Marketing attribution tools: four categories wearing one label

A typical attribution shortlist has HubSpot on it, a multi-touch tracking platform, Google Ads' own conversion reporting, and something doing media mix modelling. They arrive as one ranked list with star ratings, as if they were competing for the same purchase.

They are not comparable. Those are four different product categories that answer four different questions, and the reason attribution evaluations stall is usually that nobody has established which question is being asked.

Sorting that out first removes most of the difficulty. It also exposes an uncomfortable fact: for a large share of companies, the reason the numbers are wrong is not the absence of a tool.

The four categories

CRM-native reporting is the one most companies should exhaust first. It is included in software you already pay for, it connects to actual closed revenue rather than to conversions, and it is the only category that natively speaks in pounds invoiced. Its weakness is that it only knows what was recorded, which is why it collapses when the source field is optional.

Dedicated multi-touch platforms capture the journey across channels and devices, stitching sessions to identities. They see more than the CRM. What they add is resolution on the path; what they cannot add is certainty about cause, and their output is only as good as the tagging discipline feeding it.

Ad platform reporting is free, immediate, and structurally optimistic. Each platform sees only its own touches and claims credit under its own rules and attribution windows. Add up the reported conversions from three ad platforms and you will exceed the number of deals you actually closed — not because anyone is lying, but because the same deal is being counted by each of them.

Modelling and incrementality — media mix modelling, geo holdout tests, controlled experiments — is the only category that attempts the counterfactual. It is the only one that can tell you what would have happened without a channel. It works on aggregates rather than individuals, needs a long and varied spend history, and answers budget questions rather than operational ones.

Every one of them needs the same thing from you first

Here is the part vendors are quiet about. All four categories depend on being able to join a marketing touch to a revenue outcome. That join needs a stable key.

In practice, most organisations do not have one. The same buyer appears as a cookie on the ad platform, an email address on the form, a lead record in the CRM, a separate contact record created by a rep who did not search first, an opportunity under a company name spelled differently, and an invoice under the registered legal entity. Six identities, one person.

No attribution tool resolves that for you. Every one of them assumes it has been resolved and produces confident output regardless. That is the failure mode worth planning around: the tool does not report an error when identity resolution is broken. It reports a number.

Which is why the honest first question is not "which tool" but "can we currently trace a single closed deal from its first recorded touch to its invoice, by hand?" If the answer is no, buying a platform converts a data problem into a data problem with a subscription. Fixing CRM data quality is unglamorous and comes first.

The question that picks the category

Work from the decision you need to make, not from the feature list.

That last row is the most common presenting complaint and the one least likely to be solved by a purchase. Three systems reporting three different figures for the same channel are usually applying three different credit-splitting rules correctly. The model is a policy you chose, not a measurement, and buying a fourth tool adds a fourth number.

What to check before you sign

Five questions that expose most of the risk:

1. 01What is the join key between a touch and a closed deal? If the vendor cannot describe this in your stack specifically, the demo was running on their data, not yours.
2. 02What happens to offline touches? Events, calls, conversations at a conference. In long B2B cycles these are often decisive and almost never tracked. A tool that silently drops them will systematically over-credit digital.
3. 03Which model does it default to, and can that be changed? A platform with one hard-coded rule has made a policy decision on your behalf.
4. 04Can it report on revenue, or only on conversions? A conversion is a form fill. Revenue is an invoice. Tools that stop at the former cannot answer budget questions.
5. 05What is the implementation actually going to require from us? The answer is usually tagging discipline, CRM field enforcement and someone's ongoing attention. Budget for that, not just the licence.

What "best" means here

There is no best attribution tool, in the same way there is no best measuring instrument. There is a best fit between a category and a question, and a large amount of preparatory work that determines whether any tool in that category will produce something trustworthy.

The sequence that works:

1. 01Decide what a lead, an opportunity and a closed deal mean, and enforce it.
2. 02Establish the join key so one buyer is one identity across systems.
3. 03Exhaust CRM-native reporting, which is free and revenue-connected.
4. 04Then, if a specific question remains unanswered, buy the category that answers it.

Most companies attempt this in reverse, and the tool inherits the mess. The order is the recommendation.

If you want the underlying concepts rather than the shopping list, marketing attribution covers how the measurement is built and revenue attribution models covers choosing the credit-splitting rule itself.