The post-signup survey: asking 'how did you hear about us?'
One optional question on your signup form catches the attribution your analytics can't see — and it costs nothing. Here's how to ask it so the answers are actually usable.
Analytics attribution has holes you can't patch: the customer who read your blog on their phone in March and signed up from their work laptop in June, the one who heard about you in a group chat, the one whose privacy settings ate the referral data. Self-reported attribution — one question at signup — sees through all of it, because it asks the only witness who was present for the whole journey. It's the cheapest instrument in marketing measurement, and most small businesses either skip it or run it in a way that produces mush.
Where to ask: after the commitment, not before
Put the question immediately after the conversion — on the thank-you page, the post-signup screen, or the booking confirmation. Never as a required field before the button: every field before the commitment costs conversions, and this one isn't worth that price. After the commitment, it's frictionless — the customer has decided, the form is done, and answering one optional question feels like nothing. Expect somewhere between a third and two-thirds of people to answer; even the low end builds a usable picture within a couple of months.
How to word it: open first, options second
There are two designs, and the right one depends on volume. Low volume (a few signups a week): use a plain open text box — 'How did you hear about us?' — because verbatim answers are gold at small scale, and you'll read every one. Higher volume: use options plus an 'other' text field. Keep options few and phrased the way customers think, not the way marketers report:
- Searched Google (or another search engine)
- ChatGPT or another AI assistant — worth its own option now; it's the answer analytics is worst at catching, and AI recommendations are invisible in referral data when people search your name afterward.
- Friend, colleague, or word of mouth
- Social media (one option — split it only if you actively invest in multiple platforms)
- An article or blog post
- Other (with a text box)
Randomize the option order if your tool allows it — first options get lazy picks — and resist adding options for channels you don't use. The list shapes the answers; a short honest list produces short honest data.
How to read it: against analytics, not instead of it
The survey and your analytics measure different truths. Analytics records the last technical touch; the survey records what the customer remembers mattering. The gaps between them are the insight. When analytics says 'direct' and the survey says 'Google', that's brand searches after content exposure — your SEO working invisibly. When the survey says 'a friend' and analytics says organic search, the friend probably sent a link to a post — word of mouth carried by content. Track the survey's channel split month by month next to your analytics split; movements in the survey usually confirm (or expose) what the per-post conversion data is telling you.
Analytics tells you the last click. The customer tells you the real reason. Run both and read the disagreements — that's where the insight lives.
The mistakes that ruin the data
- Making it required — you'll trade conversions for slightly more data and resent both.
- Ten options with sub-options — precision theater; respondents pick the first plausible line.
- Reading single answers as trends — at small volume, judge quarters, not weeks. Twenty responses is a hint; a hundred is a pattern.
- Treating it as the whole system — self-report complements conversion tracking; memory is honest but fuzzy, and 'Google' can mean an ad they saw or a post they read.
- Collecting it and never reading it — schedule the answers into the same monthly 15-minute review as your other five numbers.
Why this matters more every year
Attribution is degrading everywhere at once — privacy rules, cookie loss, cross-device journeys, and now AI assistants recommending businesses with no referral trace at all. Every year, more of your best channels look like 'direct' in analytics. The survey is the one instrument that improves as the technical picture blurs, because its data source is the customer's memory, not the browser's cooperation. One question, on the page every customer already sees, feeding one column in a spreadsheet — for the price of nothing, it's the best deal in measurement.