Measuring what works

Marketing attribution: what it answers, why it's hard, and what a small team should do

Marketing attribution is the practice of working out which marketing activity led to each sale, so that the channels that earned it get the credit.

About the Google Analytics 4 connectorAbout the Google Ads connectorAbout the TikTok Ads connectorAbout the Shopify connector

Why it matters

Attribution tries to answer what you got for the money. Each platform answers for itself, and the answers overlap. A customer who saw a TikTok ad and later clicked a Google ad before buying is claimed by both. Attribution is the attempt to settle that claim once, on your terms rather than the platforms'.

The decision it changes is where next month's money goes. Get the credit wrong and the budget follows it.

How it works

Attribution starts with a path: the touches a customer had with your marketing before they bought. A rule, the attribution model, then splits the sale's value across the touches. Last click gives everything to the final touch; the other models spread it.

The difficulty is building the path in the first place. Here is an illustration with made-up numbers of what a small shop sees in one month.

Each platform counted every order it touched, and many orders were touched by more than one. The claims add up to 1,400 against 1,000 real orders, so at least 400 of the credits are shared. Attribution is the work of dividing the 1,000 real orders between the three, and the platforms' own figures can't do it, because each one only sees its own touch.

Three things make the path hard to reconstruct.

Cross-device journeys break the chain. A customer who sees an ad on their phone and buys on a laptop at work looks like two people, and the touch on the phone never joins the sale on the laptop unless something ties them together.

View-through conversions are counted on faith. A platform that showed someone an ad they never clicked, and saw them buy a week later, will often count that as a conversion. Whether the ad did anything is unknowable from the data.

Walled gardens keep their paths to themselves. The large ad platforms report what happened inside their own systems and share little of it, so the touches you can see are the ones that landed on your site with a tag attached. The rest is modelled, by them, in ways you can't inspect.

What it can't tell you

Attribution divides credit; it does not measure cause. A touch can appear on every path and have changed nothing. Incrementality testing closes that gap by withholding a channel from a control group and measuring what the channel added, rather than how often it was present.

It also can't see channels without touches. Podcast ads, out-of-home and word of mouth leave nothing in the path. Marketing mix modelling handles those by working from total sales and total spend over time, with no path at all.

When to use it

A small team should do less of this than the tools invite. Pick one model (last click is fine) and apply it in one place, usually Google Analytics 4, so that there is a single set of channel numbers that add up to the real total. Each platform's own reported conversions are useful for optimising inside that platform and no use for comparing across them.

Then spend the effort saved on one holdout test a year, on the channel whose numbers look best.

Move to MMM only when the channel count makes testing impractical and you have the history to feed it.

Tracking it in a spreadsheet

The inputs are conversions by channel from your analytics tool and actual orders, with each platform's own reported conversions beside them for the comparison above. A scheduled Fresh Sheets report can keep Google Analytics 4 key events by channel group, Google Ads and TikTok Ads conversions, and Shopify orders each current in their own tab. Fresh Sheets replaces each tab's contents on every refresh, so build the claimed-against-actual comparison on a separate tab.