Method / Stage gates

From first hypothesis to validated system.

A programme moves through six stages. Each ends in a gate with a plain test: can the next team rely on what was produced?

01
Define
Turn a broad ambition into a question that can be answered with a measurement.
Outputs
- — Written hypothesis and scope
- — List of constraints: packaging, cost, schedule, regulation
- — Success criteria agreed before any testing
Risks
- — Vague targets that no measurement can confirm
- — Hidden constraints that surface late
02
Model
Build the simplest model that could be wrong in an informative way.
Outputs
- — Model with stated assumptions and validity range
- — Parameter sensitivity ranking
- — Verification against known analytic cases
Risks
- — Over-detailed models that hide their assumptions
- — Parameters tuned to one result
03
Prototype
Make the smallest physical artefact that can test the riskiest assumption.
Outputs
- — Scale or sub-system prototype
- — Manufacturing and tolerance notes
- — Test fixture design
Risks
- — Prototype differs from the model in ways nobody recorded
- — Fixture dominates the result
04
Instrument
Decide what to measure, at what rate and with what uncertainty — before the test.
Outputs
- — Channel list with ranges, rates and filters
- — Clock and synchronisation plan
- — Calibration records
Risks
- — Missing the channel that would have explained the result
- — Unsynchronised clocks across loggers
05
Validate
Compare prediction with measurement, with the scatter of the measurement known.
Outputs
- — Prediction-versus-measurement report
- — Updated model and validated domain
- — Open questions, ranked
Risks
- — Adjusting the model until it fits one test
- — Treating agreement outside the tested range as proof
06
Handover
Pass on what is known, what is assumed and what still needs real-world verification.
Outputs
- — Dossier: data, models, assumptions, limitations
- — Recommended next tests
- — Reproducible analysis scripts
Risks
- — Results separated from the conditions that produced them
- — Assumptions lost between teams

Stage gates

G01DefineHypothesis and acceptance criteria signed off.
G02ModelModel verified; sensitive parameters identified.
G03PrototypePrototype as-built matches the documented configuration.
G04InstrumentDry run produces clean, aligned, repeatable data.
G05ValidateAgreement within pre-agreed criteria, or a documented reason for the gap.
G06HandoverReceiving team can reproduce a headline result from the dossier.

Principles

Write the assumption down.

Every model leaves something out. The useful ones say what, and how large the omission might be.

Decide agreement beforehand.

Acceptance criteria written before the test cannot be quietly adjusted to fit the result.

Keep the raw data.

Filtered and validated channels are derived, never substituted. Every change is logged.

Discuss a programme

Stage gates

G01DefineHypothesis and acceptance criteria signed off.

G02ModelModel verified; sensitive parameters identified.

G03PrototypePrototype as-built matches the documented configuration.

G04InstrumentDry run produces clean, aligned, repeatable data.

G05ValidateAgreement within pre-agreed criteria, or a documented reason for the gap.