The attribution trap, and what to measure instead

There is one rabbit hole that swallows more founder time than any other in growth measurement, and it is attribution. The promise is seductive: if you could only know which channel produced each customer, you would know exactly where to spend. The reality is that B2B attribution is largely a fiction you are paying to maintain, and chasing it sends you after precisely the wrong numbers.

Last-click lies

The default attribution model, last-click, credits whichever channel a buyer touched immediately before converting. In B2B that is almost always a trackable channel, search or a branded visit or the product itself, so those channels scoop up all the credit. What actually moved the buyer happened long before, and most of it left no footprint at all. This is the dark funnel: the podcast they heard you on, the community where your name came up, the peer who recommended you over coffee, the referral that never touched a tracked link. None of it shows in last-click, so last-click systematically over-credits the channels you can see and erases the ones you cannot.

Self-reported attribution, asking buyers how they heard about you, recovers some of the dark funnel, and it is worth doing. But it is a signal, not the truth. People misremember, and they over-weight whatever happened most recently, so the answers skew toward the last thing they touched rather than the first thing that mattered. Treat self-report as a useful hint about the channels working below the surface, never as a precise ledger you can optimise against to the decimal.

Measure what you control

The honest move, and the one that keeps a lean team out of the measurement rabbit hole, is to stop chasing channel credit you cannot trust and measure the conversion rates between the stages you do control. You may not know with confidence which channel produced a lead, but you know exactly what happens to that lead once it enters your funnel. You know your MQL-to-SQL rate, your win rate, your velocity. Those are yours, they are clean, and they are actionable in a way channel attribution never quite is.

This is not a counsel of despair about attribution, it is a reallocation of trust. Spend your measurement energy on the rates inside the funnel, where the data is reliable and the levers are real, rather than on reconstructing an attribution story from signals that are noisy by their nature. The dark funnel is real and it matters, but the response to an unmeasurable input is not to measure it harder, it is to measure the conversion you can see and let the inputs do their work.

For a founder running growth with a small team and a stack of AI agents, this is liberating. You are not building a multi-touch attribution model that takes a quarter to stand up and lies to you anyway. You are watching five rates you can trust, finding your constraint, and pulling the levers that move it. That is a measurement system you can actually run, and it does not collapse the moment a buyer hears about you somewhere you cannot track.