Find what is worth building. Build it with agents.
Your analytics tool pulls a cohort in seconds. It does not separate the 5 to 7 groups that truly drive the product, find the one screen, in the one session, that decides who stays, or hand you a ranked plan you own. We did all three on a clean sample of 16,531 users of our own app, where the top 22 percent drive 61 percent of the trips. Then we ship the work with an agent-run delivery system we run on that same app every day.
Most teams optimize for a user who does not exist.
The average user is an average of opposites.
On our own app, the average hid two completely different people: a commuter who opened it twenty times a week, and a trial user who opened it once and never came back. Tuning the product for their midpoint helped neither of them.
Three questions most analytics setups cannot answer on their own:
- Which users actually drive the product, and what do they have in common?
- Where exactly do the rest drop off, and in which session?
- Of the users you could keep, which ones are worth the effort?
And one that lands on a different desk: once you know the answer, can your team ship against it before the answer goes stale?
| Dimension | The average way | The FAMANEO way |
|---|---|---|
| Optimizes for | the average user | the 22 percent who drive the product |
| Finds the drop-off | guesses | the exact screen and session |
| You leave with | a slide | SQL you own |
What we found on our own app
We did not learn this from a client. We learned it on our own product.
TARANSIT is our own public-transit app. For this analysis we took a clean sample of 16,531 users, with 160,000 trip searches between them. We had plenty of analytics and no idea which of them mattered.
Our first attempt at segmenting them was wrong. Forty-four percent of users fell into one meaningless bucket. It took ten passes over the data in BigQuery before the groups actually separated by behavior.
What came out was seven behavior groups. The top three, 22 percent of the sample, generate 61 percent of all trips and stay at 80 to 93 percent retention after a month. The largest group, one-time trial users, is 37 percent of the sample and retains at 18 percent. The gap between the most and least committed users at day 30 is 74.5 points.
Then we found the moment that decides it. 63.2 percent of trial users never reach the connection screen in their first session. The ones who do reach it in their second session come back the next day 95.4 percent of the time. That single screen, in that single session, is the difference between a retention strategy and a guess.
We left with a ranked list of seven experiments, and we knew exactly which screen, which session, and which users each one was for.
| Segment | Users | Share of users | Day-30 retention |
|---|---|---|---|
| Committed Commuter | 566 | 3.4% | |
| Light Commuter | 1,035 | 6.3% | |
| Power Explorer | 2,011 | 12.2% | |
| Weekend Traveler | 754 | 4.6% | |
| Engaged User | 1,290 | 7.8% | |
| Casual User | 4,743 | 28.7% | |
| Trial User | 6,132 | 37.1% |
Why not just ask your analytics tool
Your analytics tool can answer a lot of this on its own now. A prompt gives you a cohort in seconds. So here is the honest version: if one query solves your problem, you do not need us.
We are for the part the tools cannot do. The segments that took ten passes to separate, not one prompt. The single screen in the single session that decides whether a user stays. And the tracking underneath all of it, which has to be right before any answer, human or machine, is worth trusting.
Product Growth
We run the same thing on your product.
Three ways to start, depending on where you are.
Where you lose users.
The one screen and session where most new users disappear. About one week.
Learn more Product GrowthWhich users drive your product.
The 5 to 7 groups that drive your product as SQL you own, the activation gaps, and a ranked experiment backlog. About three to five weeks.
Learn more Product GrowthGrowth, run monthly.
We keep your segments current and your backlog ranked, every month.
Learn moreYou keep everything we build.
The FAMANEO growth system: know which users drive the product, keep them, then charge for the value. Each step is a service.
And once you know what drives the product, there are two further things worth doing with it. We can run the operating system that acts on the findings: strategy and focus, the KPIs and the feature-engagement metric tree, continuous discovery, the experiment loop, as a fixed-scope audit, a fractional lead, or your interim Head of Product. We set this up for a client whose first-day keep-rate then rose 16.7 points and whose day-30 keep-rate rose 20.3 points.
And we can build the pricing on it, because when you know which users get the most from your product, you can charge in proportion to it.
Then it has to get built.
This is where most of this work dies. The analysis is right, the backlog is ranked, and the team is already three sprints behind. A quarter later the segments have moved and the plan is a document nobody opens.
Most teams buy the analysis from one firm and the build from another, then spend a quarter translating between them. We think the team that found what is worth building should be able to build it. That is the whole reason both of these come from one place.
So the second half of what we do is the delivery side: how an engineering team ships with AI agents reliably, and how you find out which parts of your pipeline agents can actually take.
Product Growth
The half you have just read. It runs on your event data and it leaves the model with you.
- Starts with the session where you lose people.
- Runs through the 5 to 7 groups that drive the product.
- Lands on a ranked backlog and the pricing built on it.
Ship With Agents
The half below. It runs on your delivery pipeline and it leaves the system with you.
It is proven on our own product, not yet on a published client result, and we say so before we say anything else.
Ship With Agents
A delivery system, not a pile of prompts.
First, the honest part.
This line is proven on our own product, not yet on a published client result. We would rather say that than oversell it. The client outcome numbers are not in yet, and we will not pretend otherwise.
What we can show you instead is the system itself. We run a 15-agent, ontology-grounded delivery system on TARANSIT: an orchestrator, an architect, a planner and a developer passing artifacts between stages, a verify loop capped at three revise rounds, 16 enforced entity types, and an evaluation set with a minimum of 20 ground-truth rows behind it. It runs daily, on the same product the numbers above came from, and we will open any part of it you want to see.
The agent delivery pipeline in five stages: a product lead agent writes the spec, a system architect designs, an implementation plan is drawn, a developer builds, and a verify loop of test plus code review checks the result with up to 3 revise rounds.
Four ways to work with it.
Where agents actually pay off.
Which parts of your delivery pipeline agents can reliably take, ranked against your own repository and workflow data. About one to two weeks.
Learn moreShip one workflow with agents.
One real delivery workflow running with agents in your environment, with the evaluation harness that keeps it honest. About four to eight weeks.
Learn moreTeach your team to ship with agents.
Your engineers ship with agents using an auditable method and an evaluation discipline they keep.
Learn moreKeep your agents honest.
Monthly evaluation against your own ground-truth set, so drift surfaces before your users find it.
Learn more
Who we are
FAMANEO is Martin, a product lead who builds and has shipped across MarTech (Bloomreach), navigation (Sygic), and freight (Eurowag), and now runs TARANSIT as a live growth lab.
The two halves above are not two businesses. They are two things one person actually does: the product lead who has moved real growth numbers, and the builder who ships. That is why the growth analysis arrives as SQL you can run rather than as a deck, and it is why the delivery method is one we use rather than one we read about. The reason the proof is exact is that it is our own product and our own SQL.
Book a 30-minute call.
Tell us which half you are here for, and we will tell you whether we are the right people for it. No pitch deck.
or email hello@famaneo.com