Attribution You Can Act On, Without the Fantasy of Perfect Data

The Search for a Number That Does Not Exist

Somewhere in most marketing teams there is a project to finally get attribution right. It has been running for a while. It involves a new tool, a tracking plan, and a promise that once it lands, everyone will know exactly which channel produced which sale. It rarely lands, and when it does the numbers disagree with the finance report anyway.

The reason is not incompetence. It is that a real buying journey involves conversations, a recommendation from a colleague, a podcast, and three visits from different devices, most of which no tracking system will ever see. Building a model that pretends otherwise produces confident numbers that are quietly wrong.

What Attribution Is Actually For

The point of measurement is not to assign historical credit. It is to answer one forward looking question: where should the next unit of budget go? That question tolerates far more uncertainty than most teams assume. You do not need to know that a channel produced 31.4 percent of revenue. You need to know whether it is clearly working, clearly not, or too small to tell yet.

Once the goal is framed that way, most of the complexity falls away, and a workable setup can be in place in a few weeks rather than a few quarters.

Four Things Worth Building

We install the same four pieces for almost every client, in this order.

• One source of truth for revenue, owned by finance, that marketing reports against rather than beside.
• A self reported field on every form asking how the person heard about you, in their own words.
• A consistent, boring naming convention for campaigns, enforced by one person.
• A small number of holdout tests, where a region or segment is deliberately left out of a channel for a set period.

None of these require a new platform. Three of them are process rather than technology, which is precisely why they get skipped.

The Self Reported Question Earns Its Place

The single highest value line in most measurement setups is a free text field asking how someone heard about you. It captures exactly the influences that tracking cannot see: the colleague who recommended you, the conference talk, the newsletter forwarded by a client.

Teams resist it because the answers are messy and cannot be summed neatly. That is the wrong test. Read fifty of them and patterns appear immediately, and those patterns are usually the thing the dashboard was missing rather than a duplicate of it.

Holdouts Tell You What Models Guess

If you want to know what a channel genuinely contributes, stop running it somewhere. Pick a region or a segment that is large enough to produce a signal, turn the channel off for a defined period, and compare. It is uncomfortable, it costs a little revenue, and it is the only method that produces an answer nobody can argue with in a meeting.

The discipline is in the size and the duration. A one week pause tells you nothing except that your pipeline has a lag. Six to eight weeks in a meaningful segment usually does.

Decide What the Numbers Are Allowed to Do

A measurement setup that nobody acts on is a reporting hobby. Before the first report goes out, write down the decisions the numbers are permitted to trigger and the threshold for each one. For example: a channel below a defined cost per opportunity for two consecutive months gets a budget reduction and a review. A channel above target for two months gets an increase.

Agreeing those rules in advance removes the part of the process that usually goes wrong, which is the monthly negotiation about whether a disappointing number was real. It also protects promising channels from being cut in their first slow month.

If no decision changes when a number moves, the number was a decoration.

Reporting Cadence, Not Reporting Volume

Most teams over report and under review. A weekly dashboard that nobody opens is worse than a monthly one that two people argue over for an hour, because the weekly version creates the impression of oversight without any of the substance.

We suggest a one page monthly review with three sections: what we spent, what we got, and what we are changing as a result. The third section is the only one that matters, and it is the one most reports omit entirely.

Accepting the Fuzziness

The healthiest measurement culture we see is one that treats every number as directional. It is comfortable saying that a channel looks strong, that another is unclear, and that a third needs a test before anyone defends it further. It spends its energy on decisions rather than on reconciling two dashboards that will never agree.

That posture is not lowering the bar. It is the difference between a team that keeps rebuilding its tracking and a team that keeps moving budget toward what works, which is the only outcome the exercise was ever supposed to produce.