- A one-question survey after checkout ("How did you hear about us?") captures channels that click-based tracking misses: word of mouth, podcasts, creators and video people watched without clicking.
- Design matters more than the tool: specific answer options, randomized order, a write-in "Other", and a follow-up for detail. Analyze new customers separately from repeat buyers.
- Treat the answers as one signal next to platform data, GA4 and incrementality tests. Use them to question budget splits and choose what to test, not to replace attribution.
Why add a survey when you already have attribution data?
Every ad platform reports conversions it can link to its own ads, using its own rules. Add up what Meta, Google and TikTok each claim, and the total often exceeds the orders you actually received. Meanwhile, channels that rarely produce a trackable click, such as podcasts, creators, word of mouth and video, get little or no credit.
A post-purchase survey asks the customer directly. It is self-reported data, so it has biases of its own, but they are different biases from click tracking. That is what makes it useful: when two imperfect sources disagree, the disagreement shows you where to look.
The answers are also zero-party data: information the customer chose to give you. You can use it for reporting and, with care, for segmentation.
How do the attribution sources compare?
No single source tells the full story. Each one sees part of the customer journey and misses another.
| Source | What it sees well | What it misses or overstates |
|---|---|---|
| Ad platform reporting | Clicks and views on its own ads | Overlaps with other platforms; each claims credit by its own rules |
| GA4 | Sessions, referrers, UTM-tagged visits on your site | Views without clicks, offline influence, cross-device gaps |
| Post-purchase survey | What the customer remembers as the trigger, including offline and no-click channels | Recall bias, first-versus-last confusion, non-response |
| Incrementality testing | Whether a channel causes extra sales | Covers one channel and one period per test |
| Marketing mix modeling | Long-term channel contribution from spend and sales data | Needs history and volume; slow to update |
The survey is cheap, fast and continuous. That makes it the best early warning that your attribution model is missing something, and a good input for deciding which channel to test next.
How do you write a “How did you hear about us?” question?
Keep it to one main question, with an optional follow-up.
The main question
Use plain wording: “How did you first hear about us?” Including “first” points the customer toward discovery, which is what platform tracking captures worst. Without it, many people answer with the last thing they did, such as searching your brand name.
The answer options
- Be specific. “Social media” tells you nothing. List Instagram, Facebook, TikTok, YouTube and Pinterest separately if you advertise or post there.
- Separate paid from organic only where customers can tell. Most people cannot say whether a post was an ad. “Instagram” is more reliable than “Instagram ad”.
- Include the offline and no-click channels. Friend or family, podcast, creator or influencer, TV or streaming, in a store, press article.
- Include “Google or other search”. Expect it to collect people who heard of you elsewhere and then searched. A follow-up helps.
- Randomize the order, keeping “Other” last. People favor options near the top of a list.
- Add a write-in “Other”. Review the write-ins monthly and promote frequent answers into the list.
- Keep the list short. Eight to twelve options is a practical range. Remove channels you no longer use.
The follow-up
Show a second, optional question based on the answer: “Which podcast?” or “Which creator?” or “What did you search for?” The detail makes podcast and creator spend measurable, and search answers that name your brand point back to some other channel.
Where should the survey appear?
The best placement is the order confirmation (thank-you) page, while the purchase is fresh and the customer is still on your site.
- Shopify and Shopify Plus. Shopify has moved the thank-you and order status pages to checkout extensibility and retired the old additional scripts on those pages. Use a survey app built as an app block for those pages, and check that answers are saved to the order (as a note, tag or metafield) and to the customer profile.
- WooCommerce and Magento. Survey plugins or extensions can add the question to the order received or success page. Confirm the answer is stored with the order ID.
- Email fallback. Send the same question in the order confirmation or first post-purchase flow email for customers who skipped it. Mark these answers as email responses, since recall is weaker a few days later. Our post-purchase email flow guide covers the sequence.
Wherever the answer is stored, it needs the order ID, the date, and whether this was the customer’s first order. Without those three fields, the analysis below is not possible.
Who should you ask?
Ask every customer, but analyze first-time buyers separately. A repeat customer’s answer describes how they found you months or years ago, which tells you little about this month’s acquisition spend. For acquisition analysis, filter to first orders only.
You can also skip the question for customers who already answered on an earlier order. That keeps the experience short and avoids the same customer counting twice.
How do you analyze survey answers?
Run the analysis monthly for new customers only.
- Calculate the share of answers by channel. Exclude non-responses from the shares, but record the response rate. If it falls, the results become less reliable.
- Scale shares to total new customers. If 30% of respondents chose Instagram and you had 1,000 new customers, the survey points to about 300 Instagram-driven customers. This assumes non-respondents behave like respondents, which you should state in any report.
- Compare with platform-reported new customer conversions. Large gaps in either direction are the findings.
- Calculate survey-based CAC. Channel spend divided by survey-implied new customers. Compare it with the platform-reported customer acquisition cost.
- Watch trends, not single months. A channel whose survey share rises for three months after a spend increase is a stronger signal than a one-month spike.
You can also send answers to your email platform as a profile property and to GA4 as an event parameter. Keep the raw answers in one table so reporting stays consistent.
Worked example: comparing survey and platform CAC (illustrative numbers)
All numbers below are illustrative.
A store acquires 1,000 new customers in a month. 600 answer the survey. The store scales the answer shares to all 1,000 new customers and compares them with what each platform reports.
| Channel | Spend | Platform-reported new customers | Survey-implied new customers | Platform CAC | Survey CAC |
|---|---|---|---|---|---|
| Meta (Instagram and Facebook) | $30,000 | 420 | 300 | $71 | $100 |
| Google Ads | $15,000 | 380 | 250 | $39 | $60 |
| Podcast sponsorships | $6,000 | 30 (discount code) | 150 | $200 | $40 |
| TikTok | $5,000 | 40 | 100 | $125 | $50 |
| Friend or family | n/a | n/a | 200 | n/a | n/a |
What the store should conclude:
- Platforms overlap. Meta and Google alone claim 800 of 1,000 new customers, before any other channel. The survey suggests both are over-credited.
- Podcast and TikTok look under-credited. Discount codes and clicks miss many customers who heard an ad and bought later.
- Google answers need care. Some of the 250 “Google” customers probably searched the brand after hearing of it elsewhere. The follow-up question (“What did you search for?”) shows how many typed the brand name.
What the store should not do is move budget from Meta to podcasts based on one month of survey data. The right next step is an incrementality test on the podcast and TikTok spend, with the survey as the reason to run it. Our guide to incrementality testing explains how to set one up.
What biases should you watch for?
- Recall bias. Customers remember vivid, recent or personal touchpoints (a friend, a creator) better than repeated banner ads.
- Last-touch confusion. Despite the word “first”, many will answer with the step right before buying.
- Option order and wording. Randomize, and keep wording stable. Changing the list resets your trend lines.
- Non-response. If certain customers skip the question more often, such as mobile users, the shares tilt.
- Brand-search inflation. “Google” absorbs demand created elsewhere.
Read the survey as a direction and a ratio, not an exact count.
Post-purchase survey setup checklist
- [ ] The question reads “How did you first hear about us?” and is the only main question.
- [ ] Answer options are specific, eight to twelve in number, and randomized with “Other” last.
- [ ] A write-in “Other” is reviewed monthly.
- [ ] Follow-up questions capture the podcast, creator or search term.
- [ ] The survey runs on the thank-you page using a method supported by your platform.
- [ ] An email fallback reaches customers who skipped it, and those answers are flagged.
- [ ] Each answer is stored with order ID, date and first-order flag.
- [ ] Customers who already answered are not asked again.
- [ ] Monthly analysis uses first-time customers only and records the response rate.
- [ ] Survey CAC is compared with platform CAC by channel.
- [ ] Large gaps lead to an incrementality test before budget moves.
- [ ] The privacy policy covers how survey answers are used. Confirm with counsel.
FAQ
Can a post-purchase survey replace multi-touch attribution?
No. It answers a different question: what the customer remembers as the reason they found you. Multi-touch attribution and platform data describe tracked interactions. Use the survey to spot where tracked data is blind, and use incrementality tests to settle the disagreements.
What response rate is good enough?
There is no universal threshold. What matters is that responses are numerous enough each month to be stable and that the rate does not drop sharply. If only a small share of customers answer, check placement and mobile display before trusting the shares.
Should the survey be required to complete the order?
No. Keep it after the purchase is complete and make it optional to submit. Adding questions before payment adds friction to checkout and can cost orders, which defeats the point.