Post-purchase survey: the question to ask, how to word it, and how to read the answers next to your ad reports
Pixels count clicks and views inside a time window; buyers remember the podcast, the friend and the video they watched last week. A post-purchase survey asks them. Here is the question to ask, how to word the options, where to place it, what to store with each answer, and how to read the results against your ad reports without fooling yourself.
On this page
- What a survey sees that your pixel can't
- Step 1: ask where they first heard about you
- Step 2: write options a buyer can answer in one tap
- Step 3: place the post-purchase survey where buyers will answer it
- Step 4: read the answers next to your ad reports
- A second question for VSL funnels: what almost stopped them?
- Why a supplement brand should never ask about health
- Set it up this week
- Frequently asked questions
A post-purchase survey asks each buyer one question right after they pay, usually "Where did you first hear about us?", to catch what pixels can't: the video ad someone watched last week without clicking, a podcast mention, a friend's tip. Put it on the thank-you page, offer specific options in random order with one choice, and read the answers next to your platform reports, not instead of them.
What a survey sees that your pixel can't
Every tracking tool has a blind spot shaped by how it counts. Meta's default attribution setting credits a purchase to an ad when it happens within 7 days of a click or 1 day of a view (Ad Insights reference). Meta dropped its 28-day windows in 2021 because of Apple's iOS 14 tracking changes (Meta's April 2021 changelog). A buyer who watched your video ad three weeks ago and then typed your brand name into Google doesn't appear in Meta's report unless they saw or clicked the ad again inside the window. A podcast mention or a friend's recommendation appears in no ad report at all. Our post on the Facebook attribution window covers Meta's settings in detail.
An attribution survey has the opposite problem. It hears about any channel the buyer remembers and none they forgot, and memory has biases of its own. So use three sources side by side, each for what it does well:
Three ways to see where sales came from
What each source counts, and where it goes blind
Meta, Google, TikTokThe platform's report
- Counts: sales after a click or view of its own ads, inside its attribution window
- Misses: podcasts, word of mouth, other platforms, anything older than the window
- Leans toward: itself, because each platform counts the touches of its own ads
- Use it for: comparing campaigns and ads inside that platform
Click IDs and URL parametersYour click tracking
- Counts: sales from visitors whose ad click your page recorded
- Misses: views without a click, and journeys where the click's data was lost
- Leans toward: clicks, because a view leaves nothing on your page to record
- Use it for: checking platform reports against sales on your side
The buyer's memoryThe post-purchase survey
- Counts: whatever the buyer remembers, on any channel
- Misses: buyers who skip it, and touches they forgot
- Leans toward: the most memorable touch, and search when people looked you up
- Use it for: finding channels the other two can't see
None of the three is the truth. The platform's report and your click tracking both start from interactions they could record; only the survey hears about the podcast. Where all three point the same way, act on it. Where they disagree, the disagreement is the finding.
Step 1: ask where they first heard about you
The wording decides what the answers can tell you. Ask about the first contact, because that's what your prospecting budget pays for: the moment a stranger met your brand. The classic version, "How did you hear about us?" (HDYHAU for short), sounds the same but is looser. Without "first", some buyers answer with the last step they took, which is often the search they ran to find your site again.
Here are four wordings and what each one is for:
Four survey questions and what each one answers
The sample answers are fictional
Where did you first hear about us?
The discovery question. "First" points the buyer at the moment they met you, which is what prospecting budgets pay for.
Use this one if you ask only one question. Single choice, specific options, a free-text "Other".
How did you hear about us?
It sounds the same and reads differently. It can invite the last step the buyer took, often the search they ran to find your site again.
Avoid it for attribution. If you already ask it, switch wording on a set date and note the break in your data.
What made you decide to buy today?
Open text about motivation: the reason, the moment, the final push.
"My sister had been using it for a month."
"The video answered my question about how to take it."
For copy, not budgets. The words buyers use here are hooks and angles waiting to be tested.
What almost stopped you from buying?
Open text about objections: the price, doubt that it works, shipping time, a subscription worry.
"I wasn't sure I could cancel."
"Shipping looked slow."
For the VSL and the offer. An objection a buyer barely overcame probably stopped someone else. Ask it as an optional second question.
If you ask only one question, ask the first. The others are useful, but they answer different questions: motivation and objections feed your copy and your VSL, not your budget.
Step 2: write options a buyer can answer in one tap
Options should name what the buyer saw, in their words. People remember a video on Instagram, not a campaign type. For a supplement brand that runs video ads, the list might read:
- A video ad on Facebook or Instagram
- A video on YouTube
- A video on TikTok
- Google or another search engine
- A podcast
- A friend or family member
- An article or review site
- Other (please tell us)
Six rules keep the answers usable:
- One choice. "First" means one. Multiple choice turns the answers into a list of everything a buyer ever saw.
- Random order, "Other" last. Shuffle the options for each buyer, so no channel gains anything from sitting at the top.
- Merge what buyers can't tell apart. Someone who scrolls both apps may not know which one showed them the ad, so one option covers Facebook and Instagram.
- List your channels, not every channel. Name what you actually run, plus the unpaid ways people find you, and let "Other" catch the rest.
- Read the free text. When the same "Other" answer keeps coming up, promote it to an option.
- Change the list on a set date. Adding or renaming an option mid-month breaks the comparison with the month before, so change options on the 1st and note it.
One option deserves a follow-up. Buyers who pick "Google or another search engine" can get a second, open question: "What did you search for?" Your brand name means they met you somewhere else and searched to come back; a product or a problem means search really was the first contact. The split can change the whole reading. If 48 of 400 buyers pick search and 30 of them typed your brand name, search was the first contact for 18 buyers: 4.5% of the answers, not 12%. The other 30 met you somewhere else first, and the survey hasn't yet caught where.
Step 3: place the post-purchase survey where buyers will answer it
Ask on the order confirmation page, the moment after payment. The purchase is seconds old, the buyer is still paying attention, and the answer can be stored with the order it belongs to. Make it one tap, make it optional, and never put it between the buyer and their order details. If your funnel shows one-click upsells after checkout, put the question on the final confirmation page, after the last offer, so it never competes with an upsell for the buyer's attention.
Then close the gaps. Buyers who skip it get a one-question email a day or two later. Our post-purchase email flow shows where that fits among the emails you already send. Ask first-time buyers only, or tag each answer as a first or repeat order: a repeat buyer asked where they first heard of you is remembering a different month.
The whole routine has five stages, and the third is the easiest to skip:
The survey routine, stage by stage
Open each stage for what happens, why it matters and what goes wrong
1Ask on the thank-you page
What happens: One question, one tap and a skip link, shown right after payment.
Why it matters: The purchase is seconds old, and the answer can be saved with its order.
What goes wrong:
- A required question that blocks the order details
- Three or four questions where one would do
- The same question on every repeat order
2Email the buyers who skipped
What happens: A one-question email within a day or two, with the options as links.
Why it matters: It reaches buyers who left the page before answering.
What goes wrong:
- Sending it a week later, when the memory has faded
- Burying it inside a promotional email
- Emailing buyers who already answered
Read more: Post-Purchase Email Flow: 7 Emails for Supplements
3Save the answer with the order
What happens: Order ID, date, product, first or repeat order, the answer, and the campaign your click tracking recorded.
Why it matters: Without the order ID you can count answers, but never compare them with what the platforms and your own tracking said about the same orders.
What goes wrong:
- Answers kept in a survey tool nobody exports
- No first-or-repeat flag
- Free text nobody reads
4Code the free text monthly
What happens: Group the "Other" answers into categories, and promote frequent ones to options on a set date.
Why it matters: Your channels change: a podcast you never sponsored can start sending buyers.
What goes wrong:
- Renaming options mid-month
- An "Other" bucket bigger than most of your real options
5Read it monthly next to the reports
What happens: Each channel's share of answers against the share of orders each platform's report credits to it, month after month.
Why it matters: Large gaps that last point at channels the reports under- or over-credit.
What goes wrong:
- Reading one small month as a trend
- Treating the survey as the truth instead of a third view
Read more: Marketing Efficiency Ratio (MER): Formula, Break-Even
Saving the campaign with each answer is what makes the comparison possible: it puts what the buyer remembers next to what your tracking recorded for the same order. If your VSL runs on RevvyScales and your ads carry the rvs parameter, each sale your checkout reports back is already attributed to the campaign, ad set and ad that sent the click (ad tracking with rvs). The VSL tracking guide covers the rest of that chain, from the ad click to the checkout.
Step 4: read the answers next to your ad reports
Here is an illustrative month for a supplement brand that runs video ads on Meta, YouTube and TikTok, search ads on Google and one podcast sponsorship. It took 1,000 orders, and 400 buyers answered the survey; treat both as numbers for the arithmetic, not as a response rate to aim for. The bars set each channel's share of answers against the share of the month's orders that the platform's own report credited to it:
What buyers say against what the reports credit
Illustrative month: 400 survey answers and 1,000 orders. Blue bars are survey answers.
Read the pairs, not the bars:
- Facebook and Instagram: 41% of answers against 38% of orders in Meta's report. Close enough to trust Meta's numbers for this channel.
- YouTube: 15% of answers against 6% in the report. Buyers name it far more often than the report credits it, the mark of a channel that starts journeys other channels get the credit for.
- Search: 12% of answers against 19% in the report. A search ad can catch buyers who found you somewhere else and then looked you up, so its report collects credit for demand other channels created. The follow-up question from step 2 tells you how much.
- The podcast and friends: 9% and 11% of answers, and no report at all. Without the survey, a fifth of the people who answered would have come from nowhere.
Memory has a pull of its own: even when asked about the first time, some buyers name the last thing they saw. And the people who answer may not match the people who skip. So check two things before you move any budget.
Is the gap bigger than the noise? With 400 answers, a share carries a margin of error of about ±3.5 points at 15% and about ±5 points at 41%, at 95% confidence. So 41% against 38% is within the noise, and 15% against 6% is not. The margin is 1.96 × √(share × (1 − share) ÷ answers): for 15% of 400 answers, 1.96 × √(0.15 × 0.85 ÷ 400) ≈ 0.035, or 3.5 points. With 100 answers, every margin doubles.
Does it hold month after month? A gap that lasts three months is a finding. One that comes and goes is noise.
The survey earns its keep when it changes what you test with the budget. Set each channel's share of first mentions against its share of your ad spend. Say YouTube takes 10% of the budget and earns 15% of first mentions, while search takes 25% of the budget and earns 12%. That doesn't prove YouTube deserves more money, since search also closes sales that other channels started. It tells you which test to run: shift some budget to YouTube for two full weeks, keep everything else steady, and judge the result by total revenue against total ad spend (how to read MER), not by YouTube's own report, which undercounted it to begin with.
Now try the same rules on three new cases:
Read the survey like an analyst
Three new illustrative cases; pick an answer to see why
Show the answer
Nothing yet: with 150 answers the gap is within the margin of error. At 30% of 150 answers the margin of error is about ±7 points (1.96 × √(0.30 × 0.70 ÷ 150) ≈ 0.073), so a 6-point gap is noise. With 400 answers the margin would shrink to about ±4.5 points, and the same gap would clear it.
Show the answer
Search mostly brings back people who found you elsewhere. Searching your name means the buyer already knew you. The search ad caught demand another channel created, so the search share overstates search as a first contact.
Show the answer
Not on the survey alone: test a smaller cut and judge it by MER. First mentions measure discovery, and search ads also catch buyers who met you somewhere else, so a low share of first mentions is expected. Cut one step, hold it for two full weeks, and judge the result by total revenue against total ad spend, not by any single report.
A second question for VSL funnels: what almost stopped them?
If you add a second question, make it this one, open text and optional. Buyers who sat through a long VSL and bought anyway can tell you which doubt nearly won: the price, whether it works for someone like them, the shipping time, a fear of being locked into a subscription. Each answer is an objection the video handled just well enough for this buyer, and probably not for some of the viewers who left.
Code the answers into five or six categories once a month and count them, and keep a few word for word: a buyer's own phrasing of a doubt is the phrasing other viewers will recognize. Then line them up with the video. RevvyScales analytics include a second-by-second retention chart with the pitch marked (reading your analytics). If "I wasn't sure I could cancel" tops your list and the chart drops steeply where the subscription terms come up, you know which minute to rewrite.
Why a supplement brand should never ask about health
The survey is about marketing, so ask nothing else. For a supplement brand that rule has teeth. A question about symptoms or conditions collects health information you don't need, and anything built on it can run into the ad platforms' rules. Since September 2, 2025, Meta has been rolling out checks that flag custom audiences, lookalikes and custom conversions suggesting specific health conditions, with "arthritis" and "diabetes" as its examples, and flagged ones can't be used in new campaigns (Meta on audiences, on custom conversions). An audience built from buyers who answered "joint pain" risks exactly that. If buyers volunteer health details in a free-text answer, keep those details out of your ad tools.
Beyond that, say in your privacy policy that you ask, keep the answers with the order instead of scattering them across tools, and always offer a way to skip. This isn't legal advice; privacy rules differ by country and state, so check how they apply where you sell.
Set it up this week
Work down this list once, in order:
Set up your post-purchase survey this week
0 of 9 done
Frequently asked questions
How many answers does a post-purchase survey need?
Enough that the gaps you care about are bigger than the noise. With 400 answers, a channel named by 15% of buyers carries a margin of error of about 3.5 points either way at 95% confidence; with 100 answers, about 7. If you sell a few hundred orders a month, read the survey quarterly instead of monthly, and act only on large gaps that last.
Should a post-purchase survey be required?
No. Make it optional and one tap. A required question stands between the buyer and their order details, and someone who only wants to get past it can tap any option, which fills your data with noise. A skip link costs you some answers, but the answers you keep are the ones people meant.
What response rate should a post-purchase survey get?
There is no reliable figure to aim for. The share of buyers who answer depends on where the question sits, how it is worded, how many questions follow it and who your buyers are, so a rate quoted from another store tells you little about yours. Measure your own, keep the setup fixed so months stay comparable, and judge it by the number of answers per channel rather than by the percentage.
Can I use survey answers to build ad audiences?
Be careful. Lists built from survey answers are still customer data, so your privacy policy and the privacy law where you sell apply. For a supplement brand the bigger risk is health: since September 2, 2025, Meta has been rolling out checks that flag custom audiences, lookalikes and custom conversions suggesting specific health conditions, and flagged ones cannot be used in new campaigns. Never build an audience from answers that mention a condition.
Should subscription renewals get the survey too?
No. Ask on the first order, while the first contact is fresh. A renewal usually happens without the buyer visiting a thank-you page, and asking a subscriber about their first contact months later gives you memory, not data. Tag first orders in your records so renewals never dilute the shares you compare with ad reports.