The ROTI framework · version 1.0
Research. Opportunities. Testing. Iteration.
An open framework for building products and companies on evidence: four phases in a loop, three gates, and one score that decides what gets your time. Published under CC BY-ND 4.0 by Aleksander Uznański and Roti Ventures.
01 · The loop
Four phases. One question each.
The loop is a cycle, not a funnel: Iteration sends you back to Research with better questions. Each phase exists to answer one question and has a condition for leaving it. If you cannot state the condition, you are not in the phase — you are guessing.
Research
The questionWho has this problem, how do they solve it today, and what does it cost them?
- Customer interviews — story-based, about past behaviour, never about the idea
- Market and competitor research; existing data before new data
- A problem statement one sentence long, with a named segment
- A written problem statement
- A segment you can reach and count
- The five riskiest assumptions, ranked
Leave whenThree people outside the team can retell the problem in their own words, and you can name who pays for it today.
Opportunities
The questionWhich gap is worth a bet, for whom, and what is the smallest thing that would prove it?
- Opportunity mapping: outcomes → opportunities → solutions, on one tree
- Sizing by evidence, not by TAM slides
- Choose one opportunity; write down what you are not doing
- An opportunity map
- One chosen opportunity with a success metric
- A list of assumptions to test, cheapest first
Leave whenOne opportunity is chosen, its success metric is a number, and the first test costs less than a week.
Testing
The questionDoes the evidence hold when money, time or reputation is on the line?
- Assumption tests before prototypes; prototypes before builds
- Paid pilots, pre-orders, letters of intent — commitments, not compliments
- One test per assumption, with a pass mark set before it runs
- Test results against the pass mark, written down
- A decision: build, change the opportunity, or stop
- A first version — only if the tests said so
Leave whenReal customers have committed something real — time, money or a signature — and you recorded the number.
Iteration
The questionWhat did the real signal say, and what do we change because of it?
- Ship, measure, talk to the people who used it and the people who left
- Keep, change or kill each part on evidence, not attachment
- Update the problem statement; it is never finished
- A learning log: what we believed, what happened, what changed
- A shorter list of assumptions than last time
- The next research question
Leave whenYou have a new question the data cannot answer. Go back to Research with it.
02 · The gates
Three gates. Numbers, not feelings.
A gate is a checkpoint with a number and a date. They are early and blunt on purpose: the most expensive thing anyone can do is keep working on something that has already said no. Pass, or stop — “almost” is a stop.
Evidence of demand
A waitlist that grows without you pushing, letters of intent, or a paid pilot.
If it failsStop, or go back to Research with a different segment. Do not build.
First revenue
Real money from real customers, invoiced or charged — not a promise.
If it failsStop, or change the opportunity. Six months without a single payment is an answer.
A product-market-fit signal
Retention and growth you can point at: cohorts that stay, customers who arrive without being chased.
If it failsSell, spin out or shut down. Keep the learning log; it is the asset.
03 · The score
Return On Time Invested.
For an operator, hours are the scarce capital, so every bet gets a score that prices them. The score is not a valuation. It is a ranking device: it tells you where your time goes.
- A stake you expect to be worth €60,000 in 36 months, needing 200 hours of your work, scores €300 per hour.
- A directory expected to earn €12,000 over the same period, needing 400 hours, scores €30 per hour.
- Your own rate is €150 per hour. The first bet beats client work; the second loses to it — sell it, automate it, or hand it on.
- Price your time at the rate the market actually pays you, not at what you wish it paid.
- Expected value is probability-weighted: a 20% chance of €300,000 is €60,000.
- Score every bet in the portfolio the same way, then rank. Review the ranking quarterly.
- The bottom of the ranking is sold, paused or handed to someone who can give it the time it deserves.
- One bet in build at a time. Two or three in Research and Opportunities, which are cheap.
04 · Using it
Same loop, three seats.
A founder with a first product
Start at Research even if you have shipped: you probably skipped it. Run the loop in twelve weeks — three per phase — and put the week-4 gate in your calendar before you start. Score the product against the job you left to build it.
A studio or a portfolio
Run one loop per bet, never one loop for the portfolio. Score all bets with the same rate and the same horizon, rank them, and let the ranking — not anyone’s enthusiasm, yours included — decide where the hours go.
A product team inside a company
Replace the roadmap review with a gate review: each initiative is in one phase, has a leave-when condition, and a date for its next gate. Ideas that cannot state a gate go back to Research; they are not on the roadmap.
05 · Principles
Five sentences the loop depends on.
- 01
Evidence before conviction
Opinions are cheap. Every bet earns the right to be built by passing a test with real customers.
- 02
Short loops, hard gates
Four weeks to show demand, six months to revenue, twelve to a fit signal. Miss a gate and the bet stops.
- 03
Time is the capital
Hours are what an operator invests. Price them, score with them, allocate them like money.
- 04
Serial build, parallel discovery
One thing in build. Several in Research and Opportunities, where a week of work is enough to kill or keep an idea.
- 05
Numbers, not feelings
Every phase, gate and score ends in a number written down before the work starts. Feelings are for the retrospective.
06 · Origins and credit
Where it comes from.
The loop was written down by Aleksander Uznański after a decade of product work — taking four products from zero to one and running product discovery and product-market-fit programmes for 300+ startups and 20+ product organisations through ProductTrio. It is the method Roti Ventures runs on every venture it builds, buys or backs.
“Return On Time Invested” is borrowed, with thanks, from the agile retrospectives canon — Esther Derby, 2003 — where it rates meetings. Here it rates bets. The acronym also happens to be the name of a labradoodle; that part is not part of the framework.
- Author
- Aleksander Uznański
aleksanderuznanski.com ↗ - Publisher
- Roti Ventures
roti.ventures - Engine
- ProductTrio
producttrio.com ↗
07 · License
Free to share. Not free to alter.
Creative Commons Attribution-NoDerivatives 4.0 International — CC BY-ND 4.0
- Copy and redistribute this framework — the text and the diagrams — in any medium or format, for any purpose, including commercial.
- Use it in your company, your workshop, your programme or your course without asking.
- Quote it, teach it, print it, translate it for your own use.
- Give credit to Aleksander Uznański and Roti Ventures, link to the license, and say if you changed anything.
- Not distribute a modified version — if you rework it, keep the rework to yourself or ask us first.
- Not add legal or technical restrictions that stop others doing what the license allows.
The ROTI Framework © 2026 by Aleksander Uznański and Roti Ventures is licensed under CC BY-ND 4.0. To view a copy of this license, visit https://creativecommons.org/licenses/by-nd/4.0/
Want to build on it — a derivative, a book, a course of your own? Write to hello@roti.ventures.