Word of Mouth

Start counting where users come from and one unexplained block is always left over. Subtract the people who arrived from ads, from search, and from invite links, and some remain. Ask how they heard about it at signup and the answer is that a colleague was using it, or a friend said it was good.

That block went unmanaged for a long time, on the grounds that something uncountable is also unmanageable. So growth meetings ran on ad spend and invite volume, because those were the only things anyone could touch.

The problem is that the block is usually the largest one. Half the users who arrived from an ad are gone by the end of the first month, while the ones who arrived because a colleague was using it stay. Ignore the biggest and best-retaining source and keep raising ad spend, and total user count goes up while the share who stay goes down.

Naming it word of mouth moved that block into the category of things you manage. You can't manufacture it, but you can find what's blocking it — and what blocks it is nearly always something you shipped.

Two conditions produce it, and neither of them is a reward: whether there's an experience worth recommending, and whether there's a place to recommend from.

The first is a product question. People recommend something in roughly three situations.

1. When it beat expectations. The gap between what they anticipated and what they got is what turns into speech. A merely fine product doesn't get recommended; one that is conspicuously better at some single point does, which is why broad competence tends to spread worse than one sharp advantage.
2. When they just watched someone they know have the problem. Seeing a colleague struggle to find a document produces the name on the spot. This case is created by the situation rather than by the product, which is why products used in shared workplaces spread faster than equally good ones used alone.
3. When recommending serves them. In a product with a network effect, where value grows as more people use it, recommending isn't a favor — it's self-interest.

The second is a question of surface. Willingness to recommend stops if there's nowhere to put it. Whether a shared document opens without the recipient creating an account, whether one link is enough. It's common for new signups to rise after a single no-login share link ships, and that isn't more people being persuaded — it's the path clearing for people who already wanted to recommend.

Counting it is worth knowing about. It can't be measured exactly, so three proxies get used. The how-did-you-hear field at signup gives a rough split but depends on people bothering to answer. New users unexplained by ads, search, and invites gives a residual figure that also absorbs every tracking gap you have. And the number of times a second team inside an existing customer signs up on its own gives the narrowest count but the cleanest one. The last is the most useful. When a neighboring team at the same company arrives unprompted, a human sentence happened in between, and no campaign is taking credit for it.

There are three places the concept enters a decision.

1. Splitting acquisition sources. Retention for ad-acquired users and word-of-mouth users gets read separately. When the two differ sharply, the decision about ad spend changes, because buying more users and buying more retained users stop being the same purchase.
2. Locating a product problem. High satisfaction with no word of mouth means the surface is blocked. When net promoter score (NPS) — the satisfaction score built from asking about willingness to recommend — comes in high while unexplained signups stay flat, look at the sharing flow.
3. Judging an early product. If you're still inside the cold start problem, where value doesn't exist because users don't, then having almost no word of mouth is normal. Pushing users in with ads at that stage won't make them stay, and the absence of word of mouth is how you confirm it before spending the budget rather than after.

It often runs alongside the paid version of the same idea, but the order matters. Attach a referral program to a product with no experience worth recommending and you acquire people who take the reward and leave. Check first whether you have users who reached advocacy, the stage where a customer recommends you unprompted.

"Forty percent of new signups aren't explained by ads or invites."

An estimate of the word-of-mouth share.

"Satisfaction is high but nothing is spreading. I think we need to look at the sharing flow."

Suspecting the surface rather than the product.

"This product isn't ready for a reward yet. A reason to recommend comes first."

Walking the order back.

At Pagedeck, Mia counts second-team signups inside an existing customer every quarter. Last quarter that was thirty-four, and twenty-six of them were other departments at the same company. That isn't a number ads can produce.

There was one change that moved it sharply: Noa reworked the share link so recipients could read a document without an account. Nothing was done to persuade more people — the distance shrank for people who were already recommending. Since then the team treats word of mouth as a product problem.

A common belief is that word of mouth is left to luck. It's human speech, so it looks like nothing you can touch. But what's blocking it is usually on your side: the share link demanding an account, a free tier so narrow that the person who recommended it looks foolish, no screen worth showing when they do recommend it. Removing those one at a time is as concrete as making an ad, and it usually costs less.

The other belief is that rewards increase word of mouth. Put money behind a recommendation and more of them should happen. But a rewarded recommendation changes character. The recipient reads it as advertising, and the sender knows they're sending it partly for themselves. It's common for part of the unpaid flow to move into the paid flow without the total rising at all. Rewards are better understood as something that accelerates people who are already recommending than as something that increases how many do.