Which of your marketing is working? Every model in plain words, why they are all partly wrong, and what to do instead.
CTFM Team
The question you actually have is simple: which of my marketing is working? Attribution is the industry's attempt to answer it, and the honest summary is that every method is partly wrong.
That is not a reason to skip it. It is a reason to hold the answers loosely.
Attribution is the job of deciding which piece of your marketing gets the credit for a sale, when in truth several things nudged the person along the way.
Someone reads an article, sees an ad two weeks later, gets an email, then buys. Which one caused it? All of them contributed and none of them was sufficient alone. Attribution models are different rules for splitting the credit.
Analytics versus attribution
Analytics tells you what happened: pages viewed, sessions, sales. Attribution tries to tell you why it happened, by handing out credit for the cause. The first one is measurement. The second is interpretation, and that is exactly why people argue about it so much.
All credit to the first interaction. Like handing all the credit to the friend who first mentioned the restaurant, and none at all to the review you read afterwards or the evening you walked past and looked at the menu in the window.
Last touch
All credit to the final interaction. Like crediting the waiter who handed you the menu for your decision to eat there.
Linear
Credit split evenly across every touch. It feels fair, and it is almost certainly wrong, because some of those moments clearly mattered a great deal more than the others.
Time decay
Recent touches get more credit. Reasonable for a quick decision, and misleading for a long one where something months earlier did the real work of convincing them.
Position based
Extra weight to the first and last touch, the rest shared. A compromise between the two simple models.
Data driven
An algorithm assigns credit based on patterns in your own data. Better in principle, and a black box you cannot fully inspect.
Last touch remains the most widely used, largely because it is the default in most tools. It is also the one that most reliably misleads.
Why last touch quietly wrecks decisions
Last touch systematically over-credits whatever happens closest to the purchase. That means retargeting ads and branded search look brilliant, because they catch people who had already decided. Meanwhile the content and campaigns that created the demand look worthless. Cut those on last-touch data and your pipeline dries up two months later for reasons the dashboard cannot show you.
If your tracking seems less reliable than it used to be, that is not your imagination.
Change
Effect on measurement
Apple's App Tracking Transparency
Users can decline cross-app tracking, so mobile app-based measurement lost a large share of its signal
Browser restrictions on third-party cookies
Following a person across different sites became much harder
Consent requirements under privacy law
Users who decline tracking are invisible to your analytics, and they are not a random sample
Email open tracking distortion
Pre-loaded images register opens that did not happen
Chrome's plans for third-party cookies have changed direction more than once. So check the current state yourself. Do not trust an article on that detail, including this one.
The practical consequence is that a meaningful share of your customers now appear as "direct" or "unknown". They did not arrive by magic. You just cannot see the path.
Serious teams stopped trying to track every individual and moved to methods that measure effect rather than paths.
Turn something off, or run it in some regions and not others, and compare. Genuinely tells you whether a channel caused anything, because you created a control group.
The best method on this list, and the one most people avoid because switching off spend feels frightening.
Just a free text box on your form, read once a month, costing you nothing at all to add. The answers will surprise you.
Cost per customer, per channel, roughly
Total spent on a channel divided by customers you can reasonably tie to it. Imprecise and directionally useful.
Whether the total is going up
Total enquiries and total customers, month over month. If the total rises while you cannot explain which channel did it, something is working. Do not stop everything to find out.
One holdout test a quarter
Pause one thing for a few weeks and see whether anything changes. This single habit teaches you more than any dashboard.
The most useful attribution question
Not "which channel gets credit" but "what would happen if I stopped doing this". That reframing turns an unanswerable accounting problem into a testable one.
