How it works
Equipment-side sensing + predictive ML + contractor dispatch.
Step 1
A partnered contractor installs two small boards. A hub at the outdoor unit clamps onto the compressor and fan wiring, straps temperature probes to the refrigerant lines, and reports over cellular — no Wi-Fi password, nothing for you to configure. A node at the air handler reads supply and return air, humidity, blower current, static pressure across the filter and ducts, and the drain pan.
About 25 minutes per board for residential; 2–4 hours for commercial, which adds refrigerant pressure probes.
Phase 1 — Rule-based (launch)
Thresholds on current draw, temperature split, runtime, short-cycling. High precision, conservative.
Phase 2 — Anomaly detection (months 3–9)
Per-system baselines using z-score on residuals. Catches slow degradation that rules miss.
Phase 3 — Predictive ML (months 9–18)
Supervised models forecasting time-to-failure for specific failure modes.
Step 2
After ~30 days of data, the platform has learned your system's normal pattern — runtime cycles, temperature split, static pressure, and current draw at different outdoor temperatures. Every compressor start is recorded too: a weakening run capacitor makes starts slower week by week, the earliest warning most systems give. Each system gets its own baseline; we don't compare you to a generic average.
Rule-based alerts fire from day one. Anomaly detection kicks in once baselines are established. Predictive models arrive as labeled data accumulates.
Step 3
When an alert crosses a confidence threshold, you get a push notification with the predicted failure mode, estimated cost to ignore, and a one-tap option to schedule a visit from your contractor.
The contractor receives the diagnostic data with the dispatch ticket — not just "customer says it's warm." This means faster, cheaper visits and fewer "couldn't find anything wrong" calls.
Example alert
Compressor current anomaly
Your compressor is drawing 18% more current than baseline. Likely cause: low refrigerant or fouled coil.
Estimated cost to ignore: $42/month in extra energy. High risk of compressor failure within 60 days.
Step 4
Monitoring pays for itself in lower utility bills, not just avoided breakdowns. A system losing refrigerant or with a fouled coil can consume 20–40% more energy for months before failing. We quantify that.
Power is metered at the equipment, so monthly reports show measured kWh — not an estimate — against your baseline, with dollars saved and CO2 avoided. Residential customers saving $25–40/month on energy are cash-flow positive from month one.