Setting attribution in ChatGPT Ads: click window, view window and time basis in three pictures

A conversion count isn't a measurement like the number of clicks. It's the result of a counting rule: which purchases do you credit to an ad — and to which day? In ChatGPT Ads this rule has three dials. This article explains each with a picture, shows which metrics change as a result, and how to set the rule once for your ad account in Chatvertise.

Why a rule is needed at all

Time passes between ad and purchase. Someone clicks today, thinks it over, compares and orders in two weeks. Someone else only sees the ad, doesn't click and buys that evening anyway. Whether those purchases are credited to the ad can't be measured, only defined. That's exactly what the click window and view window do. The time basis then decides on which day the conversion appears in the report.

1. The click window: how long does a click count?

The click window defines how many days after a click a purchase is still credited to the ad. The options are 7, 14 or 30 days; the default is 30.

A short window shows what the ad triggers directly. A long one also captures purchases someone thought about for longer — and therefore reports more conversions. For low-priced products with quick decisions, 7 days is the more honest number; for expensive or consultative products, 30 days.

2. The view window: does seeing without clicking count?

The view window determines whether a purchase also counts when the person only saw the ad but didn't click. There are two settings: 1 day (default) or off.

View-through conversions are the most disputed part of any attribution: whether the ad triggered the purchase or the person would have bought anyway stays open. “Off” is therefore the stricter way of counting, because only clicks count. If you're evaluating reach campaigns, leave the window on; if you calculate cost per sale strictly, turn it off.

3. The time basis: which day owns the conversion?

The time basis doesn't change whether a purchase counts, but when it's booked: on the day of the ad interaction (default) or on the day of the purchase.

“Ad interaction” answers the question: what did my September budget deliver? “Conversion” fits reconciling with the shop, which counts by order date. Within a fixed period, the time basis shifts which purchases fall in or out — the difference is most visible at month end.

What changes as a result — and what never does

• Changes: conversions, cost per conversion (CPA), conversion rate, ROAS and attributed revenue.
• Never changes: impressions, clicks, spend, CTR and CPM. These values don't depend on who gets credit for a purchase.

A real example from one account over two months: with the default, the period reports 121 conversions. With a 7-day click window it's 118, without the view window 119, with both 116. Switch the time basis to “conversion” as well and 110 remain. None of these numbers is wrong; they answer different questions. It only goes wrong when you compare numbers counted under different rules.

How to set the rule in Chatvertise

Because attribution is a counting rule and not a view option, in Chatvertise it's a setting of the ad account. You'll find it on the Conversions page under “Attribution for this ad account”: choose click window, view window and time basis, save, done.

• The setting applies to everyone on the team — two people always see the same conversion count for the same campaign.
• It applies to every view: overview, campaigns, ad groups, ads, custom dashboards, PDF reports and CSV exports.
• If an account deviates from the default, a small label with the active rule appears on the pages, so nobody is surprised by the numbers.
• Automation rules keep using the default. That way existing thresholds don't suddenly fire differently; the Automation page carries a note about it.

Which setting for whom?

• Impulse buys and low-priced products: 7-day click, view off. That shows what the ad causes directly.
• Consultative or expensive products: 30-day click, 1-day view. Decisions take time, and ChatGPT is often where the research starts.
• Reconciling with the shop backend: time basis “conversion”, so the months match the order date.
• Comparing with Meta: 7-day click and 1-day view, because that's the default there. More in Counting conversions across channels.

Two final details

First: OpenAI's conversion reporting — the click/view split and the table by event — doesn't apply the “conversion” time basis to its totals. Those values can therefore differ slightly from the other conversion metrics; the windows, by contrast, work the same everywhere. Second: in the hourly view for “today” the API returns no conversions at all, regardless of the setting.

The most important rule stays simple: decide once, then stick with it. If you switch the counting method mid-quarter, you're no longer comparing campaigns but rules. Why the same campaign shows different numbers depending on the window is explained in Attribution windows in ChatGPT Ads.