Measure what you cannot see: self-reported attribution

The channels that move B2B buyers most are often the least trackable. A podcast appearance, a recommendation in a private Slack, a forwarded newsletter, a conversation at an event, none of these leave a clean referrer for your analytics to record. The buyer hears you on a show, types your name into Google a week later, clicks an organic result, and converts, and your dashboard proudly credits organic search for a lead the podcast actually created. The tracked attribution is not wrong, exactly; it is just blind to the touch that mattered. Optimising only what you can track means optimising a partial picture, and pouring more budget into the channel that closed the loop while starving the channel that opened it.

I will not put a precise percentage on how much pipeline hides in these dark-social channels, because the honest answer is that it varies and the figures get quoted with more confidence than they deserve. What is not in doubt is the direction: a material share of B2B pipeline originates in places your software cannot attribute, and any founder optimising purely on tracked conversion data is steering with one eye shut.

The fix is almost embarrassingly simple and almost nobody does it consistently. Ask the buyer how they heard about you, at the form or on the discovery call, while the answer is fresh, not in a post-purchase survey three months later when the memory has gone. A single free-text or short-list field, 'How did you hear about us?', captures the human truth that no tag can. The answers will surprise you, and they will frequently contradict your analytics, because the buyer remembers the podcast even when your software remembers the Google click.

This is how a founder discovers that a podcast appearance invisible to every analytics tool is the real origin of a large slice of qualified leads, and then does the obvious thing: more podcasts. Without the question, that channel stays invisible, underfunded, and undervalued, while the tracked channels that merely caught the last click get all the credit and all the budget.

The goal is not to abandon software attribution, it is to stop trusting it alone. Software attribution is good at what touched the visitor on their way to converting, the mechanical last and assisting clicks. Self-reported attribution is good at what actually moved them, the influence that started the journey. Run both, and read them together. Where they agree, you have confidence. Where they disagree, the self-reported answer usually points at the real driver, and the gap between the two pictures is precisely where your most undervalued channel is hiding. That gap is one of the most useful things a lean operator can know, because it is where the next pound of marketing spend is most underpriced.