The tools aren’t missing.
The discipline is.
αlphaβeta is a thinking partner for empirical work. Talk through what you're trying to do; it pushes back, maps the problems, and holds you to what you said before the data existed.
Open the app See how it worksMost teams don't lose to bad statistics. They lose to moving goalposts.
Wins declared after the fact
A test comes back mushy. Someone finds the segment that moved. Six months later nobody remembers what "success" was supposed to mean.
Ideas that never had a problem
Features get built because someone wanted them. Nobody wrote down what they were supposed to fix — so nothing can prove they didn't.
Zombie projects
A loss with no agreed meaning doesn't die. It mutates into "the next version." The cost is every test slot it keeps occupying.
Every experimentation platform tells you whether you can run a test. None of them ask whether you should — or remember what you promised.
One loop, five moments.
Talk
Dump what you're thinking, in plain language. No forms.
Map
It builds the picture as you talk: goals, problems, open questions, ideas — and asks what each idea is actually for.
Commit
Before a test or launch: why you believe it, how you'll know, and the result that makes you walk away. Written down, dated, before results exist.
Resolve
Outcome lands next to the prediction. The bucket is computed from what you committed — not from the mood in the room.
Calibrate
Over time: how good your judgment actually is, and where it's reliably wrong.
Goals flow down. Evidence folds back up.
Everything you discuss becomes a node on a living map: goals at the top, the problems blocking them, the questions you can answer with a lookup instead of a test, the solutions competing to solve each problem, and the bets you've committed to run.
Answer a question and it folds into the problem it informed. Lose a bet and the branch it was gating visibly prunes. Kill an idea before testing it — the map remembers that as a win, because it was one.
[visual: the canvas — altitude cascade, folded questions, a pruned branch]
Seven changes, one test, no way to learn.
Before a year-end campaign, a nonprofit shipped a full donation-form redesign — seven simultaneous changes bundled into one A/B test. Revenue dropped 69.6%. The post-mortem: "test individual elements separately to identify which changes caused the decline."
The test ran fine. The bet was malformed: seven mechanisms in one wager means no result — win or lose — can say which change did what. That's knowable before a single visitor arrives.
αlphaβeta's admission step asks the questions that were skippable here: what problem does each change solve, what's the mechanism, and can this test actually resolve it? Sometimes the honest answer is a smaller test. Sometimes it's no test at all — which is the cheapest win there is.
Source: NextAfter research library, Buckner International donation-form experiment.
Built for the person who cares. Happens to protect them when a colleague doesn't.
It listens instead of demanding forms
Every tool like this has died of homework. Here the writing-down is done for you by a partner that pushes back while you think. You get sharper thinking now; the record is a free side effect.
Honesty is rewarded, not policed
Changed the plan mid-way? Recording it counts in your favor — the way aviation turned confession into a safety system. Nothing blocks you. Everything is remembered.
It judges the "should," not the "can"
It asks what an idea is for, whether the evidence supports it, and whether the test can even resolve — and cheerfully kills bad ideas before they cost six weeks of traffic.
Yours
Runs privately. No account. Your data never leaves your machine unless you choose — and it works alongside whatever platform you already test on.
People who run tests and would like to be held to them.
- Domain experts with a stats-shaped gap — researchers, designers, marketers who own the decision and want the method handled honestly.
- Practitioners who know the process and want the commitment loop without the ceremony.
- Small teams pushing back on decisions that arrive without reasons.
Not for: exploratory or creative work where the goal is discovered in the making. Pre-commitment is the wrong tool there, and alphaBeta says so.
- Nudge, never gate. Everything is recorded; nothing is blocked.
- Commitments are dated and tamper-evident — but the lock is a property of the record, not a wall in your way.
- Discipline stats are private by default and never a leaderboard.
- Platform-agnostic. Method-agnostic. A/B tests are one instrument, not the product.
In development, built in the open.
αlphaβeta is being designed through live use — every session shapes the next version. If you run experiments and recognize the problems above, we'd like to hear how they show up where you work.
Open the app Follow the work