Discussion starter: a B2B and SaaS SEO question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.
- SaaS SEO
Self-reported attribution versus analytics: how are you reconciling ‘How did you hear about us?’ answers with organic and AI data?
Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.
1 reply
-
adminAdmin
Do not try to force self-reported attribution and analytics to agree, because they measure different things. Analytics records the last click, or a modelled path of clicks, that led to a session. “How did you hear about us?” records what the customer remembers as the reason they came, which is often an earlier influence: a colleague, a podcast, a community thread or an AI assistant that recommended you. The value is in the gap between them. Where customers say “AI assistant” or “a friend” and analytics says “direct” or “organic brand”, the self-reported answer is telling you where the demand came from, and analytics is telling you how it arrived.
What each source captures
Source Captures well Misses Analytics (last click or data-driven) Search clicks, paid clicks, referrals with a referrer Word of mouth, podcasts, AI answers without clicks, dark social, offline influence Self-reported attribution Memorable influences, including those with no click Precise channel detail; small touches people forget; bias towards the most recent or memorable Search Console Search demand and clicks, including branded searches Why people searched Designing the question well
- Ask at a meaningful moment, such as sign-up, demo request or first purchase, not on every form.
- Use a free-text field, optionally with a short list of options. Free text captures detail such as the specific podcast or assistant; options make analysis easier.
- Include AI assistants explicitly, and consider naming the main ones, because people increasingly cite them.
- Separate “search engine” from “AI assistant”, since users often do not distinguish them otherwise.
- Code free-text responses consistently into categories, and review the coding scheme every quarter.
Reconciling without forcing agreement
- Compare shares by category. What share of new customers say “Google” or “search”, versus the share whose first recorded session was organic search?
- Look at AI specifically. Compare the share saying “AI assistant” with the share of sign-ups from identifiable AI referrals. The gap is a rough estimate of AI influence that analytics cannot see.
- Check branded search. Customers influenced by podcasts, communities or AI often arrive through branded search. A rise in branded search alongside a rise in those self-reported answers supports the connection.
- Report both views side by side, without adding them together.
Hypothetical example:
A B2B software company finds that analytics credits most sign-ups to organic search and direct traffic, with identifiable AI referrals under 2%. Its “How did you hear about us?” field, coded monthly, shows about 12% naming an AI assistant, most of whom arrived through a branded search or by typing the URL. Rather than choosing one number, the company reports AI as a discovery channel with self-reported share and identifiable referral share side by side, and treats branded organic search partly as the capture of demand created elsewhere. Budget discussions shift towards the sources that create demand, not only the channels that capture it.
What this means for SEO reporting
Organic search, especially branded search, often captures demand that other channels created. Self-reported data helps show this honestly: SEO gets credit for capturing and converting demand, while the channels that created it, including AI assistants, PR and communities, get credit for influence. Non-brand organic search, where people discover you by searching for a problem, tends to show up in both sources, which makes it the clearest measure of SEO’s own contribution to discovery.
Watch for this:
Self-reported data can be skewed by form design: the first option in a dropdown gets chosen more often, and “Google” is a default answer for many people who cannot remember. Randomise option order where possible, keep a free-text route, and treat “Google” answers with some caution.
Sample size and timing
Self-reported data needs enough responses to be meaningful by category. A business with a few dozen sign-ups a month should look at quarterly or half-yearly totals rather than monthly swings, and should resist reading much into a category with a handful of responses. Track changes over time more than absolute shares: a rise in customers naming AI assistants over two or three quarters is a more reliable signal than any single month’s figure. Also note when the question or its options change, since adding an “AI assistant” option will itself increase the number of people who choose it. Record the date of every change to the form so the trend can be read correctly later.
Using both for decisions
For budget decisions, use analytics to judge how efficiently each channel captures demand and self-reported attribution to judge which activities create it. A channel that rarely appears in analytics but frequently in self-reported answers, such as AI assistants or a podcast programme, may deserve more investment than click data suggests. A channel that dominates analytics but rarely appears in customers’ own answers may be capturing demand rather than creating it.
Ask the question well, code the answers consistently, compare them with analytics by category rather than forcing a match, and use the gap to understand where demand starts, especially for AI assistants that analytics mostly cannot see.
Need help with this on your own site?
Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.
Add a reply Cancel reply
Replies are open to approved members. Log in or apply to join.