How to Run a Co-Founder Trial Project

How to Run a Co-Founder Trial Project

A co-founder trial should be a bounded piece of real customer and product work that remains useful even if the pair does not continue.

Define the question, scope, time, roles, decisions, communication, expenses, confidentiality, data, intellectual-property boundary, review, and exit before starting. The goal is to observe behavior under ambiguity without pretending the trial replaces legal agreements.

Choose work that can teach both people

Avoid a fake exercise or an open-ended request to "build the company together."

Choose a two-to-four-week problem with one user decision, one product or delivery decision, and one meaningful disagreement or tradeoff. The result might be a set of workflow observations, a private prototype, a manual service test, or a technical feasibility record.

The work should create legitimate learning for both people. Do not use a trial to extract unpaid labor from someone who lacks clear terms, real choice, a fair path to stop, or access to the result.

The question is not "Can we get along for a month?" It is "What happens when we make promises, meet evidence, divide work, disagree, and recover?"

Write the pre-brief

If a subject creates legal uncertainty, pause and get current advice before the work.

YC's official Startup School video index includes a session on successful co-founder relationships. Harj Taggar's current YC profile links a talk about finding the right co-founder. These are useful attributed perspectives, not substitutes for a written trial boundary.

Keep the scope small enough to finish

A trial that tries to launch a company will mainly reveal that the scope was impossible.

Pick one learning goal and a deliverable that can be reviewed. Define what is out of scope. Set an hours expectation that respects existing work and care obligations. Decide how a missed commitment will be surfaced and renegotiated.

Include at least one moment where evidence can override a preference. For example, the pair might disagree about the user segment and agree in advance to run three representative workflow conversations before choosing.

flowchart TD
    A["Write mutual trial question"] --> B["Agree scope, roles, rights, and exit"]
    B --> C["Complete real user and product work"]
    C --> D["Record decisions, misses, conflict, and repair"]
    D --> E["Write separate retrospectives"]
    E --> F["Compare evidence together"]
    F --> G{"Mutual decision"}
    G -->|"Continue"| H["Obtain formal legal and tax review"]
    G -->|"More evidence"| I["Run another bounded trial"]
    G -->|"Stop"| J["Close work, data, expenses, and users cleanly"]

Observe behavior without diagnosing personality

Record what each person promised, what happened, how updates arrived, and how recovery worked after a miss.

Look at user judgment, product or technical judgment, financial restraint, ethics, documentation, listening, decision clarity, handling of bad news, confidentiality, attribution, and care with data.

Conflict is not automatic failure. The useful evidence is whether the pair can surface the disagreement, understand its basis, choose a decision method, act, and repair trust.

Avoid labels such as "not founder material" or amateur psychological diagnoses. Describe behavior and its consequence for the work.

Test complement and ownership

Complementary skill does not mean permanent silos. Each person should have work they can own and enough shared understanding to challenge important decisions.

Observe whether one person becomes the default owner of every ambiguous task. Notice whether ownership matches authority. Notice whether both people can explain the user, product, business, and risk even when they contribute different specialties.

[[How to Map Product-Founder Fit]] can help place the trial inside the venture's wider capability gaps.

Discuss ambition and company shape

Two people can work well and want different companies.

Discuss intended customers, venture scale, funding, profitability, geography, pace, time commitment, compensation needs, risk, public profile, hiring, exit goals, control, and which outcomes each person would find unacceptable.

Do not force agreement where a difference can remain visible. Record which subjects need another trial, external evidence, or professional review.

Protect users, data, and existing obligations

Use the least sensitive data and narrowest access possible. Obtain user consent for trial research. Do not carry employer, client, or confidential material into the project without clear authorization.

Review current employment agreements, invention-assignment terms, confidentiality duties, outside-work policies, client commitments, and conflicts before creating material that could become disputed.

Do not assume a personal account, free tool, shared drive, or verbal understanding answers ownership or privacy.

Hold separate and joint retrospectives

Each person should first write what increased trust, what reduced it, which work felt complementary, where ownership failed, how conflict changed the decision, what they learned about the desired company, and which risk remains unresolved.

Then compare records. Give each person room to decline without persuasion or punishment.

The valid outcomes are continue toward formal review, run another bounded trial, change roles, remain collaborators without founding together, or stop.

Cross the formal boundary deliberately

Before equity, incorporation, intellectual-property transfer, fundraising, employment, contractor arrangements, compensation, bank access, binding customer commitments, or securities activity, obtain appropriate legal and tax review.

The SBA launch guide separates structure, registration, tax identifiers, licenses, banking, and insurance. These choices depend on the people, activity, and jurisdiction.

The trial produces evidence about collaboration. It should end with a clean decision, not accidentally create commitments nobody reviewed.

This page was developed with AI assistance from the E056 transcript and linked primary sources, then structured for human legal, employment, intellectual-property, compensation, privacy, founder, tax, and editorial review. It is not an equity, formation, employment, contractor, IP, securities, or compensation agreement.