Resource
·
Referral
Measuring word of mouth
The short answer
Word of mouth is measurable if you ask the right question at the right time, and most "how did you hear about us" fields ask it wrong. The two questions to ask instead, when to ask them, the four numbers they give you, and what those numbers are worth.
Published
Oct 2, 2026
·
Updated
Oct 2, 2026
Why does "how did you hear about us" under-report?
Four reasons, and each has a fix that takes an hour.
It is asked at signup. The person has known you for four minutes and cannot reconstruct how they got there. Ask after they have reached the product's first value, when they can remember.
It is a dropdown of your channels. Google, Facebook, Instagram, Other. Word of mouth is either absent or buried under Other, so it is recorded as Other and ignored. Make the field open text.
It is optional. The people who answer an optional field are not the people who signed up. Make it one required question, which costs little once it sits after value rather than before it.
It asks the wrong thing. "Where did you hear about us" invites a channel name. "What made you look us up today" invites the real reason, which is often a person.
What question actually works?
Two fields, asked once, 7 to 14 days after the value event.
"What made you look us up?" Open text, required, one line.
"Was there a particular person who told you about us?" Yes or no, with an optional box for who.
The second field is the measurement. The yes rate is your word-of-mouth share, and because it is a direct closed question rather than a channel guess, people answer it accurately. The first field is the material: it tells you which moment, which phrase and which situation produced the mention, and it is the best source of landing page copy you will ever have.
Code the open text weekly against a fixed list of six to eight codes, one person, ten minutes, with the verbatims in a column beside the codes. The codes are for counting; the verbatims are for writing.
What are the four numbers?
Word-of-mouth share.
people who answered yes to the person question, divided by everyone who answered, in the same period.
Note the denominator: answerers, not all new customers. Record the response rate beside the figure every time, because a 40% share on a 20% response rate is a different claim from 40% on a 90% response rate.
Referrals per customer per period.
referred new customers in the period, divided by active customers at the start of the period.
This tells you whether word of mouth can carry growth on its own. Sustained above 1 is a loop that grows without spend, which almost no company has. The useful information is in the trend and in which segments sit higher.
Time to first referral.
the median number of days from a customer's value event to their first successful referral.
The most actionable of the four and the one nobody computes. A median of 6 days puts the ask in the first week, in the product. A median of 70 days means an in-product ask at the value moment fires too early and email is the right surface. That is how you place the ask with evidence rather than instinct, and it feeds the referral moment.
Referred cohort value index.
90-day retention of referred customers, divided by 90-day retention of non-referred customers who signed up in the same weeks. Then the same ratio for revenue per customer.
Above 1 on retention means referral is bringing better customers. Below 1 usually means a reward is doing the persuading, which is the argument in when a referral program is just a coupon. It needs the referrer stamped permanently on the referred customer's record, the field most programs forget.
How do you check the survey against something?
Self-report is biased, so triangulate with two things you can count.
Branded search. Weekly impressions and clicks for queries containing your company name, from Google Search Console. Someone told about you by a person usually searches your name, so a rising branded line next to flat paid spend is independent evidence.
Proximity matches. New signups whose source is direct or unknown, where an existing customer at the same email domain, or the same city for a consumer product, reached the value event in the previous 14 days. Count them as a series rather than attributing any single one.
If all three move together across a quarter, trust the direction. If they disagree, read the verbatims.
What are these numbers not?
They are not attribution. Never put a self-reported word-of-mouth share in a board deck as a channel next to paid: self-report over-credits whatever a person remembers last and under-credits the ad they saw and forgot, and used that way it ends up justifying a cut to a channel that was working.
Use them for the direction of travel quarter on quarter, the placement of the ask via time to first referral, and the quality comparison between referred and non-referred cohorts. Those three hold up. A precise share of revenue does not.
There is also no reliable public benchmark for word-of-mouth share, and I would be careful with anyone who offers one. Every company measures it with a different question, at a different moment, with a different denominator, so the figures are not comparable even between two businesses in the same category. Compare yourself to your own trend.
The size of the underlying effect is better established. Nielsen's survey of more than 29,000 consumers across 58 countries, fielded February and March 2013, found 84% completely or somewhat trusted recommendations from people they knew. McKinsey Quarterly in April 2010 put word of mouth as the primary factor behind 20% to 50% of purchasing decisions. Those say the channel is real. Only your own measurement says how much of it you get.
How I use it
A Series A B2B SaaS company I worked with had a "How did you hear about us?" dropdown on the signup form with five options, and the quarterly report said 4% of signups came from word of mouth. That number had been used twice to argue against doing anything about referral.
We changed one thing. The dropdown came off the signup form, and in its place an in-app question fired 10 days after an account's first completed report: one open text line, "What made you look us up?", and a yes or no on whether a person had told them about us.
The response rate was about 64%, mostly because it appeared after the product had done something useful. Of those who answered, roughly 31% said yes to the person question. The 4% had been an artifact of a dropdown and a form position.
The verbatims were more useful than the number. Four phrases came up repeatedly, all describing the same situation: somebody had seen an export in a meeting. That told us where the moment was, and two of those phrases went into the signup page headline more or less as written.
Time to first referral came out at a median of 11 days from the first completed report, which settled an argument about whether the ask belonged in onboarding or in a quarterly email. It belonged in week two, in the product.
What did not work: a third question about how likely they were to recommend us. It produced a score nobody could act on and lowered the response rate on the two questions that mattered. We removed it after a month.
Takeaways
Take the "how did you hear about us" dropdown off your signup form this week and put two questions 7 to 14 days after the value event instead.
Compute time to first referral from your existing data, and use the median to decide whether the ask belongs in the product or in an email.
Record the response rate next to every word-of-mouth share you report, and never present the share as a channel in a revenue breakdown.
Related
Up to the stage
More from this stage
Questions
When should you ask how someone heard about you?
After they have reached the product's first value, usually 7 to 14 days in, not on the signup form. People cannot answer accurately before they know what you are.
Is NPS a measure of word of mouth?
What is a good word-of-mouth share?