events directly to ad networks via Meta CAPI or Google Ads Conversion API, matched to click identifiers captured at conversion time.
- Readout: Compare incremental ROAS, cohort LTV, and BAU conversion once cohort maturity allows a fair comparison beyond the platform reporting window.
The data warehouse holds durable history; it does not replace server-side delivery. Without warehouse-backed joins between touch IDs and delayed revenue, you cannot prove whether value signals improved outcomes the platform never credits inside its default window.
Category variants
| Vertical | Window tension | Practical note |
|---|---|---|
| Ecommerce / DTC | 7-day click vs repeat purchase on day 14–30 | Platform ROAS can overweight first-order promos |
| Subscription | Trial start attributed; renewal value arrives later | Separate trial vs paid value in signal design |
| Mobile app | 7-day install window vs D30/D90 IAP | MMP window may not match in-app subscription billing |
| SaaS / PLG | Long sales cycles exceed default click windows | Data warehouse attribution fields critical for pilot readout |
Common mistakes
- Treating platform ROAS as finance truth. Windowed attribution is a reporting credit rule, not net revenue.
- Mixing windows across tools. Meta 7-day click compared to Google 30-day click without normalization.
- Conflating attribution window with optimization window. Longer attribution settings improve credit visibility in reporting; they do not fully extend learning potency (Churney observes ~36–48h post-anchor decline for delivery impact).
- Ignoring view-through credit. Inflates upper-funnel campaigns when not reconciled to incrementality tests.
- Evaluating pLTV pilots inside platform reporting windows only. Delayed value may never show in short credit windows even when cohort economics improved.
- No data warehouse copy of touch IDs. Cannot join delayed outcomes to the ad that started the relationship.
Advertiser lens
| Role | Cares about |
|---|---|
| UA / performance | Which window the platform uses for reporting credit vs optimization learning |
| Growth analytics | Consistent window definitions for experiment readouts |
| Finance | Revenue recognition timing vs attributed conversion dates |
| Data engineering | Storing click IDs and campaign keys with enough history for joins |
FAQ
What is an attribution window?
The maximum time after an ad interaction during which a conversion can be credited to that touchpoint in platform or analytics reporting.
How is an attribution window different from an optimization window?
Attribution window defines credit in reporting: which touchpoint gets assigned the conversion in Ads Manager. Optimization window defines learning and delivery impact: whether a new event still changes who the algorithm targets. They overlap in platform UI settings but answer different questions. Churney observes optimization potency fading ~36–48 hours post anchor conversion even when longer attribution windows still grant reporting credit.
What are common default windows?
Meta: often 1- or 7-day click plus optional view-through. Google Ads: 30-day click-through default. Mobile: frequently 7-day install attribution via MMP settings.
Why do attribution windows matter for pLTV?
True customer value often matures after the default reporting window closes, or after optimization potency has already faded. pLTV sends early predicted value while delivery can still learn, while data warehouse history validates longer outcomes.
Can I extend attribution windows?
Many platforms allow longer click windows in settings. Extension improves reporting visibility but does not replace sending value signals early enough for optimization impact. Some networks also extend optimization horizons separately; treat each change on its own terms.
How should teams align windows for pilots?
Document platform window, data warehouse join logic, and maturity horizon for readout. Compare incremental ROAS and cohort economics, not platform totals alone.
Not the same as
| Term | Difference |
|---|---|
| Optimization window | Period when new events still influence learning and delivery; not the same as reporting credit lookback |
| Multi-touch attribution (MTA) | Model assigning fractional credit across many touches |
| Lookback window (modeling) | Feature engineering horizon in pLTV training, not ad platform UI setting |
| Cohort maturity | Time until outcomes stabilize for evaluation, often longer than any attribution window |
