From Raw Readings to Actionable Signals
Collect data at sensible intervals, respecting battery life and privacy. Apply light filtering, derive features—like start current, runtime distribution, and temperature deltas—and compare against recent baselines. Flag persistent deviations instead of single spikes. Summarize with friendly language that explains what shifted and why it matters. Attach a quick checklist—clean, tighten, inspect, replace—and a link to documentation. Each alert should feel like a coach’s nudge, not a scolding interruption.