Two to four weeks of real work together before committing anything. The cheapest insurance in the whole company.
A time-boxed project — typically two to four weeks — where you work together on something real, with a defined deliverable, before any equity is agreed.
It has a goal, split responsibilities, agreed hours, an agreed way of communicating, a deliverable and a scheduled review at the end.
Because you cannot learn how someone works from conversations. You learn it from a deadline.
Four weeks costs almost nothing. A wrong co-founder costs the company. This is the highest return on time available anywhere in the founding process.
Goal: validate the scheduling problem and produce a clickable prototype.
Responsibilities: A runs twelve customer interviews; B builds the prototype from what A learns.
Hours: 15/week each, agreed in advance.
Communication: daily message, one call on Wednesdays.
Deliverable: an interview summary and a working prototype by the 28th.
Review: two hours on the 29th — did we deliver, how did it feel, do we want to do this for four years?
What you actually learn: whether they hit the hours they promised. Whether they raise problems early or hide them until the deadline. Whether disagreement is productive. Whether they do the unglamorous half.
None of that comes out in a conversation, and all of it predicts the next four years.
First-time founders often run a trial that is too easy or too vague — 'let's explore some ideas together for a month'. With no deliverable and no deadline, nothing gets tested, because pressure is the thing you are trying to observe.
The second mistake: not holding the review honestly. If it went badly, say so. Continuing out of politeness is how people end up with a co-founder they already knew was wrong.
Work you genuinely need done. Not a made-up exercise — the stakes are part of the test.
Two to four weeks with a hard end date. Open-ended trials drift into de facto commitment without a decision.
Each person owns a deliverable. You are testing whether they deliver theirs, not whether you can help them.
Then observe whether reality matches. This one signal is worth most of the exercise.
Two hours. Did we deliver? How did it feel? Would I want this for four years? Say the real answer.
That is the entire point. A trial you were never going to act on is theatre.
With anyone you have not worked with professionally. Non-negotiable.
You can skip it with someone you worked closely with for years — you already have the evidence. Still have the conversation.
Apply this to your own startup in My Full Journey (free account).