How Did You Hear About Us? Survey Options and Setup Guide

The "How did you hear about us?" question is a self-reported attribution instrument: a form field asking each new signup to name the source that brought them to you. It captures word of mouth, communities, podcasts, and dark social that click-based tracking cannot see, and it is the cheapest demand signal a startup can collect.

Most advice about this question treats it as a polite form nicety -- a box to check so the signup page feels complete. That framing is wrong and it is why the data usually sits unused. This post treats HDYHAU as what it is: a lightweight, survey-style attribution method that complements (and sometimes corrects) pixel and UTM-based tracking. We cover why teams add it, where in the funnel it belongs, the open-text-versus-picklist tradeoff, a concrete recommended option list with exact wording, how to reduce response bias, how to reconcile self-reported answers against GA4 and your CRM, how to store and report it as a lead field, and how to let the answers drive real budget decisions. Follow the same discipline you would apply to any other attribution model for startups rather than treating it as a one-off form field.

What Is The "How Did You Hear About Us?" Question For?

The question exists to capture self-reported attribution: the source a person says led them to you. Click attribution (UTM tags, pixels, GA4 sessions) works only when a digital trail exists and is correctly tagged. A large share of early-stage B2B and consumer demand arrives through channels that leave no clean trail: a podcast mention, a Slack or Discord community, a conference hallway conversation, a friend's recommendation, or a newsletter someone screenshotted and forwarded. These are sometimes called "dark social" and they are exactly the channels where startups find their earliest adopters.

Self-reported attribution does not replace click data; it fills the hole click data cannot see. When your GA4 reports a fat "Direct" or "Unassigned" bucket -- often 30 to 60 percent of sessions depending on how strict your tagging is -- that bucket is not "no channel," it is "channel we cannot name." The HDYHAU field is your direct line to naming it. For a seed-stage team running lean, one well-placed question per signup is dramatically cheaper than a multi-touch attribution vendor.

Two cautions up front. First, self-reported data is subjective: people misremember, guess, or pick the most recognizable name. Second, it measures perception of source, not causal incrementality. Someone may say "Google" because they typed your name into Google to find you -- but the thing that put your name in their head may have been a podcast. Treat the field as directional signal, not proof.

Where Should the Question Appear in Your Funnel?

Placement determines both how many people answer and how honest the answers are. The table below contrasts the common placements on the three things that matter: response rate, data quality, and best use.

The consistent recommendation: ask at the point of commitment, not at the point of curiosity. A top-of-funnel email capture is the worst place because the visitor has not yet decided anything and will either skip it or answer randomly. A demo request form or a post-signup / thank-you step is where recall is freshest and motivation highest. For self-serve SaaS, the post-signup confirmation screen is ideal: the user just chose you and can still remember what sent them.

If you run both a demo form and a post-signup step, pick one primary capture point to avoid double counting. A common pattern is: ask on the demo form for sales-led motion, and ask on the onboarding screen for product-led motion. Keep them consistent in wording so the two datasets merge cleanly downstream.

Should You Use Open Text, a Picklist, or Both?

This is the central design decision. Each format trades structure against discovery.

Open Text

A free text box lets respondents type anything: "saw your CTO on the Latent Space pod," "my colleague Priya told me," "the Hacker News thread last Tuesday." The upside is discovery -- you learn sources you never listed. The downside is unmanageable data: thousands of variants of the same thing ("google," "Google," "googled it," "search engine"), and heavy cleanup before it is reportable. Open text is excellent for the first few hundred responses when you are still learning the channel landscape, and weak for ongoing reporting.

Fixed Picklist

A closed set of radio buttons or a dropdown gives clean, aggregatable data and near-zero cleanup. The cost is that you can only count what you thought to list; a real channel you omitted gets squeezed into "Other" or into a wrong bucket. Picklists also anchor respondents toward whatever you show, which can inflate listed options.

Hybrid (Recommended)

The robust pattern is a picklist with a meaningful "Other (please specify)" free-text escape hatch, plus an optional "If you searched, what did you search for?" follow-up. You get clean aggregation on the common channels and a discovery channel for the long tail. Route "Other" answers into a review queue monthly so emerging channels can be promoted into the picklist. This single change is what turns the field from a dead input into a living source of truth.

Whatever you choose, never make the field required in a way that blocks conversion. A required attribution question on a signup form is a conversion-killer; make it optional but prominent, and consider pre-filling nothing so you capture genuine recall rather than a default.

What Answer Options Should You Offer?

Keep the list short enough to scan and specific enough to be useful. Aim for 7 to 12 options including Other. Too few and you force bad buckets; too many and completion drops and the rare options get no signal. Order matters: put your strongest known channels near the top, but rotate or randomize across cohorts so order bias does not permanently inflate the first items (see the bias section).

Here is a concrete recommended list with wording you can copy:

• Search engine (Google, Bing, etc.)
• A podcast or video
• A newsletter or blog you already read
• A professional or industry community (Slack, Discord, Reddit, forum)
• A friend, colleague, or word of mouth
• A conference, meetup, or event
• Social media (LinkedIn, X, etc.)
• An investor, accelerator, or advisor
• A customer or partner referral
• A paid ad you saw
• Other (please tell us)

For the question itself, use plain, low-friction wording. Copy-paste examples:

• "How did you first hear about us?"
• "What led you to sign up today?"
• "Where did you come across [Company]?"
• "Who or what pointed you to us?" (useful when word-of-mouth is a hypothesized channel)

Note the deliberate separation of "Search engine" from "Social media" from "Paid ad." Collapsing them into one "Online" bucket destroys the very signal you are trying to capture. If you run podcasts or communities as a strategy, list them explicitly rather than hiding them under "Other" -- you cannot manage budget toward a channel you never named.

How Do You Avoid Biased Answers?

Self-reported attribution carries predictable biases. Name them so you can design around them.

Recency Bias

People over-credit the last touch they remember. If they heard a podcast three weeks ago, forgot it, then Googled you yesterday, they may say "Search engine" even though the podcast drove the intent. Mitigation: ask "How did you FIRST hear about us?" rather than "What brought you here today?" First-heard wording pulls the answer toward origin.

Brand-Name Capture

Big, recognizable names swallow the credit. "Google" becomes the default answer for anything vaguely digital because it is the tool people use to navigate. This is why your "Search engine" bucket will always look large -- much of it is really assisted navigation, not discovery. Counter it by adding the follow-up "What did you search for?" so you can distinguish branded search (navigational) from non-branded discovery.

Order and Anchoring Bias

The first options in a list get more clicks simply for being first. If "Search engine" is always item one, it accumulates excess share. Mitigation: randomize option order per respondent, or rotate the list order across weekly cohorts, then aggregate. Most form tools support randomization; if yours does not, rotate manually on a schedule.

Social Desirability and Guess Bias

Some respondents pick the option that sounds most legitimate, or guess to finish the form. This is why the field must stay optional and why you should weight it as directional. Pair it with a small "not sure" or "can't recall" option so you do not force a false answer -- forcing an answer manufactures bias rather than removing it.

Non-Response Bias

The people who skip the question are not random. Highly motivated buyers may answer; casual browsers skip. Track your response rate by cohort and segment, and treat low-response segments' silence as its own signal rather than assuming they match responders.

How Do You Reconcile Self-Reported Data with Click Attribution?

The goal is not to pick a winner between self-reported and click data; it is to use each where it is strong. Click attribution is precise on tagged digital paths but blind to dark social and word of mouth. Self-reported data names the blind spots but is biased and incomplete. Reconcile them in the CRM.

1. Store the HDYHAU answer as a dedicated lead or contact field (for example, hear_about_us) at capture time, not just in a form tool, so it travels with the record into your CRM.
2. Normalize the values on ingest: map free-text "Other" entries and variants into your canonical channel taxonomy so the field is queryable alongside UTM source and medium.
3. Join the record's self-reported channel against its GA4 / UTM-derived channel for the same session where both exist, and compute agreement rate by channel.
4. Build a reconciliation view that shows, per channel, the share from click attribution versus the share from self-report, and flag channels where self-report is materially higher (your dark-social candidates).
5. Use the self-reported "Other" and community/podcast buckets to size the "Direct / Unassigned" gap in GA4; treat the difference as an estimate of unattributed demand, not a precise number.
6. Report both columns side by side to stakeholders so the conversation is "where do these disagree and why," not "which one is right."

A practical rule: when self-reported and click data disagree on a channel, trust click data for precisely tracked digital paths (you can see the UTM), and trust self-reported data for the unstructured channels (communities, word of mouth, podcasts) that UTMs cannot capture. The disagreement itself is the insight -- it tells you where your tracking is blind.

How Do You Turn the Answers into Budget Decisions?

Data that sits in a field is not a decision. The point of HDYHAU is to reallocate effort toward what actually drives signups. Do it in a disciplined loop.

1. Set a reporting cadence -- monthly is right for most early teams -- and produce one table: self-reported channel, count, share of responses, and response rate by channel.
2. Separate assisted navigation (branded search) from discovery. If "Search engine" is mostly branded, do not credit SEO for demand you already created; credit the upstream channel that generated the brand recall.
3. Compare self-reported share against what you are actually spending. If communities and word of mouth are 35 percent of answers but you spend zero there, that is a reallocation opportunity, not a reason to cut the paid channel that is measurable.
4. Run small, deliberate experiments on the under-invested channels the data surfaces -- sponsor the community, appear on the podcast, host the meetup -- then watch whether self-reported share for that channel rises in following months.
5. Close the loop by feeding the winning channels back into the picklist wording and into your UTM taxonomy so the next quarter's data is cleaner and the blind spots shrink.

The discipline that matters: make one budget move per insight and measure the before-and-after in the same field. If you add podcast sponsorship and the "podcast" share goes from 4 percent to 11 percent over two months while spend there is modest, that is a repeatable channel. If it does not move, the self-report was sentiment, not signal, and you have saved budget.

How Long Until the Data Is Trustworthy?

Wait for a meaningful sample before acting -- roughly 100 to 200 responses per segment you care about, which for most startups is a few months of signups, not a week. Below that, single responses swing shares wildly. Trend over time beats any single month.

Should Sales Reps Ever Ask It Verbally?

Yes for high-value B2B deals where the form is skipped. A rep asking "What made you look into us?" on a discovery call captures the same self-reported signal and often richer detail. Log it in the same CRM field so it merges with form data; just label the source as "verbal" so you can weight it.

For the broader survey program, including timing, question examples, and Shopify and B2B setup, see our post-purchase survey guide.

Frequently Asked Questions

What Is The "How Did You Hear About Us" Question?

It is a self-reported attribution field on a signup, demo, or purchase form that asks a new user to name the source that led them to your company. Unlike click attribution, which relies on tags and pixels, it captures word of mouth, communities, podcasts, and other dark-social channels that leave no digital trail. Used as a lightweight survey instrument, it helps startups name demand their analytics tools cannot see, especially the large Direct or Unassigned buckets in GA4.

Where Should I Place the Attribution Question in My Funnel?

Place it at the point of commitment rather than curiosity: on a demo or contact form, a post-signup onboarding screen, or a post-purchase thank-you page. Avoid the top-of-funnel email capture, where visitors have not formed an impression and will skip or answer randomly. Ask at the moment recall is freshest and motivation is highest, keep it optional to protect conversion, and use one primary capture point if you run multiple surfaces to prevent double counting.

Should the Options Be Open Text or a Fixed Picklist?

Use a hybrid: a fixed picklist of 7 to 12 specific channels with a meaningful "Other (please specify)" free-text escape and an optional search-term follow-up. The picklist gives clean, aggregatable reporting; the open "Other" preserves discovery of channels you did not list. Route "Other" answers into a monthly review so emerging channels can be promoted into the list. Never make the field required in a way that blocks signup, or you will trade conversion for noisy data.

How Do I Reconcile Self-Reported Answers with Click Attribution?

Store the answer as a CRM lead field at capture, normalize free text into your channel taxonomy, then join it against the record's UTM or GA4 channel and compute agreement by channel. Trust click data for precisely tagged digital paths and trust self-reported data for unstructured channels like communities and podcasts. Report both columns side by side and treat their disagreement as the signal that reveals where your tracking is blind, rather than declaring one source correct.

Key Takeaways

• HDYHAU is a self-reported attribution instrument that names dark-social and word-of-mouth demand click tracking cannot see, not a decorative form field.
• Place it at the point of commitment (demo form, post-signup, thank-you page), never on top-of-funnel email capture where recall and motivation are lowest.
• Use a hybrid picklist of 7 to 12 specific options plus an "Other" free-text escape, and keep the field optional to protect conversion rate.
• Counter recency, brand-name, order, and non-response bias with "first heard" wording, a search-term follow-up, option randomization, and response-rate tracking.
• Reconcile the field against UTM and GA4 in your CRM, trusting click data on tagged paths and self-report on unstructured channels, and treat disagreement as blindness signal.
• Turn answers into budget only after a few hundred responses per segment, by reallocating toward under-invested channels the data surfaces and measuring the before-and-after in the same field.