TL;DR
Meta's reporting tells you what Meta can see, which is Meta's view of the world. A one-question post-purchase survey, "How did you hear about us?", tells you what the buyer remembers, which is often a different story. Run it on your thank-you page or in your post-purchase email, keep it to one question with a short list of answers, and compare the results with your platform reporting every month. The gaps are where your budget is being misread. The survey will not replace your analytics. It will keep them honest.
The problem with asking the platform
Meta has a structural conflict of interest. It sells the ads, delivers the ads, measures the ads, and reports on the ads. That does not make its numbers fake. It makes them partial. Attribution windows, modelled conversions, and view-through credit all lean in the direction of the platform claiming the sale, because that is what the system is built to do.
The result is familiar to anyone who has run ads for a while. Meta claims more conversions than your store recorded. The gap is not always fraud or error. Often it is a customer who saw your ad, then Googled you, then came back through an email, then bought. Meta saw the ad view and claimed the sale. Your analytics saw the email click and claimed it too. The customer, asked directly, will usually say something simpler, like "my friend recommended you" or "I saw you on Instagram". That answer is not perfect either. It is a different lens, and two lenses beat one.
How to run the survey
Keep it brutally simple. One question, on the thank-you page right after purchase, when the memory is fresh: "How did you hear about us?" Offer five to seven answers, no more: Instagram ad, Facebook ad, Google search, friend or family, influencer or creator, email, other. Add an optional free-text box for the people who want to tell you more, and do not make the question mandatory. A forced question gets random answers from people who just want their order confirmation.
If you sell by lead form rather than checkout, put the same question in the confirmation step or the first follow-up email. Response rates will be lower, and that is fine. You need a few hundred answers a month for the pattern to mean anything, not a perfect census.
Resist the temptation to ask more. A five-question survey gets abandoned. A one-question survey gets answered. The extra questions are for your curiosity, and your curiosity is not the point. The point is a clean, comparable monthly read on where buyers think they came from.
What to do with the answers
Compare, monthly, the survey share against the platform share. Three patterns show up again and again.
Meta claims more than buyers remember. If Meta takes credit for 60 percent of sales but only 30 percent of buyers name a Meta ad, part of that gap is view-through and modelled credit, and part of it is people forgetting. The truth sits somewhere between the two numbers, and the gap is a reason to be careful about scaling on Meta's numbers alone.
A channel Meta cannot see is doing real work. Word of mouth, creators, and offline mentions never appear in Ads Manager. If 20 percent of buyers say a friend recommended you, that is a channel worth feeding, and it is invisible to every dashboard you own.
The answers disagree with your targeting. If your ads target 25 to 34 year olds but buyers keep saying they found you through a creator whose audience is 45 plus, your creative is reaching someone your targeting did not intend. That is not a failure. It is information about who actually buys.
None of this means you should cut Meta's budget because the survey says so. The survey measures memory, not causality. A buyer who saw five of your ads and then Googled you will often say "Google". That does not mean the ads did nothing. It means last-touch memory is not the whole story either. Use the survey as a check on the platform, not as a replacement for it.
The mistakes that ruin it
Asking too many questions is the first. Leading answers is the second: do not list "Instagram ad" first in bold with a sparkle emoji and bury everything else. Randomise the order if your tool allows it. And the third is reading a week of data as a verdict. Survey answers are lumpy. A creator shout-out on Tuesday will spike one answer for a week. Read monthly, look at trends, and ignore single-week swings.
One more: do not survey only your happiest buyers. If the question only appears in a post-purchase email that goes to subscribers, you are sampling subscribers. Put it on the thank-you page where every buyer sees it, and the sample stays honest.
FAQ
What is a post-purchase survey?
A one-question survey shown to buyers right after purchase, usually "How did you hear about us?", with a short list of channels to choose from. It captures what the buyer remembers, which is a different view from platform attribution.
Why do I need a survey if Meta already reports conversions?
Because Meta reports its own view, and its view leans toward claiming the sale. The survey is an independent check. It will not replace analytics, but it catches the cases where the platform takes credit for demand it did not create.
How many responses do I need for the survey to mean something?
A few hundred a month is a reasonable floor. Below that, the shares swing wildly from week to week and you will read noise as signal. Read monthly, not daily.
Where should the survey appear?
On the thank-you page right after purchase, when the memory is fresh and every buyer sees it. For lead generation, use the confirmation step or the first follow-up email.
What if buyers misremember where they found me?
They will, sometimes. Memory is not perfect, and last-touch recall overstates search and understates ads. That is why the survey is a check on the platform, not a replacement. The gap between the two numbers is the useful part.
Should I cut Meta spend if the survey credits it with fewer sales?
No, not on survey data alone. The survey measures what buyers remember, not what caused the purchase. Use it to question scaling decisions and to spot channels the platform cannot see, not to make budget cuts by itself.
MetaMaxd is a Meta ads agency. We trust Meta's numbers and verify them, because a platform that grades its own homework deserves an independent examiner. If your attribution needs a second opinion, that is what we do.


