How to Run a How Did You Hear About Us Survey
Ask any marketing team where their last hundred customers came from and you will get a confident answer. Look closer and that answer is built on data that quietly leaves out a third of the story. Your ad platforms report what their pixels were allowed to see. They never saw the podcast someone heard on a commute, the friend who texted a link, or the TikTok a customer watched three weeks before they ever clicked.
The oldest and lowest-tech tool in marketing recovers more of that missing third than any modeling algorithm: you ask the customer directly. A "How did you hear about us?" survey — HDYHAU for short — puts the question to the one person who actually knows the answer.
Done well, it produces attribution you can move budget on. Done carelessly, it produces a tidy-looking chart that lies to you. Here is the difference.
What a HDYHAU survey actually measures
A pixel measures touchpoints it was permitted to observe. A HDYHAU survey measures what the customer remembers as the reason they showed up. Those are not the same question, and the gap between them is exactly where your hardest-to-track channels live — word of mouth, dark social, podcasts, influencers, and offline.
That gap is not random noise — it is concentrated in your highest-intent, relationship-driven channels, the ones a pixel structurally cannot see. A modeling algorithm can interpolate the missing conversions, but it can never tell you a customer bought because a friend recommended you. Only the customer can.
A pixel tells you where the click happened. The customer tells you where the decision started.
It is also a durable signal. When Apple's App Tracking Transparency arrived with iOS 14.5 in April 2021, the vast majority of users declined to be tracked — Flurry measured opt-in at around 11% worldwide and just 4% in the US in the weeks after launch — and ad platforms lost line-of-sight into a large share of conversions overnight. We unpack that shift in measuring channels without cookies. A survey answer survives all of it. No cookie, no pixel, no platform permission required. The customer volunteers the information.
The response rates you should expect
Self-reported data is only useful at volume, so the first question is whether anyone answers. Placement decides almost everything.
Placement decides almost everything. A single question on the order-confirmation page — while the customer is still on your site, still feeling the glow of a purchase, with the answer one tap away — gets a dramatically higher response than the same question emailed two days later, where it competes with a crowded inbox and a faded memory. On-site post-purchase surveys routinely clear response rates that emailed surveys never approach; we break the placement-by-placement numbers down in post-purchase survey benchmarks.
The takeaway: ask on the thank-you page, not in a follow-up email. Recall is sharpest and motivation highest in the seconds after checkout.
Designing the question so the answer means something
One question. Ask "How did you first hear about us?" — the word first matters, because you are trying to capture the start of the journey, the awareness moment your pixel never recorded, not the last click it already takes credit for.
A few rules protect the data:
- Offer 6–10 options that match channels you actually run. Listing channels you do not invest in just invites noise.
- Always include an "Other" field with free text. It is where you discover the creator or podcast you did not know was working for you.
- Keep it to a single screen. Every extra question costs you completion.
- Do not order options by your own bias. Rotate them if you can, so position does not manufacture a winner.
A worked example (hypothetical)
Picture a DTC coffee brand spending most of its budget on Meta. The pixel reports Meta as the clear top performer and paid search a close second. Then the HDYHAU survey comes back: a third of new customers answer "a friend or family member," and another sizable slice name a single podcast the brand sponsors for a modest flat fee.
Neither of those channels shows up in the ad dashboards at all — word of mouth has no pixel, and the podcast listener heard the spot, searched the brand name, and converted on what the platform logged as "branded search." The survey did not contradict the pixel so much as reveal the channels feeding it. The realistic move is not to slash Meta; it is to fund the referral motion and the podcast that were quietly doing the top-of-funnel work for free.
Reading the data without fooling yourself
Self-reported attribution has predictable biases. Name them so they do not mislead you.
Branded-search contamination. A customer who discovered you on TikTok, then googled your name to buy, will often answer "Google." The pixel agrees — last click was search — so both sources confidently miscredit the channel that closed for the channel that introduced. This is precisely why the "first" framing matters, and why you read HDYHAU as an awareness signal, not a conversion one.
Recency and salience skew. People over-report whatever is top of mind. A heavy TV week inflates "TV" answers slightly. Treat single-week spikes with suspicion and trust trends that hold over weeks.
Small-sample slicing. It is tempting to filter responses down to one product, one region, one week — until each bucket holds twenty answers and noise swamps the signal. Stay at channel level until volume earns finer cuts.
Self-reported data is not precise. It is unbiased about the channels your pixel is blind to — and that is the half of the map you were missing.
Where HDYHAU fits next to your pixel
The mistake is treating the survey as a replacement for platform analytics. It is not. It is a second, independent instrument, and the value is in reconciling the two.
Use your pixel for what it is good at: in-channel optimization, creative testing, real-time bid and budget moves. Use HDYHAU for the question the pixel cannot answer — which channels create demand in the first place. When the two agree on a channel's importance, act with confidence. When they disagree on a top-of-funnel channel — and for word-of-mouth-driven brands they almost always do — the survey is the more trustworthy witness, because it is the only instrument that was ever allowed to see that channel.
That reconciliation — survey answers, pixel data, and UTMs read together instead of in isolation — is the difference between knowing where your customers come from and guessing. It is exactly what we built Rauxdata to do.