Is asking customers how they heard about us good attribution data?
It is useful and unreliable at the same time. People misremember, default to whatever is easiest to say, and often name the thing that reminded them rather than the thing that found them. Self-reported source works best as a check on tracked data, especially for offline influence that tracking cannot see, not as a replacement for it. Open questions produce better answers than a drop-down list.
Why the answers skew, predictably
Human recall of marketing exposure is weak and biased in known directions. Recent touches crowd out earlier ones, so people name the search they just did rather than the recommendation three weeks ago. Socially comfortable answers get over-reported, which is why referral is claimed more often than referrals occur. And people name channels they have a word for, which is why online is a common answer and remarketing display never is.
None of that makes the data worthless. It makes it a different measurement, with its own bias profile, which is useful precisely because it is biased differently from your tracking.
The intake process distorts it further
When a customer service rep is asking during a busy call, the field gets filled with whatever is fastest. A drop-down list defaulted to the first option collects a lot of the first option. A required field with no good match collects a lot of other.
Two changes fix most of this. Make the field optional but ask the question conversationally, and capture the answer as text rather than a category. Then classify the text afterward. Where calls are recorded, call analysis can extract stated source from the conversation itself, which removes the pressure on the rep entirely and produces more faithful wording than a picklist ever will.
Use it as a discrepancy detector
The right way to read self-reported data is against your tracked data. Put the two distributions side by side and look for structural gaps rather than agreement.
If ten percent of customers say they saw your truck or a yard sign and your tracked data has no category for it, you have found influence your attribution silently reassigned to search. If far more people claim referral than your tracked referral sources show, you are probably understating word of mouth. If self-report says radio and your tracking shows nothing, that is the expected result, not an error. Storing the stated source on the customer record alongside the tracked source, as customer intelligence work does, lets you run that comparison on demand instead of rebuilding it each quarter.
Where it is the only tool you have
- Offline media with no click path. Radio, mail, sponsorships, vehicle wraps, community presence.
- Word of mouth. No system will ever observe this directly.
- Long-delayed influence. A mailer kept on the fridge for eight months outlives every cookie and every lookback window.
- Sanity checks on a new channel. If nobody mentions a channel you just launched, that is worth knowing early.
Topics: self-reported attribution · surveys · call intake · offline influence
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.