What is ad attribution?
Attribution is the practice of deciding which marketing touch gets credit for a conversion. Everything difficult about it follows from the fact that there is usually more than one honest answer.
Definition
A working definition
Ad attribution assigns credit for a conversion to one or more of the touches that preceded it. A touch is any interaction you can observe — an ad click, an impression, an email open, an organic visit. The conversion is whatever you actually care about: a purchase, a trial, a booked call, an enrolment.
The hard part is not the arithmetic, it is the joining. Before any model can allocate credit it has to know that the phone session on Tuesday and the desktop checkout on Friday were the same person. Most attribution disputes are identity problems wearing a modelling costume.
Attribution is a decision tool, not an accounting record
No model produces the objectively correct answer, because credit is a judgement about influence rather than a fact you can measure. A good model is the one that makes your next budget decision better than the alternative — which is why the model you choose should follow from the decision you are trying to make.
Why the numbers disagree
Why platform pixels under-report
Every ad platform grades its own homework, using only the data its own pixel could see. Four structural gaps explain most of the disagreement with your own numbers.
Cross-device journeys
A click on a phone and a purchase on a laptop are two anonymous sessions unless something durable links them. Where the platform cannot recognise the person, the conversion is simply never credited.
Cookie lifetime and blockers
Browser storage is short-lived and routinely cleared, and tracking-prevention features shorten it further. A funnel whose purchase lands weeks after the click loses the record that would have connected them.
Attribution windows
Platforms only count conversions that land inside a fixed window after the click. Long consideration cycles fall outside it entirely, so the campaigns that start slow-burning journeys look like the worst performers.
Browser-only events
Events fired from the browser are lost whenever a script is blocked, a tab closes early, or a checkout completes inside an in-app browser. Server-side events survive all three.
Attribution models
The four attribution models you will meet
Each one answers a different question. Reading them as competing estimates of the same truth is the most common way teams get attribution wrong.
| Model | How credit is assigned | Best for answering | Watch out for |
|---|---|---|---|
| First-touch | The entire conversion is credited to the earliest recorded touch in the journey. | Which channels create demand and introduce you to people who did not know you existed. | It ignores everything that closed the sale, so it flatters awareness channels and starves the ones that convert. |
| Last-touch | The entire conversion is credited to the final touch before it happened. | Which channels capture demand that already exists and sit closest to the money. | Branded search and retargeting look extraordinary because they sit at the end of journeys other campaigns paid for. |
| Linear | Credit is split evenly across every recorded touch in the journey. | A quick, defensible view of which channels appear in your winning journeys at all. | It treats a decisive touch and an incidental one identically, and it rewards channels that simply appear often. |
| Data-driven | Credit is allocated by a model fitted to your own converting and non-converting journeys. | Splitting budget across many channels once you have enough conversion volume to fit a model. | It needs volume and clean inputs; on sparse or partial data it produces confident numbers built on very little. |
Whichever model you choose, keep one as your decision-making default and read the others as second opinions. Switching models to explain away a bad month is how attribution loses its credibility inside a company.
Signal quality
What makes a model worth trusting
Durable identity
One person, one profile, across devices and sessions. Without it, every model is allocating credit between fragments of the same journey and calling them different people.
Server-side events
Conversions sent from your own systems survive blocked scripts, closed tabs and in-app browsers. They are also the events your finance team can reconcile.
Outcomes, not proxies
Feed the model the event that represents money — paid conversion, closed deal, enrolment net of refunds — rather than the cheap proxy that happens to fire soonest.
Live analyzer
See which of these gaps applies to your funnel
Describe your funnel in a sentence or two. The analyzer writes back which attribution gaps are most likely costing you credit, and what to instrument first. Free and anonymous, no account.
FAQ
Attribution questions, answered plainly
No. Tracking is the collection of events; attribution is the interpretation of them. You can track everything and still attribute badly, and you can attribute sensibly from a modest set of well-joined events. Collection is a plumbing problem, attribution is a decision problem.
Start from the decision. If you are choosing where to open a new channel, first-touch tells you what creates demand. If you are deciding what to cut this week, last-touch sits closer to the money. If you have real conversion volume and need to split a fixed budget across many channels, a data-driven model earns its complexity.
Because they are measuring different things by design. Ad platforms count conversions they can see within their own window and often include view-through credit; your own systems count money whenever it arrives. Aim for a stable, explainable gap rather than an exact match — a gap you understand is more useful than a total that agrees by accident.
They make third-party tracking harder and first-party measurement more valuable. Attribution built on your own consented data, server-side events and aggregated modelling remains workable. What no longer works is quietly following individuals around sites you do not own.
Keep reading
Keep exploring the attribution stack
How attribution approaches compare
There are four broad ways to answer “which ad produced that sale”. Each is genuinely good at something, and each fails in a way the others do not.
Read moreAd attribution for SaaS and subscription products
Trials are cheap to buy and expensive to keep. Attribution built for SaaS scores campaigns on the revenue that survives month three, not the signup that spikes on day one.
Read moreAd attribution for e-commerce stores
Blended ROAS hides the truth: some campaigns are credited twice while others are never credited at all. Store-level attribution puts one order behind one click.
Read moreFind the gap before it costs you another quarter.
Run the free analyzer on your own funnel, then continue on our partner platform when you want a full attribution stack behind it.
Hyros is an independent attribution explainer and demo site. The analyzer writes a live read-out of your described funnel; the full report step is illustrative and is not connected to your ad accounts.