Attribution Window: Types, Platform Defaults and How to Pick One

An attribution window is the period after someone clicks or sees an ad (or visits from another channel) during which a conversion can still be credited to that touchpoint. With a 7-day click window, a purchase made 5 days after the click counts for the ad; one made 10 days later does not. The window you choose changes how many conversions each channel gets, and with them your ROAS, CPA and budget decisions.

This guide covers the window types, the defaults in GA4, Google Ads and Meta, how to pick a length from your own conversion lag (with the arithmetic), why the window you set is not always the window your tracking can see, and what it all means for organic search and AI assistant traffic.

What Is an Attribution Window?

Attribution is the act of giving credit for a conversion to the ads, clicks and visits that led to it. Every attribution system needs a cutoff: how long ago can a touchpoint be and still count? That cutoff is the attribution window.

Three things define a window:

• The trigger. What starts the clock: a click, an ad impression, a video view, an app install.
• The length. How long the clock runs: 1 day, 7 days, 30 days, 90 days.
• The conversion. What has to happen before the clock stops: a purchase, a lead form, a sign-up, an app open.

Platforms wrote windows into their tools because buyers rarely convert at the moment of the click. Someone sees an ad on Monday, compares options, and buys on Thursday. Without a window, the ad either gets credit forever or only for same-session sales. Both are wrong.

Attribution Window vs Lookback Window vs Conversion Window

The top guides do not agree on names. Some describe the lookback window and the attribution window as two different things. In the tools you actually use, they are the same setting with different labels:

Looking back from the conversion or forward from the click gives the same answer: a 30-day window means a touchpoint must be no more than 30 days older than the conversion. When a vendor uses the two terms for different things, ask for the definition in days and in direction. Then you can compare.

Types of Attribution Windows

Click-through window

Credits a conversion that happens within a set time after someone clicks the ad or link. This is the most common window and the strongest signal, because a click shows intent. Typical lengths run from 1 to 30 days, and Google Ads allows up to 90.

View-through window

Credits a conversion when someone saw an ad, did not click, and converted later. Platforms use it for display, social and video, where many people see an ad and come back another way. It is usually short, often 1 day, because an impression is a weak signal and impressions are cheap to pile up. A long view-through window can credit an ad for sales that would have happened anyway.

Engaged-view window

A middle ground for video: the person watched for a minimum time but did not click. Google Ads and Snapchat both offer one. Google’s default sits between its click and view defaults (see the table below).

App install and re-engagement windows

In mobile apps, a mobile measurement partner such as AppsFlyer or Adjust decides which ad network gets credit for an install. The window is the time a network has between its click or impression and the install. Installs outside every window count as organic. Re-engagement windows do the same for users who already have the app and come back through a retargeting ad, and deep-link windows cover how long a link stays valid for the in-app event.

Custom windows

Most platforms let you change the length per account, per conversion action or per ad set. That flexibility is useful and is also the main reason two tools disagree about the same campaign.

Default Attribution Windows by Platform

Defaults matter because most accounts never change them. Here is what each platform uses out of the box, with the source for each:

• GA4. The lookback window lives in Admin, under Data display, Attribution settings. Per Google’s attribution settings documentation, it applies to every attribution model and every key event, and a change applies going forward only. Changing the attribution model, by contrast, rewrites historical reports.
• Google Ads. Each conversion action has its own window. Google’s conversion window help page lists the defaults above, recommends at least 7 days, and confirms that a new window is not applied retroactively.
• Meta. “7-day click, 1-day view” means a purchase counts for the ad if it happens within 7 days of a click on it, or within 1 day of seeing it without a click. Meta has removed longer options over the years, so check what your Ads Manager offers today rather than relying on old guides.
• Other ad platforms. Snapchat, TikTok and Amazon Ads all publish their own windows and settings in their help centers. Read them as the platform’s own view of its ads, not a neutral count.

Attribution Window vs Attribution Model

The two settings are often mixed up. They do different jobs, in order:

1. The window decides who is eligible. Every touchpoint older than the window is dropped from the path.
2. The model splits credit among the rest. Last click gives it all to the final eligible touch. Data-driven spreads it, which is why GA4 shows fractional credit such as 0.4 and 0.6 that sums to 1 conversion.

So a longer window helps multi-touch and first-touch views the most: it lets early touches like a blog visit or a prospecting ad stay in the path long enough to earn a share. Under pure last click, the window only matters when the final touch is older than the cutoff. If you want to see how models move credit between channels, our guide to last click attribution runs the same orders through four models.

How Window Length Changes Your Numbers

Window length never changes how many people bought. It changes how many of those purchases each platform or channel claims.

Short windows make lower-funnel channels look efficient and upper-funnel channels look weak. Long windows do the reverse: more attributed conversions, higher reported ROAS and lower CPA for awareness campaigns, and more overlap between channels. That is why a window change can move budget even though nothing about the campaigns changed.

How Long Should Your Attribution Window Be?

Start from how long your buyers take, not from the platform default. As a rough guide:

• Fast, low-priced purchases (most DTC ecommerce, subscriptions, apps): 1 to 7 days click, 1 day view at most.
• Considered purchases and lead generation (furniture, travel, SaaS trials, local services): 7 to 30 days click.
• Long sales cycles (B2B with several stakeholders, high-ticket services): 30 to 90 days, with the deal itself tracked in a CRM back to the original source.

Shorten the window when you are judging bottom-of-funnel tactics that should close quickly, such as cart-recovery email or branded search. Lengthen it for products with long research phases. Then stop guessing: measure your conversion lag and set the window from data, as below.

Worked Example: Pick a Window From Your Own Conversion Lag

Conversion lag is the time between a touchpoint and the conversion it led to. Google Ads shows it in the time lag report in its attribution reports. You can also compute it from a CRM or order export with the first-touch date and the order date. The numbers below are illustrative, not from a real account: 200 orders from one paid social campaign, average order value $120, spend $8,000.

Each cumulative share is cumulative orders divided by 200. For example, 166 / 200 = 83%. Now compare what the campaign reports under two windows:

Same campaign, same buyers, and the 30-day view reports 26 more orders and $3,120 more revenue. If your break-even ROAS were 2.6, the window alone would decide whether the campaign “works.”

One trap: the lag you measure is limited by the window you already use. If your window is 7 days, you will never see orders that came on day 12. Before you measure, set the window to the longest option for a full cycle, then read the distribution.

The Window You Set vs the Window You Can Actually See

A 90-day lookback only works if your tracking can recognize the same person 90 days later. Often it cannot.

• Safari caps script-set cookies. Since Intelligent Tracking Prevention 2.1, WebKit limits persistent cookies created through JavaScript (document.cookie) to a 7-day expiry. Analytics tags that store their visitor ID that way can lose a Safari visitor who returns after a week. To the tool, the return visit looks like a new person, and the first visit’s channel is cut out of the path, whatever window you set.
• Devices change. Research on a phone and purchase on a laptop are two strangers unless the person logs in or a platform links them with its own signed-in data.
• People clear cookies or use private browsing. Each reset ends the visible path.
• Consent choices. Where visitors decline analytics storage, earlier visits may not be stored at all.

The practical effect: past the first week or so, a long web analytics window mostly extends credit for people who were already recognizable, such as logged-in users and non-Safari browsers. Ad platforms with signed-in audiences see further back for their own ads, which is one more reason their numbers run higher than your analytics. Treat the setting as a maximum, not a promise. For a deeper look at tracking without long-lived identifiers, see cookieless conversion tracking.

Why Platforms Report Different Conversions for the Same Campaign

When Meta, Google Ads and GA4 disagree, the window is often one of several causes. Check them in this order:

1. Different windows. Meta at 7-day click plus 1-day view, Google Ads at 30-day click, GA4 at 90 days: each counts a different set of conversions.
2. Views vs clicks. Ad platforms can count view-through conversions. Web analytics only sees visits, so it cannot.
3. Self-crediting. Each platform credits its own ads. A buyer who clicked a Meta ad and a Google ad can count as a conversion in both, so the sum of platform reports exceeds your real orders.
4. Different models. Last click in one tool, data-driven in another.
5. Different dates and time zones. Some reports place a conversion on the day of the ad click, others on the day of the purchase. GA4 uses the property time zone and Google Ads the account time zone.

Aligning windows across tools narrows the gap but never closes it. Pick one tool as the referee for budget decisions, usually your analytics or your order system, and read platform numbers as each platform’s own claim. For channels that leave no click at all, such as podcasts or word of mouth, add a “how did you hear about us?” question: it has no window.

Failure Mode: Recent Days Look Worse Than They Are

Reports that place each conversion on the date of the ad interaction keep changing after the fact. In Google Ads, the standard Conversions column works this way, and the “by conv. time” columns place conversions on the day they happened instead. With a 30-day window, yesterday’s clicks can still collect conversions for a month.

Using the illustrative lag table above, a campaign reviewed 3 days after a set of clicks shows only 71% of the orders it will eventually get (the cumulative share at 3 days). Pause it on that reading and you cut a campaign that was performing.

• Judge a period only once it is older than the window, or at least older than the point where about 90% of your lag has passed.
• For fresh data, compare like with like: the last 7 days this month against the same 7-day age of last month’s data.
• Write down the date you change any window. In GA4 and Google Ads the change is not retroactive, so data before and after the change is not comparable.
• Use the same window when you compare two campaigns, two channels or two periods.

Attribution Windows for Organic Search and AI Assistant Traffic

Most attribution window advice is written for ads. Organic search and AI assistants behave differently, and the window treats them differently.

• No view-through. Organic results and AI answers produce no impression your analytics can see. Only a visit to your site starts the clock, so these channels compete under click-style rules only, while ad platforms can also claim views.
• They tend to start journeys. Someone reads a guide found on Google or clicks a source in a ChatGPT answer, leaves, and comes back later by brand search, email or a typed URL. If that return falls outside the window, or the visitor is no longer recognized, the organic or AI visit gets nothing.
• Direct swallows missing referrers. When an assistant sends no referrer, the visit arrives as Direct. GA4 gives Direct no credit whenever another channel is in the path, but a visit that was never labeled AI cannot be credited to AI either. A longer window does not fix a missing source. See AI conversion tracking for what can and cannot be identified.
• Landing-page reports are session-based. A report of conversions by session landing page credits the session in which the conversion happened. Credit that crosses sessions only appears in reports that use an attribution model and its lookback window.

From Windowed Conversions to Value by Landing Page

Counting conversions is half the job. Each conversion type needs a value, and then the window’s effect becomes a dollar figure you can act on. An illustrative example: a B2B site values each demo request at $150 (how to set that number is in how to calculate conversion value). These are demo requests credited to organic landing pages under a first-touch view with a 30-day and a 90-day lookback:

The pricing page barely moves (+2 requests) because people who land there convert soon. The how-to guide goes from $900 to $2,250, a 2.5x jump, because its readers are early in research. With only the short window, the guide looks like a weak page. With the long one, it is worth about two-thirds as much as the comparison page. Whatever window you use, report it next to the number so nobody compares a 30-day figure with a 90-day one. For the full setup, see SEO conversion tracking. SEOConversion reports conversions and their value by landing page for organic search and AI assistants, and a visit with no referrer stays Direct instead of being guessed.

FAQ