Your campaign dashboard can look efficient or wasteful depending on one setting that never changes what customers actually did: the attribution window. Make it too short and paid installs appear organic. Make it too long and an old ad touch can receive credit for a conversion that was going to happen anyway.
The useful question is not “Should we use seven or 30 days?” It is “How long does this campaign's influence remain plausible, measurable, and comparable across the systems making budget decisions?”
What is an attribution window?
An attribution window is the period after an eligible ad click, impression, or engagement during which a later conversion can receive credit for that touch. A seven-day click-through attribution window, for example, allows a click to qualify for an install or purchase that occurs within the next seven days; a conversion outside the window is not credited to that click.
The term is often used interchangeably with *conversion window* or *lookback window*. Google defines a conversion window as the period after an interaction during which a conversion is recorded, and its app conversion documentation applies *lookback window* to the same eligibility rule—while noting that a mismatch between two systems' windows is itself a routine cause of reporting differences.
A window determines which touches may receive credit. The attribution model determines which eligible touch actually receives it. A 30-day window combined with last-click attribution and a 30-day window combined with multi-touch attribution can therefore produce different answers from the same journey.
How an attribution window changes reported performance
Suppose a user clicks an ad on August 1 and first opens the app on August 10:
- A seven-day click through attribution window classifies the first open as outside that click's eligibility period.
- A 14-day window keeps the click eligible, so the install may be attributed to the campaign.
- If another eligible click happened on August 8, a last-click model may credit the newer touch even when both are inside their windows.
The user behavior did not change. Only the rule used to label it changed.
That label flows into cost per install, paid-versus-organic mix, cohort revenue, ROAS, and channel comparisons. A shorter window usually reduces attributed conversions and can raise reported acquisition cost. A longer window usually captures more delayed conversions, but also increases the chance of assigning credit to weak or incidental touches. Google makes the same trade-off explicit in its App campaign guidance: longer windows include more conversions and more potentially organic activity, while shorter windows can miss incremental conversions.
This is also why platform and MMP totals diverge. Google documents app conversion discrepancies caused by differences in window length, touch-date versus conversion-date reporting, cross-network deduplication, and install definitions. Before treating a mismatch as broken tracking, compare the effective rules in both systems.
Click-through vs view-through attribution windows
Clicks and views are not equally strong evidence of intent, so they should rarely share the same window.
| Window | Starts when | What it claims | Practical starting hypothesis |
|---|---|---|---|
| Click-through | A person clicks or taps an ad | The click influenced a later conversion | Longer than view-through because the action shows intent |
| View-through | A person is served or views an impression without clicking | The exposure influenced a later conversion | Short and separately reported because exposure is a weaker signal |
| Engaged-view or engaged-click | A platform-defined video or rich-media engagement occurs | A meaningful engagement influenced conversion | Follow the network definition, then validate against conversion lag |
| Reattribution | An existing user engages after an inactivity period | The campaign caused a return or later event | Keep separate from new-user acquisition |
Vendor defaults are reference points, not universal truth. AppsFlyer's current lookback-window documentation lists a default seven-day click window and 24-hour view window for many non-self-reporting sources, while also showing that network defaults vary. Seven-day click and 24-hour view is a common industry pairing rather than a standard. Google Ads applies different defaults again, tracking first open or install conversions within 30 days of the ad interaction and in-app conversions within 90 days. Two reasonable systems can disagree before a single event is mismeasured.
Views deserve extra skepticism at scale. A person can see several ads, receive an email, search for the brand, and install within the same day. A one-day view through attribution rule may classify the install as influenced; it does not prove the impression caused it. Break out view-through conversions, compare them with click-attributed results, and use lift testing when the budget decision depends on causality.
How long should an attribution window be?
Use observed conversion delay, interaction strength, campaign promise, and platform constraints to choose the shortest window that captures a stable share of plausible conversions. Do not copy a default without checking whether it matches the behavior you are measuring.
Start with the event, not the channel
“Conversion” may mean first open, registration, trial start, purchase, or renewal. Each has a different delay distribution. An install campaign can have a short click-to-first-open delay and a much longer click-to-purchase delay, so one setting should not silently stand in for both.
Write the measurement question in one sentence: “Did this click plausibly influence a first open?” is different from “Did this acquisition source produce revenue within 30 days after install?” The first needs an attribution window. The second usually needs a post-install cohort horizon as well.
Measure the conversion-lag curve
For a mature cohort, calculate elapsed time between the eligible touch and conversion. Then review the cumulative share converted within hour 1, day 1, day 3, day 7, day 14, and day 30.
Choose a candidate point where the curve materially flattens, then segment it by:
- click versus view;
- install versus downstream event;
- campaign objective and creative promise;
- country, operating system, and acquisition source;
- new-user acquisition versus re-engagement.
Google recommends using its “Days to conversions” reporting to guide App campaign window choices. AppsFlyer also publishes a distribution of first launches in its window documentation, but your own lag curve is the relevant evidence because app category, funnel friction, and campaign design change the shape.
Match the promise and decision cycle
A one-hour flash offer should not need a 30-day click window. A considered finance or subscription product may reasonably show delayed installs after high-intent clicks. Keep impression windows tighter than click windows unless an experiment demonstrates durable view-through lift.
Treat these as starting hypotheses, not benchmarks:
| Campaign pattern | Click window to test | View window to test | Why |
|---|---|---|---|
| Flash promotion or immediate utility | 1 hour–1 day | Off–a few hours | The promised action is urgent |
| Mainstream app-install campaign | 1–7 days | Up to 24 hours | Captures common delay without a broad exposure claim |
| High-consideration app or subscription | 7–30 days | Up to 24 hours | Click intent may persist while view evidence stays weak |
| Re-engagement campaign | Based on return lag plus inactivity rule | Usually short or off | Existing-user behavior needs separate classification |
The table is an experiment plan, not a claim that every app should use those values.
Choose an attribution window with a five-step test
1. Freeze a mature touch cohort
Select a fixed set of clicks or impressions and wait until the longest candidate window has elapsed. If you compare a seven-day window on mature data with a 30-day window that is only ten days old, the shorter rule gets an artificial advantage.
2. Replay multiple window lengths
Apply candidate windows to the same raw touch and conversion data. Compare attributed conversion count, CPI, ROAS, organic share, and the share of conversions added by each extension. Look for diminishing returns rather than the largest total.
3. Audit the marginal conversions
Inspect the conversions that appear when moving from seven to 14 or 30 days. Are they concentrated in high-intent campaigns, or do they appear mostly after low-salience impressions? Do they have comparable retention and revenue? A longer window is more defensible when its marginal conversions match a plausible journey and produce credible quality.
4. Align network and MMP settings
Record the configured click, view, engaged-view, install, and reattribution windows for every partner. AppsFlyer recommends matching its self-attributing-network windows to the partner's settings to reduce discrepancies. Also align time zones, reporting basis, event definitions, and whether conversions are grouped by touch date or event date.
Deeplinkly gives app teams raw campaign data through exports, APIs, and webhooks so they can examine click-to-install delay without relying only on a dashboard total. Its deterministic approach leaves installs unattributed when no supported signal exists, which makes the measurement rule explicit instead of stretching a weak match with device fingerprinting.
5. Validate with incrementality
Attribution measures assigned credit; incrementality estimates what happened because the campaign ran. Use geo tests, audience holdouts, or platform lift studies to calibrate windows for high-spend channels. If conversions added by a wider view window do not correspond to lift, keep them as an assisted reporting signal or shorten the claim.
Document the result as a versioned policy: event, touch type, window, attribution model, effective date, data owner, and reason. Revisit it after material changes to onboarding, media mix, pricing, or privacy measurement.
Privacy frameworks impose their own clocks
Not every window is freely configurable. Apple's AdAttributionKit currently gives a person 30 days after a click and 24 hours after a view to install for eligible ad attribution. Apple documents those limits in its attribution and postback timeframes.
Do not confuse that pre-install eligibility period with Apple's post-install conversion windows. AdAttributionKit supports three conversion windows—days 0–2, 3–7, and 8–35 after first launch—with delayed postbacks and data detail that depends on privacy conditions. One clock determines whether an ad interaction is eligible for the install; another controls when post-install values can be updated and reported.
That distinction affects operational reporting. Recent cohorts may be incomplete even after the attribution decision, and a deterministic dashboard cannot manufacture user-level detail that the platform never returns. Mark privacy-preserving, modeled, and deterministic rows separately when analyzing window tests.
Common attribution-window mistakes
- Using one window for every event. First open and subscription renewal answer different questions.
- Giving views the same eligibility as clicks. An impression is weaker intent evidence and creates more overlap.
- Changing settings during an experiment. Branch notes that new window settings apply to future attributions, so record effective dates instead of treating history as one policy.
- Comparing immature cohorts. Longer windows need more time to complete, including privacy-related reporting delay.
- Optimizing for the highest ROAS. Expanding a window can raise attributed revenue without creating one additional purchase.
- Ignoring platform mismatches. Different windows can make an ad network, MMP, and analytics tool disagree even when every event was delivered correctly. Use an attribution discrepancy audit before escalating to engineering.
- Treating assigned credit as causality. Pair attribution with incrementality testing when deciding whether to scale spend.
Frequently asked questions
What is the difference between an attribution window and a lookback window?
They usually describe the same eligibility period from a touch to a conversion. “Lookback” emphasizes searching backward from the conversion, while “attribution window” emphasizes how long the earlier touch remains eligible.
What is a good default attribution window?
Seven-day click and 24-hour view are common mobile starting points, not guaranteed best settings. Start there only when platform rules allow it, then replay your own conversion-lag data and shorten or extend each window separately.
Should click-through and view-through attribution use the same window?
Usually no. A click demonstrates more intent than an impression, so click-through windows are generally longer. Keep view-through reporting separate and validate it with lift evidence where possible.
Does a longer attribution window increase conversions?
It can increase *attributed* conversions because more historical touches remain eligible. It does not create additional customer actions; it changes how existing conversions are classified.
Should an attribution window match the sales cycle?
It should reflect the plausible delay from the measured touch to the selected conversion, which may be shorter than the full sales or retention cycle. Use post-install cohort windows to measure later revenue rather than stretching install attribution to cover every downstream event.
Choose the narrowest window your evidence supports
A defensible attribution window captures meaningful conversion delay without turning every old touch into a claim. Start with event-specific lag data, separate clicks from views, respect platform clocks, align settings across reporting systems, and test the conversions gained at the margin.
If your team cannot replay window choices from raw touch and conversion data, fix that measurement layer before debating seven versus 30 days. Then adopt one documented policy for cross-channel decisions—and use incrementality to test whether the credited media actually changed outcomes.