Guide
How did you hear about us survey: options, template and tips
Updated · 5 min read · Sources checked Nov 18, 2026
Short answer
A “how did you hear about us” survey asks new signups or customers one question: where they first heard of you. Ask it right after signup or purchase, offer six to ten specific options plus “Other”, and shuffle their order. Its value is catching what analytics can't see, like podcasts, word of mouth and AI answers. Read it next to your tracked first-touch source; neither one alone is the full story.
Founders call it HDYHAU, and marketers call it self-reported attribution. It's the cheapest attribution tool there is: one field on a form. It's also easy to get wrong, with vague options, a fixed order that biases answers, and results nobody compares with the data they already have. Here's how to ask it well and what to do with the answers.
Why ask when you already have analytics?
Analytics records the click that brought someone to your site. A survey records what made them go looking. Those are often different things. Someone hears you on a podcast, searches your name a week later and signs up: analytics says Organic Search or Direct, the survey says Podcast.
| Situation | Tracked attribution sees | The survey sees |
|---|---|---|
| Clicked a Google result | Organic Search | Google search |
| Heard you on a podcast, typed your URL | Direct | Podcast |
| Friend sent your link in Slack | Direct (dark social) | Friend or colleague |
| ChatGPT recommended you, searched your name | Organic Search (branded) | ChatGPT or AI assistant |
| Clicked a tagged newsletter link | Email, with the exact campaign | Newsletter, often without which one |
Tracked data is precise and complete, but it only sees clicks. Survey data sees causes, but it's partial: people forget, pick the first plausible option or skip the question. Use the survey to find channels your tracking is blind to, and tracking to size the channels it can see.
“How did you hear about us” options that work
Name specific places, not categories. “Social media” tells you nothing you can act on; “Reddit” does. A good default list for a software product:
- Google or another search engine
- ChatGPT, Claude, Perplexity or another AI assistant
- YouTube
- A podcast
- X (Twitter)
- Reddit or a forum
- A newsletter or blog post
- A friend or colleague
- Other (please tell us where)
Adjust the list to where your buyers actually spend time: an ecommerce store might add Instagram, TikTok and “Saw it in a store”; a B2B product might add “A conference or event” and “A sales email”. Keep it under about ten options so people read them all. Separate AI assistants from search engines: they're different channels that need different content, and you'll want to see each one move on its own.
Shuffle the options for each respondent and keep “Other” last. On self-administered surveys, people tend to choose items near the top of a list, which is why Pew Research Center randomizes its answer order (Pew).
Multiple choice or free text?
- Multiple choice is easy to answer and easy to count, but it only finds channels you already listed.
- Free text finds surprises (“the ProductHunt newsletter”, “your talk at a meetup”), but you have to clean up the answers by hand.
- Both: a list with an optional “Where exactly?” box when someone picks Podcast, Newsletter, Friend or Other. This gets you countable data plus names.
Early on, with a few signups a day, free text alone is fine: read every answer. Once you have more answers than you'll read, switch to multiple choice with the follow-up box.
Where and when to ask
| Placement | Pros | Cons |
|---|---|---|
| Signup form | Every new account answers while the memory is fresh. | Adds friction to the form; includes people who never pay. |
| Post-purchase page | Answers come only from paying customers. | Fewer answers; some time has passed since discovery. |
| Onboarding email or in-app prompt | No friction at signup. | Lower response rate. |
If you can only pick one, ask on signup and keep the answer next to the account. You can always filter to the ones who paid later. Make the field optional or offer a “Don't remember” choice; forcing an answer produces random clicks.
A copy-paste template
How did you first hear about us?
( ) Google or another search engine
( ) ChatGPT, Claude, Perplexity or another AI assistant
( ) YouTube
( ) A podcast → Which one? [_______]
( ) X (Twitter)
( ) LinkedIn
( ) Reddit or a forum
( ) A newsletter or blog → Which one? [_______]
( ) A friend or colleague
( ) Other → Where? [_______]Store the answer as a short, fixed value (google, ai_assistant, podcast…) next to the free-text detail. Fixed values are what make the answers countable and joinable with revenue later.
Read the answers next to tracked attribution
Put the two views side by side for the same period: the share of new customers who chose each answer, and the share of revenue your analytics credits to each channel by first-touch attribution. Then look for the gaps:
- Survey high, tracking low: a channel without clicks, like podcasts, word of mouth or AI answers. Your tracking can't credit it; the survey is your only measure.
- Tracking high, survey low: usually the last step rather than the cause, such as branded search or Direct. Check which answers those customers gave.
- Both high: a channel that both starts and closes journeys. Put more into it.
Tracked side: revenue by first-touch source
Orbit Analytics · synthetic data · last 30 days
Common mistakes
- Vague options like “Internet” or “Social media”.
- A fixed order, which inflates whatever is listed first.
- Only asking buyers when you want to know which channels bring signups that don't convert, or the reverse.
- Not normalizing free text: “chatgpt”, “Chat GPT” and “openai” are one answer.
- Treating it as precise. A survey is a sample of memories. Use it for direction and for channels you can't track, not for exact revenue splits.
See revenue per survey answer in MetricRaven
Send the answer to MetricRaven as a goal with a property, from the browser or from your backend, and every answer gets its own revenue. Open the goal under Traffic → Conversions → Goals: each property value shows the people who chose it, how many of them paid afterwards and what they paid (revenue per property value).
window.analytics.track("heard_about_us", { answer: "podcast" });- Use the queue-safe helper from Goals & funnels so the call works before the script loads, or send it from your server with server events.
- Click an answer to filter the whole dashboard to the people who gave it, then open Acquisition → Channel to see where tracking says they came from. The gap between the two is what your tracking misses.
- Compare with Traffic → Acquisition sorted by revenue, which credits each payment to the customer's first tracked visit.
FAQ
What are good “how did you hear about us” options?
Specific places your buyers spend time: Google search, AI assistants like ChatGPT, YouTube, podcasts, X, LinkedIn, Reddit, newsletters, a friend or colleague, and Other with a text box. Keep it to about ten and shuffle the order.
Should “how did you hear about us” be required?
Usually not. A required field gets answered, but some answers are random clicks to get past it. Make it optional or add a Don't remember choice, and you'll get fewer but more honest answers.
What is self-reported attribution?
Attribution based on what customers say about how they found you, usually from a how did you hear about us question, instead of tracked clicks and referrers. It catches word of mouth, podcasts and AI answers that tracking can't see, but it's less precise.
Should I ask at signup or after purchase?
Signup if you can only pick one: everyone answers while discovery is fresh, and you can filter to paying customers later. A post-purchase survey gives answers only from buyers but fewer of them.
Sources
Figures from these pages were checked on Nov 18, 2026. Charts marked “Live demo data” come from MetricRaven's public demo, which runs on synthetic data, not a real customer's.