The Problem With Time Based Attribution

Advertising platforms have trained marketers to accept one of the strangest ideas in measurement: if an ad happened before a conversion, inside a platform defined window, the ad gets to claim the sale.

One day click. Seven day click. One day view. Fourteen days. Twenty eight days.

These numbers sound scientific because they are precise. But precision is not truth.

A seven day click window does not prove an ad caused a sale. It only proves the sale happened within seven days of a click. A one day view window does not prove influence. It only proves the ad was shown before the purchase.

That is the foundation of most platform attribution.

Not causality. Timing.

The Window Is Not The Cause

Time based attribution is built on sequence. An ad happened, then a conversion happened.

That sequence may be useful, but it does not answer the real business question: would the sale have happened without the ad?

Most platform dashboards are not built to answer that. They are built to answer something much easier: did the user click or view an ad before converting within the selected attribution window?

That is not incrementality. That is timestamping.

And timestamping is a weak foundation for deciding where to invest ad spend.

The Seven Day Problem

The seven day click window has become so normal that marketers rarely challenge it. But there is nothing magical about seven days.

A click six days before a purchase can be counted. A click slightly outside the window may not be. The conversion did not suddenly become less influenced because the clock passed an artificial line. The platform simply stopped counting it.

The same issue exists with one day view attribution. A person can scroll past an ad, buy later, and the platform claims the result. Maybe the ad mattered. Maybe it did not. The time window itself does not prove either side.

The window creates a rule. The rule creates a report. The report creates ROAS.

But the rule is not proof.

Attribution Windows Shape The Algorithm

Attribution windows are not only reporting settings. They can also shape optimization.

If a platform is told that success means getting purchases within a one day or seven day window, the system will naturally look for users who are likely to convert inside that window.

That does not always mean the ad created the sale. It may mean the platform found people who were already close to buying.

This is the hidden danger of platform attribution. A campaign can look strong because it placed ads near conversions, not because it created new demand.

In ecommerce, users move between social platforms, search, email, direct visits, reviews, and the website itself. The closer someone gets to buying, the easier they become to attribute.

But easier to attribute does not mean more incremental.

Sometimes the best attributed media is simply the media closest to the cash register.

One Sale Cannot Have Five Truths

Platforms often say there are many ways to measure performance. In one sense, that is true. There are different touchpoints, windows, models, and user journeys.

But at the order level, there is still only one sale.

A customer may see a TikTok ad on Monday, click a Google ad on Wednesday, get retargeted by Meta on Thursday, open an email on Friday, and purchase that same day.

TikTok, Google, Meta, and email may all claim influence, but the business did not receive four sales. It received one.

Each platform can say it touched the customer based on its own rules. The advertiser still has to answer the harder question: which dollars created revenue that would not have existed otherwise?

That is the difference between attributed revenue and incremental revenue.

Attributed revenue says, “We touched it.”

Incremental revenue says, “We caused it.”

Those are not the same thing.

View Through Attribution Makes It Worse

Click attribution at least includes an active user action. The user clicked, visited, and later purchased. That still does not prove causality, but it is a stronger signal than a view.

View through attribution is much more fragile.

A user scrolls past an ad, the platform records a view, the user later buys, and the platform claims the conversion.

A view does not always mean attention. Attention does not always mean influence. Influence does not always mean the sale would have disappeared without the ad.

This becomes even more questionable as the window gets longer. One day view attribution is already difficult to interpret. Seven, fourteen, or twenty eight day view windows become even harder to defend without real incrementality testing.

At some point, the platform may not be measuring influence. It may be measuring coincidence at scale.

Modeled Conversions Make The Black Box Bigger

There is another layer advertisers rarely see clearly: not every reported conversion is fully observed.

Because of privacy changes, cookie loss, browser restrictions, app to web gaps, consent rules, and missing event data, platforms often rely on modeling to fill in the blanks.

Modeling is not automatically bad. In many cases, it is necessary. The problem is transparency.

Advertisers are often not shown clearly how many reported conversions were directly verified versus modeled. They are not shown how this changes by platform, campaign, device, browser, market, or attribution window.

That is one reason reported results can shift over time. The number shown in real time is not always the final number. As more data comes in and models update, platforms may revise performance.

That does not mean the platform is lying. But it does mean the advertiser is looking at a black box. And when that black box is also grading its own performance, the incentive problem is hard to ignore.

The Real Measurement Question

The advertising industry spends too much time asking when the conversion happened.

Within one day. Within seven days. Within twenty eight days.

But the real question is not when. The real question is whether.

Would this sale have happened without the ad?

Time based attribution cannot answer that by itself. A timestamp can tell us an ad came before a conversion. It cannot tell us the conversion depended on the ad.

That is why incrementality matters. Holdout tests, geo tests, conversion lift studies, and causal models are not perfect, but at least they are trying to answer the right question.

Time based attribution asks whether the platform touched the user before they bought.

Incrementality asks whether the platform changed the outcome.

Those are completely different standards.

Attribution Was Built For Credit. Advertisers Need Investment Truth.

Time based attribution is not useless.

It can help understand user behavior. It can show how quickly people convert after clicking. It can compare short and long consideration cycles. It can provide directional signals when interpreted carefully.

The mistake is treating it as ROI.

ROI is not whether an ad appeared before a sale. ROI is whether the money spent created more profit than would have existed without that spend.

Platform dashboards usually show attributed purchases, attributed revenue, and attributed ROAS.

They rarely show how many of those sales were truly incremental, they rarely separate observed and modeled results clearly, and they rarely explain why the chosen attribution window should be trusted as the basis for investment decisions.

That is the gap.

Advertisers are trying to answer the most important question in paid media: where should the next dollar go?

Platforms are often answering a different question: which conversions can we claim based on our rules?

The Future Should Be Causality, Not Timing

The future of measurement should not be based on who touched the user last, or who touched the user inside a platform approved window.

It should be based on whether the ad caused a different outcome.

That means proving whether the ad created a purchase that would not have happened, brought in a new customer who would not have bought, increased total business revenue beyond platform reported revenue, and grew profit after media cost.

That is what advertisers actually need to know.

The uncomfortable truth is that full incrementality reporting would make many platforms look smaller than their dashboards suggest.

A lot of reported performance is not true growth. It is credit capture.

And credit capture is a very profitable business.

The Bottom Line

Time matters.

But time is not causality.

A one day, seven day, fourteen day, or twenty eight day window does not prove an ad created a sale. It only proves the ad happened before the sale within a rule someone defined.

If platforms cannot clearly explain why that rule exists, how much of the data is modeled, and whether the conversion would have happened anyway, advertisers should be careful about calling it ROI.

The goal is not to know which platform can claim the sale.

The goal is to know which platform actually created it.