When I join a company I ask what the product is really about. Most people answer with features, or a vision statement. Ubersuggest's answer, once we found it, was neither. It was a growth engine for small businesses that could not afford to buy traffic and knew that search was the one place they could compete for it.
That made it the first rung of a ladder. The company behind it also ran an agency, and the agency's clients were businesses that had grown enough to pay for experts. Every customer who grew into that was worth more than any subscription. So the product's job was not to keep beginners busy. It was to make them good enough at search to want more help than software could give.
It was doing the opposite.
Beginners picked a fight with the professionals on day one.
Most of the people signing up were new to search. They chose the keywords they wanted to win, and those were almost always the ones they could not. Many did not yet rank for their own company's name.
Here is what the product did with them. It filled in keywords on the first screen without asking what the business was, who it competed with, or where its customers were; a company in Canada selling into Europe got the wrong country. It tracked those keywords, one of the most expensive things we ran, for tens of thousands of customers who would never rank for them. Then it left them to write the content on their own. By the time they had, the research was done and there was nothing to come back for. So they downgraded, or left.
We fitted a model to what trial users actually did, and it said the same thing in numbers.
A new project was a fresh start on the wrong footing. The people who came back to one project were the ones whose keywords were actually theirs. It was a boot camp with no plan and no coach.
Finding that out was the easy part.
A boot camp needs three things: a plan you can start this week, a coach who tells you what to do next, and something to show for the effort. We had none of them, and adding one would not have worked without the others.
Nothing on this list worked on its own.
A first screen that asked first. What the business is, who it competes with, where it sells. Then the value before the account: your site in, the gaps and a seven-week plan out, sign-up last. The plan was honest about the work, usually six or seven pieces of content to start, and about the wait. New users went from under ten tracked keywords to more than twenty, and they were ones they could win.
Projects stayed, as containers. In most software a new project means someone is coming back. Here it was a folder for keywords and the unit of pricing; a consultant with five clients needed five. What changed was what went into one on day one.
Reporting a consultant could hand to a client. Visual, exported as a PDF, instead of a spreadsheet.
A coach, which we called Ubersuggestions. A ranked list of the five things most likely to move a site, ordered by how much they would move it against how long they would take. Not always new content: content without links and authority sits still, so the list also said fix this page, add these alt tags, earn this backlink. Most of what rose to the top was new content on long-tail keywords, so we generated it, with Frase's templates on GPT-3 and our own OpenAI templates, in build from January 2022, about ten months before ChatGPT. Within six weeks of the beta, 79% of the people who saw it tried it and 17% saved a document, against targets of 60% and 10%. By August, 44% of its users said they would be very disappointed to lose it, over the 40% bar the industry uses for fit, and the people who tried it churned less than the people who had not.
Together they made the loop in the drawing at the top: ask first, the right keywords, the next five things, content published, rank, back next month. A habit, made as painless as we could make it. That is what a boot camp sells, and a customer who keeps the habit long enough is the one who comes looking for experts.
Over my eighteen months there: paying subscribers from roughly 51,000 to 81,000. Accounts from about 150,000 to 310,000. Lifetime plans grew until they were close to half of monthly revenue. First-time trials converted at 44%.
The model that found it did not exist when I arrived.
They had hired me for growth, and the company could not tell me why anyone was leaving. In October 2020, before I arrived, there was no number for churn, because nobody had built the thing that would produce one.
When we finally had it, it looked survivable. Fourteen per cent of paying customers chose to cancel every month, before counting failed payments. Bad, obviously. But a month is small, and fourteen is not a frightening number to look at.
Sixteen out of a hundred. Everything the company spent on getting customers was refilling a bucket with a hole that size.
Voluntary churn 13.9% to 5.4% a month. Involuntary churn halved as well, 7.2% to 3.6%. Same customers. Same product. Same spend on marketing.
The number was not new. The argument was.
Once the number existed, the job was getting it in front of the people who set the budget, as a limit on the whole business rather than a support metric, at the moment it could change what got funded.
So we built the thing that needed to exist first. Not a dashboard. An analytics function. Survival models that estimate how long a customer lasts and what shortens it. Cohort studies on people who had already left, to find out who comes back.
I did not run those models. I hired and built the team that did, and I made the case for why a growth budget was the right place to pay for them.
Fourteen per cent of everyone who cancels comes back; among people who cancelled by choice rather than a failed card, it is eighteen per cent. Two thirds of the returners take the lifetime plan.
More often than not, the problem described on the first call was not the one underneath, and finding the real one meant doing arithmetic nobody had thought worth doing.
I could have handed them the churn finding and gone home. It was a good finding. It was worth about one quarter.
What they had when I left was a business that could ask its own questions.
Survival models and win-back cohorts running as ordinary practice, by people who did not need me in the room to do it. I am trying to make myself unnecessary everywhere I go. It is the only version of the job I can defend.
When I left, the team was running those models without me.
One bet I could not get approved.
The content templates were the part of Ubersuggestions I believed in most. The output was rough, and customers loved it anyway: a small business that has never had content would rather have rough content it can edit than none. As a team we thought generative content was where the product was going, not a feature of it. We watched the first AI writers, and Grammarly's work on tone, and saw the same thing.
My case to the founders was to reinvest the profit for a period and make that bet.
They said no, and I understand why better now than I did then. The company had a proven way to grow: buy small, high-traffic tools, put the brand behind them, and scale each one into highly profitable revenue. It worked, and it paid for itself. I was not proposing a feature. I was proposing a different business model, at a moment when the market barely understood what data these models needed, or what they were about to become.
I was asking them to trade a model that worked for one that was still a promise.
By the time I left, a company built on the same API had become a unicorn on content alone. That says the bet was real. It does not say declining it was a mistake.
What I got wrong. I argued the potential. I did not price the risk in the owners' terms. I could not convince them against their own risk level, and that is a fact about me, not about them.
What I do now. When a bet changes the business model, not just the roadmap, I start with the owner's appetite for risk. Then I bring the smallest version that tests the idea, with a ceiling on what it can cost, so that saying yes is a decision rather than a leap.
A bet gets approved at the risk tolerance of whoever owns the balance sheet.