Your Checkout Survey Is Being Too Polite to Tell the Truth

The ‘How did you hear about us?’ box has a lovely premise: ask customers directly and they will tell you. In practice, it is often a polite fiction. Memory is fuzzy, the question arrives at a busy moment, and a lot of British shoppers would rather say a friend recommended something than admit they have been served the same Instagram ad for three weeks.

That does not make the box useless. It makes it a clue, not a confession. The fix is not a longer form or a pushier dropdown. It is a small, consent-safe routine that treats self-reported attribution as one signal among several — and reconciles it every month.

Why the answer is usually a guess

Self-reported attribution fails for three reasons. First, memory. Most people cannot reliably separate the first time they saw you from the last time. Second, politeness. A checkout is not the place for a forensic interview, so people pick the quickest acceptable option. Third, ad blindness. If your paid social campaign has been running for a fortnight, the customer may genuinely feel they ‘found you themselves’ because the ad blurred into the background.

This is why ‘a friend told me’ is not always a lie. It is often a shorthand for ‘I do not remember’ or ‘I do not want to explain.’ Treat it as a bucket to investigate, not a channel to credit.

A two-question survey that gives you something usable

Keep it to two questions and make the whole thing optional. Put it on the thank-you page or in the order confirmation email, not in the middle of payment. For service businesses, ask at booking or after the first enquiry — not during a phone call when the person is trying to get off the line.

Question one: ‘How did you first come across us?’ Leave it as a short free-text box. Do not force a dropdown. The messy answer is where the useful detail lives.

Question two: ‘Which of these played a part?’ Use multi-select with a short list: Instagram or Facebook, Google search, a friend or colleague, email, saw us in person, other, not sure. Always include ‘not sure’ and ‘prefer not to say.’ Those two options reduce the pressure to invent a tidy answer.

That is it. No star ratings, no five-point scale, no essay box. The goal is a usable response rate, not a research project.

Match it to behaviour, not to a person

Now for the consent-safe part. Do not export names and email addresses and join them to ad platform IDs just because you can. That is how a simple survey turns into a privacy headache.

Instead, match at the aggregate level. Compare survey response counts with UTM-tagged sessions, campaign flight dates and first-party analytics. If 40 people said ‘word of mouth’ in the same week your paid social spend doubled, that is worth a closer look. If ‘saw us in person’ spikes after a local event, that is a useful signal too.

If you want customer-level matching, collect explicit consent first and keep it separate from the purchase. For most small UK teams, aggregate matching is enough. It is simpler, safer and less likely to annoy people.

The monthly reconciliation ritual

Once a month, block out 30 minutes. Export the survey responses into a spreadsheet. Tag each free-text answer into buckets: paid social, search, email, PR, word of mouth, offline, unknown. Then add two columns: what analytics says happened that week, and what you think actually happened.

Look for mismatches. A cluster of ‘friend told me’ answers during a heavy paid social push is a classic. So is a wave of ‘Google search’ answers when your branded search volume barely moved. Neither proves anything on its own, but together they stop you from over-crediting the wrong channel.

Keep a running note of your assumptions. ‘We think the October Instagram push drove a lot of the word-of-mouth answers because the free-text mentions included saw it on insta and kept popping up.’ That is not hard science, but it is better than treating the checkout box as gospel.

When ‘a friend told me’ means Instagram

Here are the tell-tale signs. The word-of-mouth cluster lines up with a paid social flight. Branded search rises at the same time. The free-text answers include phrases like ‘kept seeing it’ or ‘my daughter sent me the link.’ The same discount code appears in both paid and organic orders.

None of those is a smoking gun. But if three or four line up, you can reasonably treat some of that ‘word of mouth’ as paid reach in disguise. That matters when you are deciding where the next £500 goes.

For a service business, the equivalent is ‘drove past’ or ‘a neighbour mentioned you.’ Check whether those answers cluster around a local leaflet drop or a Google Business Profile push. If they do, you have a way to separate genuine local recommendation from marketing that did its job quietly.

Keep it small and keep it honest

The point is not to build a perfect attribution model. It is to stop letting a polite checkout box make your channel mix look tidier than it is. Two questions, a monthly reconciliation habit and a healthy suspicion of ‘a friend told me’ will get you further than another dashboard nobody opens.

And if a customer does say a friend recommended you, take them at their word. Just check the calendar before you take it as the whole story.