Glossary - Acquisition and attribution

Multi-touch attribution

Multi-touch attribution (MTA) is a way of assigning credit for a sale across several of the marketing touchpoints a customer passed through before buying, instead of handing all of it to one. Each order's value is split between the channels in its journey by a weighting rule. Ecommerce operators use it to see which channels open, assist and close sales, and to avoid starving the channels that never hold the final click.

Channel credit = Sum over orders of (order value × weight given to that channel's touches in the journey)

Variable	What it covers
Order value	The revenue (or better, the contribution margin) of each order the tracking could tie to a journey.
Journey	The ordered list of tracked touches before the purchase: ad clicks, email clicks, organic visits. Untracked touches are simply missing.
Weight	The share of the order each touch receives. Weights in one journey always add up to 1, so no order is credited twice.
Weighting rule	Linear (equal shares), time decay (more to recent touches), position based (more to first and last) or data driven (fitted by an algorithm). See attribution model.

The weights adding to 1 is the useful property: total credited revenue equals the revenue the tracked orders actually brought in.

Worked example

Order 1, A$120: Meta click, then email click (linear, 2 touches)Meta A$60, email A$60

Order 2, A$90: Google click onlyGoogle A$90

Order 3, A$150: TikTok, then Meta, then Google (3 touches)A$50 each

= Linear totals: Meta A$110, email A$60, Google A$140, TikTok A$50A$360

Store revenue from the three ordersA$360

If each platform claims every order it touched: Meta A$270, Google A$240, TikTok A$150, email A$120A$780

Platform claims above real revenueA$420

Worked example / demonstrative numbers. Multi-touch attribution keeps the total honest. It does not tell you whether any of the touches were needed.

What is good multi-touch attribution?

Good multi-touch attribution is honest about what it is: a credit-sharing rule applied to the touches your tracking could see. Its real strength is structural. Because each order's weights sum to one, the channel totals add back to real revenue, which the platforms' own reports do not. Meta over-reports roughly 26% above third-party analytics on average, per the sources cited on The Math, and summing every platform's claims counts the same order several times, as platform over-claiming shows.

Its weakness is that it is still correlational. Every rule divides credit among touches that happened to occur; none can say whether the sale needed them. Cookie loss, cross-device journeys and view-through impressions also leave gaps that no weighting fixes. So the workable setup compares several rules side by side, reads the channel ranking that stays stable across them, and settles the expensive questions with incrementality tests. When the arguing is really about which platform owns a sale, the problem is covered in who gets the credit.

Multi-touch attribution vs related methods

Method	What it does	How it differs from multi-touch attribution
Last-click attribution	Gives 100% of credit to the final click.	A single-touch rule; the special case MTA was built to replace.
Marketing mix modeling	Estimates channel effect from aggregate spend and sales over time.	Top down, no user-level tracking, and it can see channels MTA cannot track.
Incrementality	Measures sales a channel caused, against a holdout.	Causal and episodic; MTA is correlational and always on.
MER	Total revenue ÷ total marketing spend.	Needs no attribution at all, so it cannot be distorted by the weighting rule.
POAS	Attributed profit ÷ ad spend.	Consumes MTA output; change the rule and POAS changes with it.

Common mistakes

Treating data-driven as causal. Algorithmic weights are fitted to observed journeys. They still describe correlation, not what would have happened without the ad.
Crediting revenue instead of margin. A channel that assists full-price orders and one that assists heavily discounted orders look the same on revenue. Weight contribution margin, per the ROAS formula and its margin versions.
Ignoring what the tracking cannot see. Podcasts, word of mouth, in-feed views without a click: they are absent from the journey, so MTA quietly gives their credit to whatever was tracked.
Reading a rule change as a performance change. Moving from linear to time decay reshuffles the channel league table. Nothing about the business moved.
Using it to set total budget. MTA divides credit within the spend you already have. How much to spend overall is a question for blended CAC, MER and payback.

Multi-touch attribution FAQ

Single-touch attribution gives the whole sale to one touch, usually the last click or the first. Multi-touch attribution splits the sale across every tracked touch using a weighting rule, so channels that introduce or assist a buyer get some credit instead of none.

None is correct in a causal sense; each is a different opinion about weighting. Linear is simple and transparent, time decay favours closers, position based favours openers and closers. Compare several and trust the conclusions that survive all of them, then test the big ones with a holdout.

Less than it was. Lost cookies, opted-out users and cross-device journeys break the chain of touches, so journeys look shorter than they are and the last tracked touch gets more credit. Use MTA for direction, and check totals against store revenue and MER.

No. It shows where a channel appeared in journeys that ended in a sale. Whether the sale would have happened without it is an incrementality question, answered by a holdout or geo-lift test.

Attribution model Last-click attribution Incrementality Marketing mix modeling ROAS MER Blended CAC calculator All 14 free tools →

Inside Blufire, S5 Acquisition runs the Model Comparison Matrix so attribution models sit side by side instead of one being trusted, Margin Triangulation checks their claims against the margin that actually arrived, and one Canonical Source Map is published for every other section to read.

Updated September 2026