The goal isn't to redesign everything.
It is to know what deserves confidence, what doesn't, and why — before important decisions become expensive to reverse.
Confidence needs a trail.
A decision is easier to trust when the reasoning around it can be followed, questioned, and revised.
- 01Context
- 02Decision
- 03Assumption
- 04Evidence
- 05Risk
- 06Recommendation
- 07Confidence
Not all evidence carries the same weight.
The point is not to eliminate uncertainty. It is to name what a conclusion is resting on, and how much of it is known.
"Traffic will remain predictable."
- 01Observed
Seen in the system or in the work as it exists.
- 02Measured
Supported by a repeatable measurement.
- 03Documented
Recorded in a decision, contract, or explicit constraint.
- 04Estimated
Reasoned from incomplete information.
- 05Assumed
Believed to be true, but not yet tested.
The outcome should be clear enough to act on.
A recommendation is not a verdict performed with confidence. It is a bounded response to the evidence and risk currently visible.
What are we committing to?
- Accept
The decision is sufficiently understood for the context.
- Accept with conditions
Proceed, but make the conditions explicit and owned.
- Investigate
Important uncertainty remains before commitment.
- Reconsider
The current decision does not fit the evidence or constraints.
- Reject
The risk is not proportionate to the value or reversibility.
- Experiment before commitment
Learn something material before making the decision durable.
A timeout is not a decision.
The method becomes useful when it is applied to the awkward detail that a diagram tends to leave implicit.
A payment service times out after the request leaves our system.
Retry the payment request automatically.
Are retries actually safe?
Idempotency keys are documented. Duplicate behavior is untested.
A second charge, or an ambiguous customer state.
Experiment before commitment.
Conditional until duplicate behavior is measured.
A useful record leaves the decision written down, the assumptions named, the evidence separated from estimates, and the recommendation clear enough to revisit.
Not every uncertainty deserves the same attention.
What happens if we are wrong?
The more consequential, uncertain, and difficult to reverse a decision is, the more deliberately it should be examined.
Know what deserves confidence, and why.