Minimum Viable Governance

There is a persistent belief that governance is what large programmes need, and that small teams are exempt until they scale. It is exactly backwards. Small teams feel the cost of ungoverned evidence soonest, because there is nobody spare to reconstruct what a departed colleague knew, and no slack to spend re-running a test that was already run. The other persistent belief is that governance means process weight. It does not have to. Minimum viable governance is three components. A team of two can run all three, and none of them requires a tool.

One gate that can refuse

Before anything is built, the hypothesis gets reviewed against three questions. What do we believe, and what evidence makes us believe it? What would change our minds? What will we do differently depending on the result?

The review must be able to send things back. A gate that waves everything through is ceremony, and a small team cannot afford ceremony. If nothing has been sent back in three months, the gate is not working, and it is better to know that than to pretend.

This is deliberately one gate, not five. The gate sits before build because that is where refusal is cheapest.

Success defined before launch

Before the test goes live, write down the primary metric, the direction and size of change that counts as a win, how long the test runs, and what you will do in each outcome: win, loss, inconclusive. Then do not touch it.

This is the cheapest protection there is against the most common corruption in experimentation, which is deciding what success means after seeing which numbers moved. It costs five minutes per test. Skipping it costs the trustworthiness of every result you report.

A decision record

One running log, in whatever tool you already use. Every experiment gets a line when it concludes: what was decided, on what evidence. Every rejected idea gets a line too: what was declined, and why.

The rejection half is the part almost nobody keeps and the part that pays best. Ideas are not rejected once; they return every six months wearing new words, and without a record each return costs a fresh debate. With a record, triage takes a minute and the reasoning compounds instead of evaporating.

What to skip at this size

No committees. No scoring models. No workflow tooling. No documentation standards beyond the log. No gates on statistical method beyond the honesty of the success definition. All of these have their place in larger programmes; at this size they are weight without value, and weight without value is how governance gets a bad name.

Why this is enough

These three components are small, but they are the load-bearing three. They give the programme the one property that matters: the ability to answer for its evidence. What did we test, what did we conclude, what did we decline, and what did we decide? A two-person team that can answer those questions is better governed than most enterprise programmes, and when the programme scales, these habits scale with it. Tooling can be added to habits. It cannot substitute for them.