The Setting Drifted vs the Part Failed Decision Tree
Why this matters
When a system reads or delivers the wrong value, you are looking at one of two stories: a setting that slowly wandered off (drift) or a part that gave out (failure). They feel similar at the panel but they are not the same job. Drift is recalibrated in minutes for the cost of your time; failure means a part and a different ticket. Calling drift a failure overcharges and over-repairs; calling a failure drift sends you back next week. This tree sorts them across HVAC, plumbing, electrical controls, appliances, and pool gear.
Start here: how did it go wrong, over time
The single best clue is the shape of the change, not the size.
- Gradual, one direction, getting slowly worse: classic drift. Calibration creeps, sensors age slowly, springs relax, contacts slowly oxidize. The value walks away from spec a little at a time.
- Sudden, with no in-between: classic failure. A part that was fine yesterday and wrong today, especially with a noise, smell, smoke, or trip, broke. Things do not "suddenly drift."
Slow and one-directional points to drift. Abrupt points to failure. That is your first branch.
If the change was gradual
Walk it from the simplest cause.
1. Measure the actual against spec and note the offset. A steady, fixed offset in one direction (always reads a little high, always a little low) is the fingerprint of drift. The device still tracks changes, it is just shifted.
2. Try to calibrate it back to spec (see the calibration method article). If it adjusts cleanly into tolerance and holds through a cycle, it was drift. Done. Price it as an adjustment.
3. If it will not hold after calibration, the adjustable element is worn out, the spring has lost tension, the sensor has aged past its range, the contact surface is too eroded to settle. Drift that cannot be corrected has become a failure. Replace the part.
4. If it cannot reach spec at all, same conclusion: the component is at the end of its life. You calibrated it as far as it goes and it is still wrong. That is a repair.
If the change was sudden
1. Look for failure forensics first (see the reading-the-scene articles): scorch, a blown or swollen component, corrosion, a leak, a burnt smell. Physical evidence settles it fast, that is a failed part.
2. Check whether it reads a fixed wrong value or no value. A sensor or control reading a stuck, pinned, or nonsensical value (full-scale, zero, or wildly out of range) is usually a dead or shorted part, not drift. Drift gives plausible-but-wrong numbers; failure often gives impossible ones.
3. Test the suspect part against its own spec, not the system output. Measure the component directly (resistance, continuity, capacitance, pressure-holding, mechanical travel). A part out of its own spec is failed, replace it.
4. Rule out an upstream cause before condemning the part. A "failed" reading can be a good part fed bad power, bad signal, or a bad reference. Confirm the part is actually bad, not just downstream of a problem, before you swap it.
The tells that separate them
| Clue | Points to drift | Points to failure |
|---|---|---|
| Onset | Slow, over weeks/months | Sudden, often with an event |
| Direction | Steady, one way | Pinned, erratic, or impossible value |
| Calibrates back to spec | Yes, and holds | No, or won't hold |
| Physical evidence | None; part looks fine | Scorch, swelling, leak, smell |
| Behavior | Tracks changes, just shifted | Stuck, dead, or wildly wrong |
| History | Long quiet decline | Worked, then quit |
If it calibrates back and holds, it drifted. If it shows damage, won't hold, or reads impossible, it failed. Most cases land cleanly on one side once you measure and look.
The mixed case: a part causing apparent drift
Watch for the trap. A slowly failing component (a sensor degrading, a valve seat eroding, a contact corroding) can look like drift because it gets worse gradually. The tell is step 3 on the gradual branch: if calibration will not hold, the "drift" is a dying part. Do not keep recalibrating a component that will not stay put, you are babysitting a failure. Recalibrate once; if it walks off again soon, replace it.
Confirm before you commit
- State your call: drifted or failed, and the evidence.
- If drifted, calibrate to spec and verify it holds a full cycle.
- If failed, confirm the part is out of its own spec and that nothing upstream caused it.
- Re-test the system. A recalibration that drifts again within the visit is a failure you mislabeled.
The honest version of this job is the profitable one long-term: you do not bill a part for an adjustment, and you do not let a tired part go out the door looking calibrated, only to bring the customer right back.
References
- Trade-standard diagnostic practice (measure, calibrate, then condemn the part)
- Manufacturer documentation for tolerance, drift limits, and component specs
- See related: Calibrating to Spec (the method); The Adjustment vs the Repair Decision Tree
- See related: Reading Scorch and Heat Marks; Reading Rust and Corrosion Patterns