Word of mouth is probably your biggest channel. Stop hiding it in everyone else's numbers

Referral and word-of-mouth sales leave no click and no tracking number, so they land in Direct / Unknown or get credited to the branded search people used to find you. Measure them by asking at first contact and recording the answer as a source. Never spread unattributed revenue across paid channels: it hands your best customers to the wrong ads.

Ask the owner of almost any local service business where their best customers come from and the answer is the same: people who were told about us. Then look at their attribution report and the channel is missing. Google Ads is there, Facebook is there, the Business Profile is there. Word of mouth, the thing everybody agrees matters most, appears nowhere.

That is not because the report is broken. It is because a recommendation is a conversation between two people you were not part of, and no tracking system was in the room. The question is what to do about it, and the most common answer — distribute the missing revenue across the channels you can see — is the wrong one.

Why referral revenue disappears

A referred customer arrives through whatever route is most convenient — the main phone line, a walk-in, or a search for your name — and none of those routes carries the fact that somebody recommended you. Without a record on the lead, the sale either has no source at all or is credited to the route it happened to use.

Follow the four most common paths a referred customer takes, and what the report makes of each:

The first two outcomes are honest: the sale had no recorded source, and the report says so. The last two are worse, because the report confidently names a channel that only supplied the phone number. We made the same point about organic search, where branded queries collect credit for demand created elsewhere; word of mouth is the largest single source of that demand for most local businesses.

So the word "referral" in a web analytics report and the word "referral" in a business owner's mouth describe different things. When a marketing report says referral traffic is small, it is usually saying that few blogs link to you. It is saying nothing about how many customers were sent by people.

Word of mouth has no channel in any analytics tool. It appears as Direct, as organic search, or as a branded search ad, and in each case the report is describing the route the customer took, not the reason they took it.

How big the missing channel usually is

You cannot know without asking, but the evidence that it is large is consistent. Recommendations from people they know top consumers' ranking of trusted channels, and referred customers have been shown to be worth more and stay longer than customers acquired otherwise.

Put those together and the stakes are clear. The channel your report cannot see is likely to be one of your largest and to contain some of your most valuable customers. Wherever that revenue ends up in the report, it will make that line look better than it is.

Measure it with a question, a code and a column

Referral revenue is measured the low-tech way: ask every new customer how they heard about you at first contact, give referral programs a code or a referrer field, and store the answer in its own source column on the lead. Then match closed sales to those leads exactly as you would any paid source.

1. Ask at first contact, not at the sale

The best moment is when the lead is created: on the web form, in the booking flow, and as the first question a receptionist asks on a call. By the time of the sale, memories have blurred and the customer is likely to name the last thing they saw. Make the question required, and design the answers carefully:

• No default answer. A pre-selected "Google" will be left as it is by half the people who see it, and your referral share will vanish into search.
• Separate the people from the routes. "A friend, family member or neighbor" and "An existing customer" are sources. "Google" is usually a route. If you only offer routes, you will only learn routes.
• Include "Other" with a text box. The free text tells you which options to add next quarter.
• Keep it to one question. This is the answer you need, so do not make it compete with others on the same form.

For calls, the question belongs in the call script and the answer in the call log or CRM, against the caller's number. A call tracking system records the number dialed; only a person can record why they dialed it.

2. Give referral programs a code or a name

If you reward customers or partners for referrals, the reward itself is the tracking mechanism: a code to quote, a referral link, or simply the referrer's name written on the new lead. Partner referrals from other businesses — the realtor who sends you buyers, the builder who sends you a roofer's work — work the same way and are covered in more depth in our post on affiliate and partner revenue.

3. Keep it in its own source column

Record the answer in the same source or channel column your other leads use, with a value such as "Referral — customer" or "Referral — partner", so referred leads sit in the same file as everything else. A referral that lives in a notes field cannot be counted; one that lives in the source column becomes a line in the report.

If somebody answers "a friend told me" on a form they reached by clicking a Google ad, that one lead carries two claims. Decide in advance which goes in the source column — we think the recommendation, because it created the demand the ad captured — and apply the same rule everywhere. When the same person also arrives as a separate lead from another route, our post on duplicate leads sets out how to choose consistently which lead counts.

One required question at first contact, answered into the source column, turns word of mouth from an assumption into a channel. Nothing else in attribution is as cheap or does as much.

What not to do: spread the unknown across paid channels

The most common way to make Direct / Unknown disappear is to distribute it across the channels you can see, in proportion to their attributed revenue. It produces a tidy report and the wrong decisions, because the customers it redistributes are disproportionately ones no paid channel touched.

Here is what proportional spreading does to a business with $1,000,000 of closed revenue, $700,000 of it matched to a source and $300,000 left as Direct / Unknown.

The arithmetic: Google's share is $350,000 ÷ $700,000 = 50%, so it receives $150,000 of the $300,000; Meta's share is $200,000 ÷ $700,000 = 28.6%, so it receives $85,714; organic search receives the remaining $64,286. Google's return rises from $350,000 ÷ $70,000 = 5.0 to $500,000 ÷ $70,000 = 7.1, and Meta's from $200,000 ÷ $50,000 = 4.0 to $285,714 ÷ $50,000 = 5.7.

Now suppose the business adds the intake question and finds that $180,000 of the $300,000 — 60% — came from referrals. None of it was created by Google or Meta. Spreading had awarded Google $90,000 of referral revenue (50% of $180,000) and Meta $51,429, and the report would have recommended moving budget toward the channel that was already receiving the most unearned credit. That is exactly backwards. We set out the general case in unattributed revenue; referrals are its sharpest example.

Spreading the unknown in proportion to the known gives the biggest channel the biggest windfall. With referral revenue in the bucket, it hands your best customers to the ads that did not find them.

What to do with the number once you have it

A measured referral line changes three decisions: how you read your paid channels' returns, whether a referral program is worth funding, and how you explain to anybody reviewing the marketing budget why the paid channels look smaller than the platforms claim.

• Paid channels get honest returns. With referrals out of the unknown bucket and out of branded search, each paid channel's return reflects customers it plausibly found, which is the figure to budget on. It is lower and it is real, the same trade we described for blended ROAS.
• A referral program can be priced. If referred customers close at a known rate and value, you can set a reward that is a sensible fraction of what a customer from a paid channel costs, and check next quarter whether it changed the referral share.
• The remaining unknown gets smaller and more honest. What is left after referrals are recorded is walk-ins, repeat customers and contacts nobody captured. That is a much better-defined bucket to work on.

Expect the referral share to vary a lot by industry. A dentist or a roofing company in a close-knit area can see a large share of new customers arrive on recommendation; a business selling mostly to first-time, one-off buyers will see less. The point is not to hit a benchmark. It is to have your own number and stop guessing.

CloseRev never spreads Direct / Unknown across channels. If your lead file has a row for each referred lead, with its phone or email and "Referral" in the source column, it matches those leads to closed sales like any other source and reports referral revenue as its own channel; without such a row, the sale stays in Direct / Unknown, where it can be seen and counted.

Start with one question this week

The whole method starts with a single required question on your forms and in your call script, answered into a source column. Everything else — codes, partner fields, matched referral revenue — builds on that one change.

It is unglamorous work, and it is the most valuable measurement change most local businesses can make. Ad platforms will never report word of mouth, because they cannot see it and have no reason to. The only people who can measure it are the ones who answer the phone.

Word of mouth is not unmeasurable; it is unasked. Ask at first contact, record the answer as a source, and let the report show your biggest channel instead of quietly giving its revenue to the ones you pay for.