Why Bootstrapped Teams Should Run No More Than Three Experiments at Once

A small team is slowed not only by work volume, but by open contexts, pending results and unfinished decisions. Limiting the portfolio to three active experiments reduces switching and increases the chance of reaching a conclusion.

Three is not a universal law. It is a simple constraint preventing abundant ideas and AI tools from creating an infinite queue.

An experiment is not a task

An experiment has a hypothesis, change or observation, metric, duration, owner, success condition, stop condition and final decision.

“Redesign the page” is a task. “Changing the promise increases qualified visitors who start the diagnostic” is an experiment.

The hidden cost of work in progress

Every additional experiment requires tracking, collection, interpretation and coordination. Results arrive at different times. People forget why work started and launch a new idea before closing the old one.

AI reduces starting cost, not the cognitive cost of finishing.

Allocate the three slots

• Value: is the customer outcome real and important?
• Distribution: can the team reach the audience and obtain commitment?
• Reliability or economics: can the outcome be delivered with acceptable quality, cost and trust?

This mix prevents the team from testing only interface variations.

The parking lot is not a fourth active column

Capture ideas with date, source and reason. Do not assign an owner or time until a slot is free. Remove ideas that no longer matter during monthly review.

Entry conditions

An experiment enters when user and problem are precise, evidence or risk justifies it, the metric can change a decision, the test is bounded, cost fits and another experiment has finished, merged or stopped.

Exit cleanly

Closing means recording outcome, gap, decision and learning. “Paused” without a condition becomes invisible debt.

Use four exits: continue, modify, stop or integrate into stable practice.

Example

Yumea tests the moat diagnostic as a magnet, a manual referral channel and the trust checklist on a sensitive workflow. A calculator idea remains parked.

After two weeks, the checklist creates conversations but the channel cannot find its audience. The channel stops and the calculator can enter with a revised hypothesis.

FAQ

Three experiments for the entire company?

For a small team sharing the same decision-makers, yes as a starting point. Independent teams may set separate limits when dependencies are controlled.

What about a long experiment?

Define decision milestones and a maximum duration. If no information arrives, revise the metric or stop.

Does a successful experiment leave the portfolio?

Yes. It becomes a decision, new experiment or stable practice. It does not remain active by habit.

Next step

List all experimental work and keep only three active cards with decision dates. Use the Three Experiments Board. For a review, email contact@yumea.fr with “THREE” in the subject.