Word of mouth promotion: it is a distribution problem, so build the sentence, the moment and the carrier
You cannot buy the message, and you can build every part around it. The four pieces of a working word of mouth programme, the line the FTC rule draws on incentives since October 2024, and the three numbers that tell you it is happening.
Published by AI Knows Us · Updated 22 October 2026
Word of mouth gets treated as a reward for being good, which is why most businesses have no plan for it. It is better understood as a distribution problem. A message has to exist in a form somebody can repeat, it has to be asked for at a moment when the customer is willing, it has to travel without you in the room, and any reward attached to it has to stay on the right side of both the law and the platforms. Four buildable pieces. The only part you genuinely cannot control is whether the product deserves it, and that part is settled before any of this starts.
This page sets out the four pieces, the rules on paying for recommendations in the United States, the shape of a referral mechanic that does not get gamed, and how to measure something that happens in conversations you never see.
The four pieces
| Piece | The job | The usual failure |
|---|---|---|
| The sentence | One thing a customer can say in their own words, in under ten seconds, without getting it wrong | A positioning statement nobody can repeat, so the customer invents a vaguer version |
| The moment | Asking when the customer feels the benefit, not when you need the lead | Asking at the invoice, which is the one point where the relationship is about money |
| The carrier | Something that survives being forwarded: a link, a code, a photograph, a searchable name | A recommendation with nowhere to land, so the lead arrives as "someone mentioned you" and dies there |
| The reward | Making it easy to do the favour, without buying the opinion | Paying for sentiment, which is now prohibited outright in the United States |
The sentence has to be repeatable, not accurate
Ask three existing customers what you do and write down the words they use. Whatever is common to all three answers is your real message, and it will be shorter and blunter than anything on your website. That version is what travels, so make it true and then make it easy.
Three things make a sentence repeatable. A number, because numbers survive retelling: next day, twelve years, two people. A surprise, because a mild contradiction is what makes somebody want to tell it: the accountant who answers the phone, the builder who quotes a fixed price. And a name, because a recommendation without a spellable name cannot be acted on. If your business name is easy to mishear, the sentence has to carry something else searchable with it.
The moment is a decision, not a mood
Pick one or two points in the customer journey and ask there every time, by default, rather than when somebody remembers. The good moments share a feature: the customer has just experienced the benefit and owes you nothing.
- Right after the thing worked. The delivery arrived, the leak stopped, the report landed. Hours, not weeks.
- After the second purchase, not the first. A repeat buyer has already decided, and their recommendation is more confident.
- When the customer volunteers praise. An unprompted thank you is the single highest yielding moment there is, and the reply that captures it is one sentence: thank you, and would you tell one person who needs this.
- Never alongside a bill, a renewal or a complaint resolution. The first two make it a transaction and the third makes it an apology.
Write the ask into the process that already happens at that moment: the completion email, the handover sheet, the closing line a staff member says. An ask that depends on somebody feeling like it happens a tenth as often.
The carrier is the part almost everyone skips
A recommendation has to survive the gap between the conversation and the action, which is usually hours and often days. Give it something to carry.
- A short link or a code that works when typed by hand, and that you can count later.
- Something physical where the conversation is physical. A card in the bag, a sticker on the box, a magnet on the boiler. It is not about the design, it is about surviving the walk to the car.
- A screenshot friendly artefact. People forward pictures far more readily than links. A clear one page price list, a before and after, a receipt with your name legible on it.
- A findable name. Most recommendations end with the receiver looking you up, so the first thing they find has to confirm the recommendation. If the search result is thin, the carrier breaks at the last step.
Rewards, and the line the law now draws
In the United States the ground moved on 21 October 2024, when the Federal Trade Commission's rule on the use of consumer reviews and testimonials, 16 CFR Part 465, took effect. Three parts of it govern anybody running a word of mouth programme.
- Conditional incentives are out. You may not give compensation or any other incentive that is contingent on the review being positive, or on it expressing a particular sentiment. A disclosure does not cure it. Asking for a review and rewarding the act of reviewing is a different thing from paying for a good one, and the contingency is what the rule turns on.
- Fabricated reviews and testimonials are out, including writing your own, buying them, and negative ones aimed at a competitor.
- Suppressing reviews is out. Using groundless legal threats, intimidation or false accusations to get a negative review removed is covered by the rule, and penalties are set per violation and adjusted for inflation.
Separately, the FTC's endorsement guides require a material connection between an endorser and a brand to be disclosed clearly: payment, free product, employment, a family relationship or an affiliate arrangement all count. Platforms then add their own rules on top, and they are stricter than the law in places: Yelp asks businesses not to solicit reviews at all, and Google allows asking while prohibiting incentives, so a programme built around one platform's rules can breach another's. Check the current policy of each place you intend to send people, which for most local businesses starts with the Yelp business page and the Google profile.
The safe design, and the one that works better anyway: reward the referrer for a referral that becomes a customer, reward nobody for an opinion, and keep reviews as a separate unincentivised ask.
A referral mechanic that holds up
- Decide one sided or two sided. One sided pays the referrer. Two sided gives the new customer something as well, which raises conversion and gives the referrer a reason to pass it on that is not about their own benefit. Two sided is usually the better opening offer.
- Pay on an event you can verify. First purchase, completed job, retained account at thirty days. Paying on a signup is how a programme gets farmed.
- Set the reward against the real acquisition cost. If a customer from paid media costs you a hundred dollars, a fifty dollar referral reward is cheap. The arithmetic behind that kind of reward, and the traps in it, works the same way as loyalty card marketing.
- Cap it and date it. A maximum number of rewards per referrer per quarter, and an expiry on the code. Both exist to stop one person turning the programme into a job.
- Exclude the obvious self dealing. Same household, same payment card, same company. Write it into the terms rather than arguing it later.
- Make claiming it dignified. Nobody wants to chase you for twenty dollars. Pay automatically and tell both sides it happened.
Three numbers that tell you it is working
- Asked at source. One question on every enquiry form and every first call: how did you hear about us, as an open text field or a short list that includes a person's name. It is the only direct measurement of word of mouth that exists, and it stops working the moment somebody turns it into a required dropdown with five marketing channels in it.
- Code redemption rate. Codes issued against codes used, by cohort. A low rate with high issuance means the carrier is wrong, not the offer.
- Direct and branded demand. People told about you look you up by name. A rising count of searches for your business name, and of direct visits, is the clearest trace that conversations are happening, and it moves on a lag of weeks.
Set a baseline for all three before you change anything, because the first month of a referral programme always looks good: your most enthusiastic customers go first, so read the rate again once that early group has had its turn.
Word of mouth promotion: the questions that come up
Is word of mouth the same as a referral programme?
A referral programme is one instrument for producing word of mouth, the one with a code attached. Word of mouth includes everything that happens without a code: the group chat, the trade contact, the neighbour. The programme is worth running because it is measurable, and most of the volume will still come from the part you cannot count.
Does it work for business to business?
It is usually the strongest channel there, because the buyer's risk is professional. The mechanics change: the carrier is an introduction rather than a code, the reward is often reciprocity rather than money, and the ask goes to a person whose own reputation is on the line, so ask only where you are confident of the work.
How is this different from influencer marketing?
The relationship. An influencer is paid for reach and the payment must be disclosed. Word of mouth comes from a customer with no material connection to you, and the moment you create one, disclosure rules apply to it too.
What about a refer a friend discount, is that an incentive the rule covers?
Paying for a referral that becomes a customer is a sales arrangement, not a review. The rule bites when the incentive is tied to someone publishing an opinion, and particularly to publishing a favourable one. Keep the referral reward and the review request in separate messages and the distinction stays clean.
When can we judge whether it is working?
Count referrals by source from the first week, knowing the earliest come from customers who already wanted to help. Judge the steady rate once the ask is embedded in a process rather than a campaign; where it settles depends on your offer and your customers. The slow part is the sentence: it takes a few weeks of listening to customers to find the one they already use.
What kills a word of mouth programme fastest?
Asking everybody, all the time, with no moment behind it. The ask is a small withdrawal from goodwill, and spending it on customers who are not yet pleased leaves you with neither the referral nor the goodwill.
The recommendation nobody in the conversation makes
Word of mouth has quietly acquired a second form. A buyer who would once have asked a colleague now asks an assistant which company to use, and gets back two or three names with reasons attached. It behaves exactly like a recommendation, carries the same authority, and is assembled from whatever can be retrieved about each business: reviews, listings, its own pages and third party write ups. That makes it the one recommendation channel a business can inspect directly rather than infer from an enquiry form. Asking the question a customer would ask, and reading who gets named, is a short exercise and it tends to be uncomfortable the first time. AI Knows Us runs those questions across the main assistants and reports which businesses come back and what the answers were built from.