The highest-return marketing decision at most companies gets made by the product team, silently, months before anyone books a campaign. It's the decision about whether using the product leaves a trace — whether ordinary use throws off something other people see, whether the thing gets more visible as more people around you touch it, whether anyone would bother mentioning it unprompted. Get that decision right and marketing has a tailwind it can't buy. Get it wrong and every dollar of paid acquisition is pushing a product that gives its own users no reason to spread it, which is the most expensive way to grow there is.
I want to draw a hard line between two things that both get filed under "marketing." Bolt-on marketing is spend that tells people about a product which, once they have it, does nothing to tell the next person. Built-in marketing is a product decision that manufactures distribution as a byproduct of use. The distinction matters because these two live on different sides of a growth equation, and only one of them touches the term that sets your ceiling.
The denominator nobody funds
Here is the mechanism, in numbers, because the whole argument hides without them. In the sequel-game post I wrote about distribution, the steady-state user count of a product with monthly churn c and steady external acquisition A converges to N* = A / c. That model quietly assumed something: that your only source of new users is the money and effort you spend to go get them. Built-in marketing breaks that assumption.
Suppose each active user, through normal use, produces some number of new users per period — they leave an artifact a stranger sees, they use it in front of a colleague, they mention it because using it made them look good. Call that per-user spread rate i. Now the base grows on its own: next period equals this period, plus what your existing users bring in, minus churn, plus whatever you buy.
N_next = A + (1 + i − c) · N
Solve for the fixed point and you get:
N* = A / (c − i)
Bolt-on marketing raises A. It moves the numerator, linearly: double the spend, double the ceiling, and the moment you stop paying, the term stops. Built-in marketing raises i, which shrinks the denominator — and a denominator approaching zero is where the interesting things happen.
Per-user spread i |
Denominator c − i |
Ceiling N* (A = 1,000, c = 0.05) |
|---|---|---|
| 0 (no built-in spread) | 0.050 | 20,000 |
| 0.02 | 0.030 | 33,333 |
| 0.04 | 0.010 | 100,000 |
| 0.048 | 0.002 | 500,000 |
| 0.05 | 0 | unbounded (escape velocity) |
Doubling your paid budget takes 20,000 to 40,000. A product change that lifts i from 0 to 0.04 — each user, over a period, bringing in four new users per hundred through the mere fact of use — takes the same product to 100,000 on the original budget. And as i climbs toward c, the leverage stops being linear and goes vertical. When per-user spread matches churn, the denominator hits zero and the ceiling is gone entirely. That is the regime every founder wants and almost nobody engineers, because it is engineered in the product, not in the marketing plan.
This is the precise sense in which product decisions set the ceiling on what marketing can do. Bolt-on spend can push you anywhere under A / (c − i). It cannot change the shape of the curve. Only the product decisions that move i can, and no amount of campaign budget substitutes for a positive i — with i at zero you're just buying a bigger numerator over a fixed denominator, forever, renting growth you never get to own.
Four ways use manufactures distribution
If i is the variable that matters, the practical question is what product decisions raise it. There are four mechanisms, and each is a design choice, not a marketing tactic.
Artifacts others see. Every document, design, link, invoice, or output a product generates is a surface that can carry the product to people who don't use it. The canonical case is old but still the cleanest: Hotmail's founders appended one line — "P.S. Get your free email at Hotmail" — to the bottom of every message a user sent. Each email was an ad delivered by the user to a hand-picked recipient, at zero marginal cost, with the implicit endorsement of "someone you know uses this." Calendly's booking page, Typeform's "powered by," Docusign's signature request, Loom's video that plays before you've installed anything — all the same move. The product decision is: does using this thing produce an object that a non-user encounters? If your output is a private artifact that dies inside the user's account, you've chosen i = 0 and you'll pay for it in the acquisition line every month.
Exposed by use. Some products advertise themselves simply by being used in front of other people. The white earbuds were the campaign. A laptop sticker, a delivery courier's branded box handed over on a doorstep, a checkout flow a second shopper watches over the first one's shoulder. The decision here is about the venue of use: does the product get used in private, or does normal operation happen where others can see it? You can design for this — make the visible moment the good moment, put the brand where the eyes already are — or you can ignore it and hand that surface to a competitor who didn't.
Status and identity. Some products spread because using them says something the user wants to be seen saying. This is where marketing collapses into a costly signal of the user, not the seller: displaying the product is a way for the user to signal taste, competence, membership, or being early. When that's true, users don't merely tolerate the artifact — they seek out the display, because the display does work for them. That's the strongest form of i, because the user is now spending their own reputation to spread you. And it's exactly why word of mouth is a consequence and not a channel: people talk about the product when talking about it makes them look good, or useful, or ahead. You don't buy that with referral bonuses. You earn it with a product decision about what using the thing says about the person using it.
Remarkable defaults. Seth Godin's Purple Cow made the point that "remarkable" is literal — it means worth making a remark about. A product with a default behavior surprising enough that a user tells someone unprompted has manufactured a unit of distribution for free. The extension I'd add is that remarkability has to sit in the default path, not in a feature you have to go find. A delightful setting buried three menus deep raises i for the 2% who discover it. A default that startles everyone on first use — the first time Superhuman loaded before your finger left the key, the first time a model answered better than you expected — raises i across the whole base. Put the remarkable thing on the road everyone walks, or it does no distribution work.
"We'll figure out marketing later" is a confession
When a founder says they'll figure out marketing later, they usually mean one of two very different things, and the difference is the whole essay. Sometimes they mean the product spreads through use and they'll optimize the paid layer once the loop is proven — fine, correct sequencing. Far more often, "we'll figure out marketing later" is an admission, not a plan: we built something with no built-in reason to spread, and we're hoping a campaign will supply the reason after the fact. It won't. Marketing that has to be bolted on afterward is almost always compensating for a product that came out of the oven with i = 0, and the campaign's job silently becomes forcing distribution the product declined to generate. That's a tax you pay every single period, and it's why the CAC never comes down.
I watched the clean version of this in the market where Kommerce operates — cash-on-delivery commerce in places where institutional trust is thin and courts are slow. Paid ads there are close to worthless on the trust dimension, because a scammer buys the same ad. What actually moves is the artifact of a delivery: a real courier handing a real box to someone on their doorstep, in front of the household, the package opened while a neighbor watches. That moment is a distribution event a competitor can't fake and an ad can't manufacture. The product decision that mattered wasn't the ad budget — it was making the delivery-and-confirmation flow into something visible and legible to bystanders, so that one honored transaction advertises the seller to everyone in the room. In a low-trust market the product's own spread through observed use isn't a nice-to-have on top of marketing. It is the marketing, and the campaign is the thing that would have been optional if we'd nailed it.
The move: make "does this spread through use" a product requirement
Before you fund the next campaign, run the decision through one question, and route the answer to the product team rather than the marketing team: what product decision would make this thing market itself? Concretely, two sub-questions, because they map to the four mechanisms:
Does using it produce something other people see — an artifact, a public surface, a moment of use that happens in front of a non-user? If not, can you make it? That's the i from artifacts and exposure.
Does it get more valuable, more visible, or more status-conferring as more people around the user adopt it? If not, is there a version that does? That's the i from identity and remarkable defaults.
Then treat "does this spread through use" as a first-class product requirement — a line in the spec that a feature can fail, the way it can fail a performance budget or an accessibility check. Not a growth-team problem handed downstream after launch. A design constraint present at the start, the same way distribution belongs in the product's design and not in a later phase. A feature that raises retention and a feature that raises i are both good; but the second one changes the denominator, and you should know which kind you're shipping before you ship it.
The reframe that should make a growth lead and a product lead argue: a large slice of what you're about to spend on marketing is misallocated product investment. The campaign is trying to buy, at retail, an i you could have built in wholesale. Move the money one department to the left. The best marketing decision you'll make this quarter isn't a marketing decision at all.