Self-Reported Attribution

The mechanics are deliberately simple.

1. A question on the lead form, ideally open text rather than a dropdown, so the answer is the lead's memory and not your list.
2. The answer stored on the contact or deal in the CRM, next to the lead source the tracking wrote.
3. A categorisation step, manual or automated, that turns 'saw Jonas on LinkedIn' and 'linkedin post' into one value.
4. A report that groups leads, pipeline and revenue by that value.

Sales teams have asked the question on calls for decades. What changed is that the answer is now captured at form time, for every lead, and read as a channel report rather than an anecdote.

• Podcasts and video

  Someone hears the founder on a podcast, remembers the name, searches it a fortnight later. Tracking sees a brand search; the lead writes 'podcast'.

• Word of mouth and communities

  A recommendation in a Slack group, a WhatsApp message, a peer at an event. There is no click to record, so no click-based model can see it.

• Organic social

  A person reads months of LinkedIn posts without clicking one, then types the URL. The influence is real and the referrer is empty.

• AI answers

  A prospect asks ChatGPT or Perplexity for tools and gets you named. The visit that follows often arrives as direct or organic; the lead can tell you it was an AI answer, and AI search tracking can show where the citations came from.

In each case the measured source is not wrong so much as incomplete: the brand search happened, but it was the second step of a journey the tracking never saw begin.

• Recency and salience biasPeople name the last memorable thing, or the most flattering one. 'Google' from someone who saw three ads before searching; 'a friend' from someone who was retargeted for a month. The answer is a memory, not a log.
• Free text needs cleaning'LinkedIn', 'linkedin post', 'saw your ad on LI' and 'Jonas' are one channel and four strings. Without a normalisation step the report has fifty rows and no signal.
• It costs conversions and coverageA required field lowers form completion; an optional one is skipped by a share of leads. Either way the sample is not the whole pipeline.
• It has no campaign, no timestamp, no valueThe answer cannot say which ad, which keyword or which week. It cannot be sent back to Google or Meta as a conversion, so it informs strategy but not bidding.
• One touch, chosen by the leadIt is a single-touch model with the lead as the model. A journey of six touches becomes whichever one they felt like naming.

The mistake is treating the two as rivals. Measured attribution is precise and blind to what it cannot record; self-reported is complete in scope and imprecise in detail. Together they cover each other's blind spot.

• Store both on every lead: the measured journey (click IDs, UTMs, referrer, pages) and the lead's answer, as two fields that are never merged.
• Compare the distributions monthly. Where tracking says 'brand search' and leads say 'podcast', the podcast is working and brand search is collecting it.
• Treat disagreement as the finding. Agreement confirms the tracking; disagreement names the channel the tracking cannot see.
• Use measured data for optimisation and platform feedback, self-reported data for budget conversations about channels that have no click.

The self-reported attribution use case shows this layout on a lead record, and customer journey tracking is the measured half of the pair.

• Ask it once, earlyPut the question on the first form a person fills, not on every one. Later forms should inherit the answer, not overwrite it.
• Keep it open text, with a nudgeA short prompt like 'Podcast, a colleague, a LinkedIn post, a search?' raises answer quality without turning it into a dropdown of your own channels.
• Normalise into a fixed listMap the free text to ten or so categories in the CRM and keep the original string. Report on the category, audit against the string.
• Pair it with hidden fieldsThe same form should carry the measured data in hidden fields: UTM parameters, the click ID and the landing page. Self-reported answers only pay off when they sit beside those.

Self-reported attribution is a cheap, honest way to see the channels that never produce a click: podcasts, communities, organic social, a colleague's recommendation, an AI answer. It is also a memory test with a response rate, so it cannot replace measured tracking, feed a bidding algorithm or say which campaign did the work. Run both, keep them in separate fields, and read the disagreement between them as the most useful report you have. Lead attribution software that stores the answer next to the journey makes that comparison a column rather than a project.