What is self-reported attribution?
Self-reported attribution is a method of learning where leads come from by asking them. Instead of inferring the source from cookies and referrers, you add a question to the form, usually "How did you hear about us?", and record the buyer's own answer. It's often abbreviated HDYHAU.
It matters more in 2026 because so much B2B research now happens where analytics can't see: in ChatGPT and Perplexity, in podcasts and in private communities. A buyer who found you through an AI answer often arrives as Direct traffic. The form field is the cheapest way to catch that. It complements, rather than replaces, the tracking described in our hub on marketing website analytics.
This guide focuses on the form itself: the field type, the wording, where to put it and how to keep it from costing you conversions. For the strategic case, read B2B attribution when buyers research in ChatGPT.
Should the field be open text or a dropdown?
Open text gets the most honest and detailed answers; a dropdown gets the cleanest data. For most B2B sites, a hybrid works best: a short dropdown of the channels you care about, plus an "Other" option that reveals a text box, or a dropdown followed by an optional "Anything else?" field.
Field type | Answer quality | Reporting effort | Best for |
|---|---|---|---|
Open text | High: buyers name the exact podcast, prompt or person | Needs manual bucketing | Low-volume, high-value demo forms |
Dropdown only | Lower: buyers pick the nearest option, sometimes at random | Easy, already bucketed | High-volume forms where you need counts |
Dropdown + "Other" text | Good: clean counts with room for detail | Light cleaning of "Other" | Most B2B demo and contact forms |
Dropdown + optional detail field | Good, with extra context from motivated buyers | Light | Teams that want both counts and quotes |
If your form gets fewer than a hundred qualified submissions a month, start with open text. You can read every answer, and the words buyers use are often more valuable than the counts. Move to a hybrid once volume makes reading everything impractical.
How do you set up a self-reported attribution field?
Setting up a self-reported attribution field takes an afternoon: choose the forms, write the question, add the field, connect it to your CRM and agree how you'll report it. The steps below work on HubSpot, Webflow, Framer and custom Next.js forms.
Pick the forms
Add the field to demo requests, contact sales and trial signups with a sales follow-up. Skip newsletter and content download forms, where the lead's value doesn't justify the friction.
Write the question
Use "How did you hear about us?" for a simple first-source answer, or "What made you reach out today?" to capture the trigger. Pick one and keep it fixed so answers stay comparable.
Choose the options
If you use a dropdown, keep it to six to eight options, include AI tools by name, and always add "Other (please specify)". Randomize the order or list them alphabetically.
Remove any default
Start the field on a neutral placeholder like "Select one" so no answer is pre-chosen.
Place it last
Put the field after name, email and company, just before the submit button. Buyers who've filled everything else rarely abandon over one more question.
Map it to the CRM
Create a contact property, such as "Self-reported source", and map the form field to it. Keep the raw text even if you also store a bucketed version.
Agree the buckets
Decide with sales how answers will be grouped (for example AI tools, Google, referral, social, podcast or event, other) before the first monthly report.
Which dropdown options should you include?
Include the channels you actively invest in, the channels you suspect are undercounted, and a catch-all. Six to eight options is the right size: enough to separate the channels you care about, short enough that buyers read them all.
- AI assistant (ChatGPT, Perplexity, etc.): name the tools so buyers recognise the option.
- Google or other search engine: kept separate from AI so you can tell them apart.
- Colleague or friend recommended you
- LinkedIn or other social media
- Podcast, newsletter or event
- Review site or comparison article
- Other (please specify): reveals a text box.
Avoid options that mean the same thing to a buyer, like "Organic search" and "Google", or internal labels like "Outbound". Buyers don't think in your channel names. If you run a podcast or a well-known content series, it can earn its own option.
How to clean and bucket open-text answers
Once a month, export the raw answers and add a bucket column. Most answers fall into an obvious group within seconds: "ChatGPT", "asked Perplexity for agencies" and "AI" all go to AI tools. "Google" goes to search, unless the buyer adds "AI Overview", in which case note it separately.
Keep a short rules list so different people bucket the same way, and leave genuinely unclear answers as "Unclear" rather than guessing. Save the most specific answers, like a named podcast episode or a quoted prompt, in a separate tab. Those are often the most useful thing the field produces, because they tell you exactly what to repeat.
Does a self-reported field hurt form conversion?
One well-placed optional field rarely hurts conversion on a high-intent form, because buyers asking for a demo have already decided to reach out. The risk grows when the form is already long, the field is required and awkwardly placed, or the question is asked on low-intent forms where any friction loses signups.
- Keep the form short overall. If you add this field, consider removing one you don't use, like phone number or job title.
- Make it optional on long forms, required on short ones where answers matter most.
- Test before and after. Compare the form's completion rate for a few weeks either side of the change, on the same traffic.
- Watch the mobile layout. A dropdown that opens off-screen or a text box that's hard to tap does more damage than the question itself.
Form friction is usually a design problem, not a question problem. A clear form with good spacing, visible labels and a single column handles one extra field fine. A cluttered one struggles with any change.
How do you use self-reported data once you have it?
Use self-reported data to find the influences your analytics can't see and to decide where to invest. Report it monthly next to your tracked sources, follow it to closed deals in the CRM, and treat specific answers as research for content and positioning.
- Compare self-reported buckets with tracked original source for the same leads. Large gaps show where analytics undercounts.
- Follow each bucket to pipeline and closed-won in the CRM, using the property you mapped. See connecting website analytics to your CRM.
- If AI tools appear often, check what they say about you. Our guide to checking what ChatGPT says about your company covers how.
- Feed specific answers back into strategy: a named podcast is a channel to repeat; a quoted prompt is a page to write.
Self-reported data is one input to your attribution model, not a replacement for it. For how it fits with first-touch and last-touch models, read what an attribution model is.
