The seven stages of a defensible decision

Most people, when they say a decision was "good", mean it worked out. That is hindsight. It is also useless as a way of getting better.

A more useful word is defensible. A defensible decision is one you can describe, six months later, in a way that a serious person — a board member, a regulator, an investor, your future self — would call reasonable given what you knew at the time. The outcome is a separate question.

Kauzio is built around seven stages. Together they are what we call the Decision Loop. They are the same on the shop floor, in the founder's kitchen, and behind an API. They are not optional, in the sense that every decision goes through them — the only question is whether you do it explicitly or by accident.

1. Suggested

The decision exists. Someone — a person, a rule, a model — has put it on the table. Order 200 units. Discount this SKU. Hire this person. Close this store. The act of writing it down is the first piece of governance. A decision that was never explicitly proposed cannot be reviewed.

2. Reviewed

A second voice sees the proposal. It may be a colleague, a model, a checklist. The point is not consensus — the point is that the case has been heard against an opposing view before it is committed. Most bad decisions die at this stage, if the stage exists at all. Most bad decisions in independent retail get made because this stage does not exist at all.

3. Approved

Someone with authority commits. In a small business that someone is usually the owner; in a larger one it is a named role. The reason this stage matters is accountability anchoring. The decision has a name attached to it. Anonymous decisions cannot be learned from, because nobody remembers making them.

4. Executed

The decision becomes an action in the world. The order is placed; the discount is applied; the person is hired. The gap between approval and execution is where most operational risk lives. A decision approved on Monday and executed on Friday is not the same decision — markets move, prices move, context moves.

5. Measured

The outcome is observed. This is the stage that almost nobody does properly. Measuring means going back, at a defined interval, and comparing the actual outcome to the expectation set at the point of approval. Without this stage you do not have a loop, you have a list.

6. Learned

The variance between expected and actual becomes a finding. Not "the decision was wrong" — that is the lazy version. The useful version is "the model under-estimated demand because we did not account for the school holiday." The output of this stage is a written, dated correction to your own thinking.

7. Closed

The decision is sealed. The receipt — proposal, review, approval, execution, measurement, learning — is signed and archived. You do not have to look at it again. But it exists, forever, in a form that someone else could pick up and understand without you in the room.

Why this matters more than "being right"

A correct decision that you cannot defend is fragile. It survives only as long as the outcome is good. The first bad outcome kills it, kills your authority, and kills the team's appetite for similar moves.

A defensible decision is durable. Even when the outcome is bad — and outcomes are often bad, especially in fields with high variance — you can stand behind the process. The board does not punish bad outcomes from good processes. They punish bad outcomes from invisible processes. There is a large literature on this in decision science, going back to work by Daniel Kahneman and others; the operational version is simpler than the academic one.

The minimum viable loop

If you run a small business, here is the version that costs nothing.

Buy a notebook. For every decision that involves more than £500 of money or two weeks of time, write four lines:

• What I am about to do.
• What I expect to happen in 30 days.
• The strongest argument against doing it.
• The date I will look back.

When the date arrives, write one more line: what actually happened. That is the entire system. Everything Kauzio adds on top — the engines, the receipts, the agent debates — is in service of letting that loop survive a real working week, when the notebook gets lost and the Tuesday rush eats the review.

A note on the word "defensible"

We use "defensible" rather than "correct" deliberately. Correctness is a property of outcomes. Defensibility is a property of process. You only control one of them. Build the loop around the one you control.