Multi-touch attribution models explained with one order

For illustration: one 120-euro order with four touches gets five different credit splits from five rules, from all 120 euros to the last touch down to 30 euros per channel, and no rule changed the sale. GA4 now offers three models: data-driven, paid and organic last click, and Google paid channels last click. Its help page says first click, linear, time decay and position-based are no longer available as of November 2023. The last section shows you how to rebuild the removed rules on your own paths.

How do five rules credit the same order?

Take one order and four touches, oldest first. The rules differ only in how they split the credit.

For illustration (invented numbers): one order worth 120 euros, four touches

  Paid Social     21 days before the order
  Organic Search  14 days before
  Email            7 days before
  Paid Search      0 days before (same day)

Credit in euros               Paid Social  Organic Search  Email  Paid Search
Last click                              0               0      0          120
First click                           120               0      0            0
Linear                                 30              30     30           30
Position-based (40/20/40)              48              12     12           48
Time decay (7-day half-life)            8              16     32           64

Google's documentation, still online for the legacy Universal Analytics reports, defines the multi-touch rules plainly. Linear "gives equal credit to each channel interaction on the way to conversion." Position-based splits the credit between the ends and the middle: one common scenario gives 40% to the first interaction, 40% to the last and 20% to the interactions in the middle. Time decay favours recent touches: a touch 7 days before the order gets half the credit of a touch on the day of the order.

Google's legacy overview works a similar four-touch example, where position-based gives the first and last channels 40% each and the two middle channels 10% each. For illustration: with a 7-day half-life, touches at 0, 7, 14 and 21 days carry weights of 1, 0.5, 0.25 and 0.125, which add up to 1.875, so the last touch gets 120 × 1 ÷ 1.875 = 64 euros.

Which of these rules does GA4 still offer?

Three, and none of them is a split like the ones above. GA4 lists data-driven attribution, paid and organic last click, and Google paid channels last click. The same page says the first click, linear, time decay and position-based models are no longer available as of November 2023. It adds that every model excludes direct visits from credit unless the whole path is direct.

Google's reasoning, in its 6 April 2023 announcement, included a usage figure: less than 3% of Google Ads web conversions were attributed using first click, linear, time decay or position-based models (Google data, global, February to March 2023, published by Google and not independently audited). That measures how little the rules were used in Google Ads. It doesn't say the rules were wrong, and it says nothing about your store.

What does data-driven attribution do differently?

It replaces the fixed formula with an estimate from paths. Google says the model uses your account's data to calculate the actual contribution of each click interaction, and describes a counterfactual approach that contrasts what happened with what could have occurred. Google's own illustration: four ad exposures lead to a 3% probability of a key event, and without the fourth the probability drops to 2%, so the fourth is credited with +50% key event probability.

The same page says the models compute the counterfactual gains of Google ad exposures by training on randomized controlled trials. It publishes no accuracy figure for the model. You get a credit split you can't reproduce by hand and a description you can't audit from the page. Google builds and reports the model and also sells the ads it credits, so treat it as one view, not the answer.

How do I rebuild the removed rules from my own GA4 data?

Do this in a spreadsheet. It takes one export and costs nothing.

1. In GA4, open the Key events attribution paths report under Advertising. Pick your purchase key event and download the data table with Share this report. The report shows paths up to 20 touchpoints long.
2. Delete Direct from every path. GA4's models give Direct no credit unless the whole path is Direct, so your rebuild has to do the same.
3. For each path row, give its purchase revenue to channels by the rule: last click to the last channel, first click to the first, linear in equal parts. For position-based, give 40% to each end and 20% across the middle.
4. Add up the credit by channel. Then open Attribution models under Advertising, pick paid and organic last click for the same dates, download it with Share this report and put it beside your last-click column.

Pass: your last-click column lands on GA4's for every channel, or differs by a gap you can name, such as paths longer than 20 touchpoints or a different reporting time. Your other columns are then arithmetic you can trust. Fail: a gap you can't name, so check Direct and path length before anything else.

Then read the spread. A channel that moves from first place to last between rules is one where the rule, not the data, is deciding. Skip time decay in the spreadsheet: it needs a date for every touch, which means the event-level BigQuery export. Don't move budget on any rule's credit alone. Change that channel's spend a little in one region and compare revenue with a region you left alone.

Sources, 30 September 2026: Get started with attribution (Google Analytics Help, 2026); First click, linear, time decay, and position-based attribution models are going away (Google Ads Help, 2023); Overview of Attribution modeling in MCF, legacy (Google Analytics Help, legacy Universal Analytics); About the default MCF attribution models, legacy (Google Analytics Help, legacy Universal Analytics); Key events attribution paths report and Key event attribution models report (Google Analytics Help, 2026).

Related answers

• Single-touch vs multi-touch attribution on the same orders
• Attribution models compared: rules, data-driven, and MMM
• Why multi-touch and last-click numbers disagree
• Causal attribution
• Multi-touch attribution in the glossary