Define the question and the decision
An Acoustic Discovery Pilot starts with a target asset or repeatable process, a practical engineering question, and the people who will review the result. Define the operating states and fault hypotheses that matter.
Agree how the team will decide whether the evaluation was useful. Avoid choosing an accuracy target or economic benefit before representative data and the operating assumptions are understood.
Check feasibility before collecting history
Assess sensor suitability, installation access, safety requirements, environmental noise, and available process context. Identify who can authorize data collection and what handling restrictions apply.
Collect representative normal operation across relevant states. Check clipping, silence, interruptions, noise quality, and recording consistency. A baseline that misses important states can make normal changes look unusual.
Review the evidence and the errors
Use representative labeled examples where they exist. Review precision and recall only when the labels and evaluation design support those measures. Also track false alerts per asset and period, review time, and sensitivity to operating changes.
Keep a record of technician findings. An event without a confirmed cause should remain unresolved rather than being counted as a successful diagnosis.
Decide whether expansion is justified
Review signal quality, operating state coverage, event usefulness, workflow fit, and engineering effort. Evaluate economic value using actual customer assumptions.
Document what worked, what did not, and what additional evidence is needed. A useful pilot produces a defensible next decision, including when a particular application is not a suitable fit.