Post-purchase survey: how to ask "How did you hear about us?" and use the answers

A post-purchase survey is a short questionnaire shown to customers directly after they buy, most often on the order confirmation page. For marketing attribution it usually has one core question: "How did you hear about us?" In the ecommerce world this is often shortened to HDYHAU.

The answer is data the customer gives you on purpose. It does not depend on cookies, click IDs or consent banners, and it covers the whole journey from the customer's point of view. That makes it the natural counterpart to pixel data, which only knows what happened in a browser.

Pixels measure clicks and page views. Many buying decisions start somewhere with nothing to click.

• Podcasts, radio and TV: the listener hears your brand, searches for it days later, and Google gets the credit.
• Word of mouth: in Nielsen's 2021 Trust in Advertising study of 40,000 people in 56 countries, 88% said they trust recommendations from people they know more than any other channel (Nielsen, 2021). No pixel sees a dinner conversation.
• Video views without a click: someone watches your TikTok or Reel on their phone, then buys on a laptop a week later.
• Tracking loss in the browser: 29.5% of internet users worldwide use ad blockers (GWI Q2 2025, via Backlinko, March 2026), the average iOS App Tracking Transparency opt-in rate was 35% in Q2 2025 (Adjust), and with a legally compliant cookie banner about 60% of visits refuse consent, so roughly 40% agree (etracker Consent Benchmark 2025).
• AI assistants and communities: a recommendation from ChatGPT, a Reddit thread or a WhatsApp group leaves no referrer you can rely on.

The platforms fill these gaps with their own view. In one German account in September 2026, Meta, Google and TikTok together claimed 2.2 times the revenue the company actually made, as described in our guide to true ROAS and MER. A survey gives you an independent second opinion on where those sales came from.

Ask one neutral question, offer a single choice from 8 to 12 specific options, randomise their order and keep Other with a text field last. Every rule below fights a known bias.

• Wording: "How did you first hear about us?" The word first pushes people past the last touchpoint, which they remember best.
• Single choice for the main question. Multiple choice inflates every channel and makes shares add up to more than 100%.
• Specific names instead of categories: Instagram and Facebook as two options, not Social media. Podcast gets a follow-up asking which one.
• Randomise the order except Other. People pick options near the top of a list more often, an effect survey research has documented for decades (Krosnick and Alwin, Public Opinion Quarterly, 1987).
• Search engine is a path, not a source. If someone picks Google, a follow-up asks what made them search.
• Keep it optional. A forced answer produces random clicks from people who just want their order number.

Put it on the thank-you page. The buyer is still on your site, happy, and the decision is fresh. Everything else is a fallback.

On Shopify, surveys on the thank-you and order status pages run as apps built on checkout UI extensions (Shopify.dev). Shopify switched off checkout.liquid and additional scripts on those pages on 28 August 2025, and script tags for stores without Plus on 26 August 2026 (Shopify.dev). Check that your survey app uses the current extension model.

Expect roughly 40% to 60% on the thank-you page and 10% to 15% by email. The figures come mostly from survey vendors, so read them as their customers' averages.

If you land well below 30% on the thank-you page, the usual causes are a survey that loads too late, a long first question or a design that looks like an ad.

Use the survey to check the platforms, not to replace them. Each source answers a different question, and the useful signal is where they disagree.

• MER (total revenue ÷ total marketing spend) tells you whether marketing as a whole pays off.
• Platform ROAS ranks campaigns and ads inside one platform.
• The survey tells you which channels customers name as their source, including channels no pixel sees.

Survey-based revenue of a channel = share of survey answers for that channel × revenue from new customers

Credit ratio = revenue the platform reports ÷ survey-based revenue of that channel

An illustrative example with round numbers: a shop makes €100,000 from new customers in a month. 30% of survey answers say Instagram or Facebook, 20% say TikTok. The survey-based revenue is €30,000 for Meta and €20,000 for TikTok. Meta reports €45,000, a credit ratio of 1.5, so Meta likely over-counts. TikTok reports €8,000, a ratio of 0.4, so TikTok probably drives more than its dashboard shows, mostly through views without clicks.

Before moving budget on such a gap, check the sample size. The margin of error for a channel share is roughly:

Margin of error ≈ 1.96 × √(p × (1 − p) ÷ n)

With 400 answers and a share of 50%, that is about ±5 percentage points at 95% confidence. With 100 answers it is about ±10 points. Big budget shifts deserve a test on top, which our guide to marketing attribution models compares with other methods.

On Shopify, dedicated survey apps are the quickest start. If you want survey answers next to pixel data and ad decisions, pick a tool that combines both.

If you only need a survey on Shopify with deep reporting, Fairing or KnoCommerce is the more focused choice. Our add-on makes sense when the answers should flow into budget and campaign decisions.

• Recall bias: people remember the last touch or the most visible one. Ask how they first heard of you and treat shares as estimates.
• Fixed answer order: the options at the top get picked more often. Randomise.
• Asking returning customers: they answer "I already knew you", which dilutes the channel shares. Show the survey to first-time buyers only or filter them out.
• One Social media option: you learn nothing about which platform to fund.
• Changing the answer list mid-month: trends break. Change lists at the start of a month and note the date.
• Reading daily numbers: 20 answers a day are noise. Read weekly totals, decide monthly.
• Treating the survey as the truth: it is one of three views, next to platform data and MER.

Marketing AIgency's AI agents steer Meta ad accounts on the company number, meaning real revenue or real leads, instead of platform ROAS. The Attribution add-on includes a post-purchase survey, a first-party pixel and sending conversions back to Meta through the Conversions API. The survey answers sit next to what the platforms claim, so a suggested budget shift comes with the channels your customers actually named. Every suggestion waits in the approval queue until you decide.

Plans have fixed monthly prices with no percentage of ad spend, and the add-on can be booked on any plan. See pricing.