Attribution Windows: 1-Day Click, 7-Day View, and Why Your Numbers Moved

Nobody changed spend. Nobody changed creative. Reported conversions fell 28% overnight — and the only thing that happened was someone clicked "7-day click" instead of "7-day click + 1-day view" in the ad account settings.

No customer behaviour changed. The attribution window changed: the rulebook that decides how long after an ad interaction a purchase still counts as "caused by the ad." Every platform ships a different default, every window quietly rewrites history when it moves, and most teams discover this the week they try to compare two channels — or two months — measured under different rules.

Click windows vs view-through windows

Two kinds of credit exist, and they differ wildly in strength.

A concrete illustration. A footwear brand spends ₹5,00,000 on Meta in a fortnight and the warehouse records 1,000 orders from customers exposed to those ads. Under a strict 1-day-click rule, 210 orders count — customers who clicked and bought within a day. Widen to 7-day click and 340 count, because considered purchases take days. Add 1-day view and the number jumps to 520: 180 more people who saw an impression, never clicked, and bought within 24 hours — many of them brand searchers and past visitors who would have converted through some other touchpoint anyway.

Same spend, same customers, three conversion counts (210 / 340 / 520) and three reported CACs (₹2,381 / ₹1,471 / ₹962). The attribution window did not measure differently — it defined differently. A view-through conversion is not a weaker measurement of the same thing; it is a claim about a different, far looser causal relationship.

Platform defaults compared

Each platform chose the default that makes its own inventory look best — which is precisely why defaults differ.

Read the incentives. Search gets a 30-day click window because someone who clicked once and bought three weeks later was clearly high-intent — and a long window captures all of it. Social gets a 1-day view bolted on because its inventory is mostly scrolled past, not clicked; without view credit, social's reported conversions would collapse. Neither default is dishonest. Both are self-serving, and comparing a 30-day-click Search ROAS against a 7-day-click-plus-view Meta ROAS is comparing a month of harvesting against a week of prospecting plus a day of bystanders.

For Indian D2C there is an extra wrinkle: conversion lag interacts with COD. A customer clicks on Monday, places a COD order on Wednesday, and it RTOs the following week. Under a 7-day-click window the platform keeps the conversion; your books never see revenue. The longer the window, the more unconfirmed COD intent it sweeps in — another reason to reconcile platform conversions against delivered orders, not placed ones.

What changes when you shorten a window

Shortening a window does three things at once, and only the first is intended.

The intended effect is honesty: the remaining 340 conversions are demonstrably stronger-causal than the 180 that left. The unintended effects are operational. Smart bidding and Advantage+ treat attributed conversions as ground truth, so narrowing the window starves the optimiser — expect a 7–14 day re-learning wobble with higher CPA before it stabilises. Reporting continuity breaks: every historical comparison across the change is invalid unless you restate old periods under the new rule (export daily click-only and view-through splits for at least 60 days before switching). And team psychology takes a hit — a media buyer whose ROAS "fell" 35% overnight will claw the old window back unless leadership pre-agrees that the drop is definitional.

Picking a window for your business

Four practical rules. First, align the click window to your measured lag, not the default: pull the click-to-purchase distribution from your analytics (median and 90th percentile lag in days) and set the window just past the 90th percentile. A brand whose 90th percentile lag is 4 days gains nothing from 28-day click except stray credit. Second, keep view-through out of optimisation targets — visible as a diagnostic column, excluded from tROAS/tCPA signals and from any number that leaves the marketing team. Third, freeze windows during tests: any creative, audience or budget experiment run across a window change is uninterpretable, because the ruler moved mid-race. Fourth, document the windows next to every number the way you would a currency — "Meta 3.1x (7d-click)" is a measurement; "Meta 3.1x" is a rumour.

If cumulative coverage hits 90% by day 5, a 7-day click window captures your reality and anything longer only imports other channels' customers into this one's report.

Frequently asked questions

Should I just turn off view-through everywhere?

For optimisation and targets, yes — optimise bids and judge creatives on click-through only. Keep view-through visible as a diagnostic, because a sudden surge in view claims often means frequency is too high (everyone "saw" the ad) rather than performance improving. Turning it off entirely blinds you to that signal.

Why do platforms default to windows that flatter them?

Because the default is the number every casual advertiser sees, and flattering numbers retain spend. This is not a conspiracy — a 1-day view genuinely captures some real influence — but the burden is on you to choose the ruler, not to inherit it. Change the default deliberately in week one of any new account.

Can I compare ROAS across platforms with different windows?

Not directly. Either restate both to a common click-only basis (7-day click is the usual lingua franca) or stop comparing platforms on platform numbers entirely and compare them on warehouse-attributed new-customer CAC instead. The second option is more work and more honest.

How do attribution windows affect smart bidding?

Directly: the optimiser maximises whatever conversion definition you feed it. Feed it view-throughs and it will buy impressions that generate views, not buyers — frequency climbs, click quality falls, and reported CPA improves while warehouse CAC worsens. Feed it a tight click window and it buys fewer, better events. The algorithm is obedient; make sure the instruction is right.

Do iOS / privacy changes make windows less reliable?

Yes — modelled conversions now fill gaps where device-level tracking fails, and modelled volume concentrates in exactly the loosest buckets (view-through, long-click). Treat any window's modelled share as disclosed uncertainty: ask each platform what proportion of reported conversions is modelled, and discount long-window, view-heavy numbers accordingly.

The summary

• An attribution window defines conversions; it does not merely measure them. Same spend, three windows: 210 / 340 / 520 conversions and three different CACs.
• Click credit is demonstrated intent; view-through credit is mostly correlation — and the home of cross-platform double-counting.
• Platform defaults serve platform incentives: 30-day click for search, 7-day-click-plus-view for social. Never compare across them raw.
• Shortening a window buys honesty but costs signal volume, reporting continuity and a 1–2 week bidding re-learn — restate 60–90 days of history first.
• Set the click window from your measured lag distribution, keep view-through out of targets, and let blended CAC and MER carry the portfolio truth.