Conversion Has Levels, and the Second Is Just as Hard

Say the first part worked. You built the growth engine, you fed it for a few months, and traffic showed up. Some weeks it doubles. You have signups now, a number you can screenshot. And revenue looks exactly the way it looked in June.

This is the second problem, and it’s a different problem. Traffic is a supply question. What happens to that traffic is a series of much smaller, much harder questions, and most founders are answering them with a single number that hides everything interesting.

Start with the number you probably don’t have

Almost every founder can tell me how many signups they got last month. Far fewer can tell me what share of the people from any given source signed up.

That distinction is the whole game. A site-wide conversion rate of 2% is an average across channels that behave nothing alike. Underneath it, one source converts at 6% and another at 0.2%, and the only decision you actually face, which is where to spend next month, depends entirely on knowing which is which.

An aggregate rate is a number you can report. A rate by source is a number you can act on.

That last part is the piece people skip, and it’s a schema change more than a marketing task. One column on the user record. It’s the difference between “we got 40 signups from the newsletter” and “the newsletter produced 40 signups and $0, while a community thread produced 6 signups and four of them are paying.”

There are at least two conversions

The word “conversion” gets used as if it names one event. It names a category, and there are at least two members.

Level one: stranger to user. Somebody who had never heard of you signs up, starts a trial, subscribes to the newsletter, or follows you. They spent attention. They didn’t spend money.

Level two: user to customer. Somebody who is already using the thing decides to pay for it.

Here’s the part that surprises people: level two is as hard as level one, and being good at one tells you almost nothing about the other. The link between them runs from weak to nonexistent.

They’re different acts by a different person in a different frame of mind. Level one asks somebody to spend two minutes on curiosity, at zero risk, on a decision nobody will ever ask them to justify. Level two asks for money, which means a decision that has to hold up. In a company, it means a decision they may have to defend to somebody else. Everything that made level one easy is absent.

Piling up free users while nobody upgrades is not partial success. It’s information, and the information is fairly specific: the thing you built is interesting, and the thing you charge for isn’t essential.

I wrote about the version of this that stings most in why your SaaS has zero users. People sign up, poke around, and live on the free tier forever. The instinct is to blame the pricing page or the paywall placement. Usually the paid features are solving the parts of the problem that people have already learned to tolerate, and somebody who has lived inside that problem knows which parts those are.

Either way, the free-user pile is a reading. Treat it as one, rather than as progress toward something.

Count the small conversions. Then follow them to money.

Newsletter signups count. A Substack subscribe counts. A follow counts. They’re real conversions with real value, they’re the earliest sign that anything is working, and at low volume they may be the only signal you have.

Stopping there is what turns them into decoration. If you count signups and follows and never connect them to a revenue event, you can’t tell which channel is doing anything. You end up with four channels producing numbers, all of them going up and to the right, and no idea which one is worth another six months.

So track through. Signup, first real use of the thing, first payment, second month of paying. For each one, keep the source attached. It’s tedious for exactly one afternoon, and then it answers questions forever.

The first time this pays off is the day a channel you were proud of turns out to produce nothing but free accounts, and you get to stop.

Why your A/B test told you nothing

Here’s the arithmetic that ends most early conversion optimization, and it’s worth doing on your own numbers before you spend a week on a headline.

Say you get 600 visitors a month and 2% of them sign up. That’s 12 signups. You change the headline, and next month you get 15. That’s a 25% improvement, and it’s also three people.

The rule of thumb I use: you want somewhere around 100 conversions per variant before a difference of a few percentage points is worth believing. At 12 a month, that’s eight months per variant. You do not have eight months per headline, so at this volume you cannot A/B test your way anywhere, and running the test anyway mostly produces confident conclusions from noise.

What works instead, at low volume:

• Change big things, not small ones. An effect large enough to see through the noise is the only kind you can detect. Rewriting the offer beats rewriting the button.
• Run in phases. Pick a change, run it for a defined window, and don’t touch anything else during it. Six weeks is usually the floor.
• Write down what you expected before you start. Otherwise you’ll find a story in whatever the numbers did.
• Fix the funnel by watching people, not by testing. At 600 visitors a month, five recorded sessions or three phone calls with real users will teach you more than any split test.

The tension with running every channel

The first post in this pair says to run all the channels at once, and I stand by it. It’s also how you end up scatter-shot, firing industrial-grade AI slop into every surface that will accept it, on the theory that volume is a strategy.

AI has produced more content, marketing and promotion than there are humans available to read it. That gap is going to widen. Anyone can now generate 30 posts in an afternoon, which means 30 posts in an afternoon signals nothing about you and costs the reader something to sort through.

Quality matters and so does quantity. It’s both, and holding both is harder than picking one. Publish enough to be findable and to give yourself a reading, and make each piece worth the reader’s time on its own. Content marketing works when you have something to say and fails when you’re producing volume to satisfy a schedule.

Measurement is what keeps the quantity honest. A channel filled with slop shows up in the data as a channel that produces signups and never a dollar. Without the revenue tracking, that channel looks like your best one for a very long time, because slop is extremely good at generating the shallow half of the funnel.

Phases, not dashboards

Early on, the work is gathering data, trying things deliberately, and running each thing long enough to reach a sample size that means something. That’s slower than the dashboard culture suggests, and it’s the actual job.

A phase looks like this: one change, one hypothesis, one window, and one number you agreed to watch. At the end, you write a sentence about what happened. Ten of those sentences is a real understanding of your own business, which is worth more than a dashboard nobody reads and considerably more than an intuition you’ve stopped questioning.

Free users are not a leading indicator of paying users. They’re a leading indicator of interest, which is a real thing to have and a different thing to have.