Multi-touch attribution: what actually remains possible
Multi-touch attribution as it is still sold no longer exists in Google's tools. Four models were removed in 2023 and only two remain: last click, and data-driven attribution, whose split is published nowhere. The choice therefore comes down to a simple rule or a black box.
- Google states that the first click, linear, time decay and position-based models are no longer supported. The conversion actions that used them were switched over automatically.
- What remains is last click, which is simple and wrong in a predictable way, and data-driven attribution, which is more accurate and unverifiable.
- No formula is published for the second. You cannot reproduce its calculation or dispute it with numbers.
- The practical consequence is not to choose the right number, it is to choose one number and stop changing it during the evaluation period.
- The question that decides a budget is not which channel deserves the credit, but how much a customer costs in total, across all channels.
On this page
Data-driven attribution
Data-driven attribution is the default way of distributing the credit for a conversion across the advertising interactions that preceded it, in Google Ads and in Google Analytics 4. Unlike the older fixed-rule models, it calculates a variable share for each touchpoint based on the account's history. Its distinctive feature is that the calculation differs from one advertiser to the next and is not published, which makes it more faithful in theory and unverifiable in practice.
What disappeared, and what it changes
The market's vocabulary is several years behind the tools. Multi-touch attribution engagements are still sold on the promise of a choice between five or six models, while Google's Help Center is explicit. The first click, linear, time decay and position-based models are no longer supported, and the conversion actions that used them were switched to data-driven attribution. Reverting to last click is still possible.
For a management team, this removes a useful illusion. The linear model gave an equal share to each touchpoint: it was not accurate, but it was understandable, reproducible and open to debate in a meeting. Data-driven attribution is probably more faithful, and no one in your organization can redo the calculation or dispute it with numbers.
The financial risk is therefore not misattributing. It is making a decision to cut or raise a budget in the belief that it rests on a measurement, when it rests on an estimate you cannot audit. That is a difference in nature, not in precision.
What to settle before funding an attribution project
Attribution is the project that consumes the most analysis time for the fewest decisions changed. Before committing a budget to it, check which precise decision you expect to make differently once the work is done.
- What concrete decision would we make differently if we had perfect attribution?
- Do we know our cost per acquired customer across all channels, before even splitting it up?
- Which attribution model are we using today, and has it been changed during the period we are analyzing?
- Do our offline closed sales feed back into the tool, or are they missing from the calculation?
- If the project changes no decision within six months, what do we stop funding?
A solid answer names the target decision before talking about tools, and points out that only two models are available. An evasive answer offers a choice among five models, promises exact attribution, or sells a third-party tool before checking that closed sales feed back in.
Deciding whether an attribution project will actually change a budget decision is settled on your real sales structure. A 90-minute consultation settles it, with a written summary your team can execute.
What remains possible
Four approaches remain, with different costs and reliability. None gives the truth, and that is the first thing to accept.
| Approach | What it gives | What it costs | When to choose it |
|---|---|---|---|
| Last click | A simple number, wrong in a predictable and consistent way | Nothing | Short cycle, one or two channels |
| Data-driven attribution | A more faithful split, not verifiable | Nothing, it is the default | Several channels, enough volume |
| Ask at purchase | What the customer says themselves | One form field | Always, as a complement |
| Turn-off test on a channel | Real causality, not a correlation | The revenue lost during the test | Heavy decision, large budget |
The last two rows are the ones the market mentions least and that decide the most. A free-text field asking how the customer heard of you produces imperfect data, but data independent of the platforms: it captures what no tool sees, word of mouth, a referral, a billboard. The turn-off test, for its part, is costly in revenue, and it is the only honest way to prove that a channel truly produces sales.
What stays in-house: confirming closed sales, especially when they close over the phone or in person, and the margin by product line. Without these two figures, no attribution is worth the time spent on it. What gets delegated: setting up the tracking, importing offline sales, choosing and maintaining a single model, the periodic reading and the design of the turn-off tests. A business that feeds its closed sales back in gets a better reading than a business that buys a third-party attribution tool.
The reading that actually decides
The trap of this topic is to look for which channel deserves the credit, when the question that commits money is elsewhere: how much a customer costs in total, and whether that number is going down or up.
This calculation is done without attribution. Take the total marketing spend for a month, divide by the number of new customers in the same month, and track that curve over twelve months. It is crude, it is insensitive to attribution windows, and it is the only number a management team can set against any platform report. The break-even logic is detailed in how to calculate real campaign profit.
Changing the attribution model in the middle of an evaluation period makes any comparison unusable, and it is a frequent mistake because the change looks harmless. Set the attribution model before you launch, note the date, and do not touch it again until the period ends. A report that compares two months measured with two different models measures only the change of model.
The indicators that change a decision are detailed in marketing indicators, reading tests in A/B testing, and the trade-off between capturing and building demand in growth marketing.
This point sits inside the approach described in what conversion rate optimization really fixes.
Bringing an acquisition spend back to a defensible return is at the heart of the Optimize the profitability of your digital campaigns goal.
Already running a marketing team? See how we plug in as reinforcement on conversion rate optimization.
Frequently asked questions about multi-touch attribution
Which attribution models are still available?
Two. The Google Ads Help Center states that the first click, linear, time decay and position-based models are no longer supported, and that the conversion actions that used them were switched to data-driven attribution. Last click remains available.
Can you verify the calculation of data-driven attribution?
No. The split varies from one advertiser to the next and the formula is not published. That is the trade-off to accept: the result is probably more faithful than a fixed rule, and no one in your organization can reproduce it or dispute it with numbers.
Should you buy a third-party attribution tool?
Only after feeding offline closed sales back into the existing tools. That is the most frequent and most costly gap, and no third-party tool fills it for you. A tool added before that step measures the same incomplete journey, at a higher price.
Which number should a management team look at?
The cost per acquired customer across all channels, tracked over twelve months. It is calculated by dividing total marketing spend by the number of new customers, without attribution. It is imprecise by channel and reliable as a trend, and it is the only number you can set against any platform report.
- Google Ads Help Center, About attribution models, official documentation, accessed July 2026.
- Google Ads Help Center, About Target ROAS bidding, on the link between attribution and automated bidding, accessed July 2026.
Gabriel almost always takes your first call and carries out your audit. He builds the strategy starting from your growth goal: where to put your budget, which market to test and how to connect each lead to a real sale in your CRM. He mainly leads engagements for three goals: Optimize the profitability of your digital campaigns, Develop a new market, and Generate demand and growth. With Geneviève, he also works on organic search (SEO), AI visibility (GEO) and conversion rate optimization (CRO). The sales a Google Ads or Meta Ads campaign brings in depend on the page that receives the click. He writes mainly about marketing strategy, paid advertising and measurement.
About Falia →
SEO
Paid advertising
Conversion
AI visibility
Web design