15 FIELD GUIDES · 2 WORKED EXAMPLES
What makes a result usable?
Translate research questions into requirements, acceptance tests, reproducible records and a reliable path to useful hardware.
Research & engineering practice
Start with the question closest to your experiment. Each guide separates the mechanism, an illustrative example, and the evidence needed for the next decision.
What makes an engineering requirement usable?
Turn an intention into a requirement with a measurable outcome, operating conditions, and a verification path. Keep rationale distinct from the obligation.
Read the guide ↗FIELD GUIDEHow do acceptance criteria turn tests into decisions?
Define acceptance criteria before testing, including thresholds, uncertainty, sample rules, and invalid-run handling. Avoid moving the goal after results.
Read the guide ↗FIELD GUIDEWhat should a stage gate actually decide?
Use stage gates to decide whether evidence supports the next commitment. Define exit criteria, unresolved risks, and owners instead of ceremonial reviews.
Read the guide ↗FIELD GUIDEHow do you choose an instrument for the decision?
Select instrumentation from the quantity and decision margin. Compare uncertainty, range, bandwidth, loading, and calibration rather than resolution alone.
Read the guide ↗FIELD GUIDEHow do error budgets differ from failure budgets?
Separate accuracy allocations from reliability allocations. Combine uncertainties and failure probabilities using models that match dependence and consequences.
Read the guide ↗FIELD GUIDEHow do you design for maintenance before deployment?
Assess maintainability through access, diagnosis, replacement, and recovery. Include realistic tools, people, downtime, and verification after intervention.
Read the guide ↗FIELD GUIDEWhat belongs in a reproducibility package?
Build a package that lets another person regenerate a result. Connect data, code, environment, commands, and expected outputs with a tested entry point.
Read the guide ↗FIELD GUIDEWhy should datasets have immutable versions?
Keep data corrections traceable with immutable versions, provenance, and transformation records. Make every reported result point to its exact inputs.
Read the guide ↗FIELD GUIDEWhat should an experiment log capture?
Log experiments so results can be interpreted later: configuration, timing, interventions, anomalies, and raw-data links matter as much as the final number.
Read the guide ↗FIELD GUIDEWhat does preregistration add to an experiment?
Record hypotheses, analysis choices, stopping rules, and exclusions before seeing outcomes. Preserve exploration while making planned tests distinguishable.
Read the guide ↗FIELD GUIDEHow can a negative result still be informative?
Interpret negative results through precision, controls, and the tested claim. Distinguish evidence against a meaningful effect from an inconclusive experiment.
Read the guide ↗FIELD GUIDEWhat makes a replication independent?
Plan replication with new evidence and explicit independence. Separate shared methods from shared failure modes and define what agreement would mean.
Read the guide ↗FIELD GUIDEWhy is a working prototype not a specification?
Distinguish what a prototype happened to do from what a product must consistently do. Capture interfaces, limits, and evidence before scaling production.
Read the guide ↗FIELD GUIDEWhat does environmental qualification establish?
Connect qualification tests to the intended environment and configuration. Track margins, exposures, sequence, and post-test function without overclaiming.
Read the guide ↗FIELD GUIDEWhy does one successful demonstration not prove reliability?
Distinguish showing that a system can work from estimating how consistently it works. Connect reliability claims to exposure, failures, and assumptions.
Read the guide ↗PUT THE IDEAS TO WORK