Your Attribution Model Is Fiction. Here's What Actually Drives Revenue.
Last-click, multi-touch, data-driven — they're all wrong. The real customer journey doesn't fit in a spreadsheet.
Last-click, multi-touch, data-driven — they're all wrong. The real customer journey doesn't fit in a spreadsheet.
Every marketing team I've worked with has an attribution model. Most of them are fiction.
Not because the math is wrong — the math is usually fine. Because the underlying assumption is broken: that customer decisions happen in trackable, linear sequences.
A B2B buyer reads your blog post. Mentions it to a colleague at lunch. The colleague Googles your company name three weeks later. Clicks an ad because it's the first result. Your attribution model gives 100% credit to paid search. The blog post that started everything? Zero credit.
This isn't an edge case. A study by Forrester found that 62% of B2B buying influence happens through channels that traditional attribution can't track. Peer recommendations, industry events, Slack communities, podcast mentions — the majority of what drives revenue is invisible to your analytics.
Multi-touch attribution was supposed to solve this. Spread the credit across touchpoints. Weight by position, by recency, by engagement. The models got more sophisticated. The answers didn't get more accurate.
The fundamental problem is that most buying influence happens in places you can't track. Slack conversations. Podcast mentions. Conference hallway chats. Word of mouth. The dark funnel isn't a gap in your tracking — it's where most of the actual persuasion happens.
Even Google's own "data-driven attribution" model — which uses machine learning to assign credit — admits in its documentation that it requires a minimum of 300 conversions per month to function. Most B2B companies don't hit that threshold. So they're using sophisticated-sounding models built on statistically meaningless sample sizes.
The worst part: multi-touch attribution makes teams feel more confident in bad data. Last-click at least has the decency to be obviously wrong. Multi-touch looks reasonable, which makes it more dangerous.
Chris Walker coined the term "dark funnel" and he's right about the core insight: the channels driving your revenue are mostly the ones you can't measure.
We ran an experiment with a £4M ARR SaaS company. For three months, we added a simple free-text "How did you hear about us?" field to their demo request form. Not a dropdown. Free text.
The results demolished their attribution data. Their GA4 model said paid search drove 45% of demos. The self-reported data said 8%. Their model said organic content drove 15%. Self-reported: 35%. The single biggest source according to actual humans? "My colleague recommended you." That channel doesn't exist in any attribution model.
The dark funnel isn't dark because we haven't built the right tracking. It's dark because human decisions are messy, social, and non-linear. No amount of UTM parameters will capture the conversation that happened at a dinner party.
The best marketing teams I know have stopped trying to attribute individual conversions and started measuring system-level metrics. Brand search volume. Share of voice. Pipeline velocity. Customer-reported "how did you hear about us" responses (and yes, they actually read them).
They still run attribution models — but they treat them as directional signals, not ground truth. The model says paid search drives 40% of revenue? Great. That means paid search is probably involved in the journey somewhere. It doesn't mean cutting paid search would lose 40% of revenue.
The metrics that actually matter: branded search growth (people specifically looking for you), share of voice in your category, pipeline velocity by source, and qualitative feedback from sales on what prospects mention first. These paint a truer picture than any attribution dashboard.
Related: Content Marketing Died When ChatGPT Learned to Write. Here's What Replaced It.
Related: Google AI Overviews Killed 40% of Your Organic Traffic. You Just Haven't Noticed Yet.
Attribution models exist because executives need numbers in spreadsheets to approve budgets. That's it. The actual customer journey is a mess of offline and online signals, social proof, timing, and luck.
The sooner your marketing team accepts this, the sooner they can stop optimising for attribution and start optimising for actual growth. Stop asking "which channel gets credit?" and start asking "are we growing?" If branded search is up 40% year-over-year, your marketing is working. You don't need a model to tell you which specific blog post caused it.
The best marketing doesn't need to be attributed. It's so obviously working that nobody asks for the report.