Rules-Based Attribution: What Each Model Credits and Why Your Agency's Report Depends on It

Rules-based attribution gives credit for a sale to ad clicks and other touchpoints using fixed rules, such as all credit to the last click or an equal share to every click. Data-driven attribution sets the credit from past data instead. The same sales produce different numbers under each, so ask which model a report uses.

By Avi Vatsa, CEO, Exchange Four Agency. Avi leads the team that installs and runs AI-leveraged revenue systems for owner-led companies. His background spans law, technology, and marketing; he also co-founded Dialora, an AI voice-agent platform for automated lead capture and booking. (Marketer of the Day #1411)

What is rules-based attribution?

Attribution is the job of deciding which ad, click or channel gets credit when a customer converts. A conversion is the action you count, such as a sale or a submitted form. Customers often touch several channels before they convert, so something has to decide how the credit is split, or whether one touch gets all of it.

Google Ads Help defines the rules-based kind in one sentence.

"Rules-based attribution models assign value to each advertising touchpoint based on predefined rules."

A touchpoint is one interaction, such as a click on an ad. Predefined means the rule is fixed before the data arrives. Give the model the same journey twice and it splits the credit the same way twice, so you can check the result by hand.

If you want the wider question of which model suits an owner-led company, read what marketing attribution model owner-led companies should use. This article stays on one narrow point: what each fixed rule credits, which ones Google still offers, and what that does to the report on your desk.

What does each rules-based model credit?

Five names come up: last click, first click, linear, time decay and position-based. The plain definitions below are ours, not quoted from Google.

This is an illustrative example. The journey, the order value and the weights are invented to show the arithmetic. They are not a benchmark and not a real customer.

A customer buys a $1,000 order after four touches, in this order: a paid social ad, an organic search visit, an email click and a paid search ad.

Every row adds up to $1,000. Linear is 1,000 divided by 4, or 250 each. Position-based gives 40 percent of 1,000 to the first and last touch, and splits the other 200 between the two in the middle.

The sale is the same in every row. Paid search looks like the hero under last click, and paid social under first click. Real tools set their own time decay and position-based weights, so your own figures will differ.

How is data-driven attribution different?

A data-driven model does not follow a fixed split. It works the split out from the data. Analytics Help describes the method this way.

"Attribution uses machine learning algorithms to evaluate both converting and non-converting paths."

Google Ads Help gives the Google Ads version.

"Data-driven: Distributes credit for the conversion based on your past data for this conversion action."

Here is our reading. You cannot rebuild a data-driven split with a calculator, and you can rebuild a fixed-rule split in a minute. That does not make one better. It means the report has to name the model.

Which models does Google Ads offer, and which were retired?

We read the Google Ads page on 30 October 2026. It shows no date of its own. It opens with a list.

"Google Ads currently offers several attribution models:"

The page describes last click like this.

"Last click: Gives all credit for the conversion to the last-clicked ad and corresponding keyword."

It describes data-driven in the words quoted in the section above. The page also says what the choice is for.

"Attribution models give you more control over how much credit each ad interaction gets for your conversions."

The four other rules from the table above had their own announcement. The Google Ads announcement is dated April 6, 2023, and its title says first click, linear, time decay and position-based were going away. It gave the reason in a number.

"Today, less than 3% of Google Ads web conversions are attributed using first click, linear, time decay, or position-based models."

Less than 3 percent means fewer than 3 in every 100 conversions. The same page says what would stay.

"Data-driven attribution, last click, and external attribution won’t be impacted."

And it says what would happen to accounts still on a retired model.

"Then, starting in September 2023, we'll switch any conversion actions still using these models to data-driven attribution."

That was a plan stated in April 2023, and we did not check what each account did afterward. A Google Ads report that spans that period can mix models, so ask which model each conversion action uses, and from what date.

What does Google Analytics 4 offer now?

Google Analytics 4 has its own help pages and its own setting, so check it separately. We read the Analytics Help page on changing the reporting attribution model on 30 October 2026. It also shows no date. On the retired rules it says the following.

"The first click, linear, time decay, and position-based attribution models are no longer available as of November 2023."

The same four rules are named in both Google notices. The page describes a Google paid channels option built on last click, and it says what happens when there is no Google Ads click.

"If there is no Google Ads click, the attribution model falls back to Paid and organic last click."

Another line on the page describes a last interaction rule this way.

"Ignores direct traffic and gives all credit to the last interaction before the customer completed the important interaction."

Two more facts from the Analytics help pages matter to an owner. One is about direct visits, meaning people who type your address or use a bookmark.

"All attribution models exclude direct visits from receiving attribution credit, unless the path to key event consists entirely of direct visits."

The other is about timing. A key event is the action Analytics counts as important, such as a form submission.

"Changing the reporting attribution model applies to historical and future data."

That last line means a model change in Analytics can rewrite numbers you have already seen.

Why do the same sales produce different numbers?

Because the model is part of the measurement. The sales do not change. The credit does.

This is a second illustrative example. The sales and paths are invented. They are not a benchmark and not a real account.

In one month, 10 sales come in. Six customers first saw a paid social ad and later clicked a paid search ad. The other four clicked a paid search ad and nothing else. Each sale is worth the same, so we count sales.

Last click gives paid search all 10. First click gives paid social the 6 customers who saw it first, and paid search the other 4. Linear splits each of the six two-touch sales in half, so paid social gets 6 times 0.5, which is 3, and paid search gets 3 plus the 4 single-touch sales, which is 7.

Paid social is credited with 0, 3 or 6 of the same 10 sales. An owner who read only the last click report would cut paid social. One who read only the first click report would fund it. Neither report is wrong. Each answers a different question. The error is comparing reports that do not say which question they answer.

The model matters in a contract too. If a fee or a bonus depends on conversions credited to a channel, the model sets the payout. How to structure a performance-based marketing agreement so it can't be gamed covers the terms. Write the model name into them.

What should you ask your agency about the model?

A report that shows channel results without naming the model is a claim, not a measurement. Ask for these.

1. The model for each platform. Name it for Google Ads and for Analytics. Do not accept one answer for both.
2. Any change during the period shown. Ask for the date. Model changes in Analytics reach back over older data, so ask whether past figures were restated.
3. The conversion being counted. A sale, a form fill and a phone call are different counts, and the model applies to whichever one is set.
4. A last click view beside the headline. Ask for the same results under last click. The gap shows how much the model choice moves the answer.
5. How direct visits are handled. All models in Analytics exclude them unless the whole path is direct, so direct visits will not show up as credited channel results.

For the wider problem of reports that show activity and not proof, see why most marketing reporting does not prove anything.

What do these sources cover, and what do they not?

All four pages quoted here are Google help pages, and Google makes the tools they describe. Only the April 2023 announcement carries a date. The other three show none, so what we say is offered now rests on our reading of them on 30 October 2026. None of the pages gives a figure for how accurate any model is, so this article gives none. The two worked examples are invented to show arithmetic. They are not data from any company.

What is the short version?

Rules-based attribution applies a fixed split to a customer's touchpoints. Last click gives everything to the final touch. First click gives everything to the first. Linear, time decay and position-based spread it by other rules. Google said in April 2023 that it was retiring first click, linear, time decay and position-based in Google Ads, and Analytics says they are gone as of November 2023. What the pages we read still describe is last click and data-driven, plus external attribution in the Google Ads notice. Same sales, different model, different channel totals. So ask which model each report uses, when it changed, and whether old numbers were restated. Then compare reports only when the answers match.