Cross-Channel Attribution Models Explained: Stop Guessing

Quick Answer

Cross-channel attribution decides how the credit for a sale is split between the channels a buyer touched on the way: for one buyer who clicked a TikTok ad, opened an email and then bought through a branded search ad, last click gives search 100% and linear gives each touch a third. None of these rules says whether a channel caused the sale; that takes a comparison with what would have happened without it.

When the same ads look brilliant one week and useless the next, the problem often isn't your creative or your offer. It's the report: most attribution models describe what happened before a sale, not why it happened, and nothing on the screen says so.

The Siren Song of Last-Click Attribution

The simplest model, and the most misleading, is last-click attribution: 100% of the credit goes to the very last touchpoint. In the journey above, it says the branded search ad did all the work, which is like giving the gold medal to the person who hands a marathon runner a bottle of water at the finish line. Budgets that follow it drift toward the channels that close sales and away from the ones that create demand.

A Parade of Flawed Models: Attribution’s Hall of Shame

Beyond last click, a family of rules-based models offers the look of sophistication. They all fail the same way: they assign credit by a touch's position or timing, not by what it did. GA4 stopped offering four of them (first click, linear, time decay and position-based) in November 2023, but plenty of other tools still do.

First-Click, Linear, Time-Decay & U-Shaped: Credit by Rule, Not by Evidence

First-Click: The opposite of last-click, giving 100% credit to the first touch. Equally wrong.

Linear: Divides credit equally among all touchpoints. A participation trophy for your marketing channels.

Time-Decay: Gives more credit to touchpoints closer to the conversion. Better, but still arbitrary.

U-Shaped (Position-Based): Gives credit to the first and last touch, with the rest split in the middle. A slightly more complex guess.

These models were built for journeys you could track end to end. Since iOS 14.5 in 2021, Apple has required apps to ask permission before tracking people across other companies' apps and websites, and many journeys now arrive with holes in them. A rule that splits credit over the touches it can see has no way to account for the ones it can't.

How to Solve This: From Correlation to Causality

Moving from correlation to causality means asking, for each channel, what sales would have looked like without it. That's causal attribution, and you can start on it with data you already have.

Three checks before you move budget between channels:

1. Put the credited sales next to your orders. Add up the purchases Meta, Google and TikTok each report for one week and compare the total with your Shopify orders for the same week. When the platforms together claim more sales than you made, they're counting the same buyers more than once: every platform grades its own homework.
2. Watch MER. Total revenue divided by total ad spend, week by week. It can't tell channels apart, but no attribution model can inflate it, so it shows whether the mix as a whole is working.
3. Test one channel with a holdout. Cut or pause its spend in one region for two weeks, keep everything else the same, and compare that region's revenue with the rest, against how they compared in the weeks before. The change is what that channel caused, with no model in between; the incrementality testing guide covers the set-up.

How sure is a Causality Engine estimate?

As sure as its data, and it says how sure: a causal attribution tool like Causality Engine gives every channel a data-health score from 0 to 100 that grades the read, not the channel. Under 90, move budget in steps, and test each step the free way: change spend a little in one region, then compare revenue.