It Worked Until Last Tuesday and Nobody Touched It

Why this matters

"It worked until last Tuesday and nobody touched it" is the sentence that starts most arguments about connected equipment. The customer is telling the truth: nobody in the house touched it. That does not mean nothing changed. This case was settled at the kitchen table in about twenty minutes, without opening a single cabinet, because two records existed: the shop's own install record and the device's event log. Read it for the method, which is reconstructing a fault backwards from what was written down, and for the gap in the shop's record that nearly let the cause slip.

The call

A heat pump customer called in early winter. Three complaints, all starting the same week: a hot-dust smell from the vents every morning, the house reaching temperature "too fast, like a furnace", and a utility app showing daily usage up sharply. The system had been installed by the shop two winters earlier and serviced in the fall with no findings. The customer was clear that nobody had changed any settings, and the office had no record of a visit since the fall service.

That cluster points one direction. The smell of hot dust and a fast morning recovery are both signatures of auxiliary heat, the electric resistance strips in the air handler that supplement the heat pump. Strips heat quickly, smell the first few times each season as dust burns off, and draw far more power than the compressor for the same heat. The question was not whether the strips were running. It was why they had started running when they had not before.

What the dispatcher did before rolling a truck

The dispatcher asked the customer to open the thermostat app and read two screens aloud: the device information page and the event history. That took five minutes on the phone and changed the visit from a diagnostic call into a settings call. The first screen showed a firmware version the shop's record did not match. The second showed the timeline below.

EVENT HISTORY (thermostat, newest last)
TUE 02:14  firmware update installed  4.2.1 -> 4.3.0
TUE 02:16  device restarted
TUE 02:17  settings migrated to new version
WED 05:30  early recovery for WAKE 06:30
WED 05:30  setpoint 68F (was 62F)
WED 05:30  heat pump stage 1 ON     outdoor 41F
WED 05:32  aux heat ON              outdoor 41F
WED 06:19  setpoint reached, aux heat OFF
WED 06:41  heat pump stage 1 OFF

Version numbers here are illustrative; the shape is what real logs look like. Three lines carry the case. The update landed at 02:14 on the Tuesday the customer named. The first morning recovery after it started an hour before the 06:30 wake time and brought on auxiliary heat two minutes after the heat pump, at an outdoor temperature of 41 F. And a line that says "settings migrated" is the device telling you it rewrote its own configuration.

The shop's record, side by side

The install record from two winters earlier listed the thermostat model, firmware version 4.2.1, and four settings changed from default. The one that mattered:

• Auxiliary heat outdoor lockout: 30 F. A lockout tells the thermostat not to use the strips when it is warmer outside than the set temperature, because above that point the heat pump alone can carry the house and strips are the expensive way to heat. The default on this model was no lockout at all.

The tech compared the live settings screen against that list. Three of the four changed settings were still at the values recorded. The lockout read "none". That is a default reset: the update migrated the settings, and one of them came out at the factory value rather than the value the installer set.

That alone explains strips running at 41 F. It does not fully explain the pattern, because with no lockout the thermostat would normally still try the heat pump first and add strips only if recovery lagged. Two minutes is not lagging.

The second change, which the record could not have caught

On the same settings screen was an option the installer had never seen, because it did not exist in version 4.2.1: an early-recovery feature, switched on, described in the app as reaching the scheduled temperature by the scheduled time. To hit 68 F by the wake time from a 62 F setback, the thermostat had decided to bring the strips in almost immediately. A 6 F recovery on a heat pump at 41 F outdoors is slow but entirely within its capacity; the feature traded that slowness for speed, and paid for it in resistance heat.

This is the gap in the record, and it is worth being exact about it. The shop's record did everything a record of changed settings can do. It could not list a setting that had not been invented yet. What would have caught it faster is a photograph of every settings screen at commissioning: comparing photo to screen shows a new line immediately, where comparing a list of four values shows only that those four values moved.

Putting numbers on it

The customer's utility app showed daily usage. The thermostat's runtime history showed auxiliary heat minutes per day. Both are records nobody had to create for this call, which is why reading them first is cheap.

• The week before the update: daily use around 38 kWh, auxiliary heat about 6 minutes a day, all of it on the two coldest mornings, when it was below the 30 F lockout.
• The week after: daily use around 49 kWh, auxiliary heat about 72 minutes a day.

The strips in this air handler are rated 10 kW on the nameplate. The extra strip time is 72 minus 6, which is 66 minutes a day. At 10 kW, 66 minutes is 66 divided by 60, times 10, which is 11 kWh a day. The rise in daily use is 49 minus 38, which is also 11 kWh. The two independent records agree, which is what lets you tell the customer with confidence that the strips account for the rise.

Two cautions belong next to that agreement. First, the comparison assumes similar weather in both weeks; the thermostat's own outdoor-temperature history showed overnight lows in the same band both weeks, so the assumption held here, and it should be checked rather than assumed. See related: Reading Energy Data Without Misleading the Customer. Second, the strips ran while the heat pump also ran, so the 11 kWh is extra strip energy on top of a compressor that kept working, not a swap of one for the other. If the compressor had dropped out, the arithmetic would not close and you would be looking at an equipment fault.

What was ruled out without a ladder

The record was strong enough to set aside the usual causes of aux heat running, and it helps to say why each one went, because each would have been a full diagnostic visit.

• A heat pump capacity problem (low refrigerant charge, a failing compressor, a blocked outdoor coil). These make the heat pump fall behind, and the thermostat then adds strips after a lag. The log showed strips two minutes after the compressor, which is a control decision, not a capacity shortfall. A capacity fault also would not start on the exact morning after a firmware update.
• Defrost. Heat pumps bring strips on during defrost to temper the air. Defrost events appear in the log as their own lines and run minutes, not most of an hour, and at 41 F with dry weather there were none.
• A wiring fault energizing the strips. That would run them regardless of the thermostat. The log showed aux heat on and off at the thermostat's command, matching the setpoint being reached.

If any of those had been in play, the tech would have opened the air handler, and that is panel work on a 240 V circuit feeding the strip heater. It is for a qualified person only: open and lock the air handler's disconnect, tag it, and prove the circuit dead with a meter proved on a known live source before and after, per NFPA 70E-2021, 120.5, with work practices at 29 CFR 1910.333(b)(2). Strips stay hot for several minutes after they shut off, so wait before reaching in. In this case none of that was needed.

The fix and how it was confirmed

The tech set the auxiliary lockout back to 30 F and switched the early-recovery feature off, after explaining to the customer what it did: "It gets the house warm before your wake time by using the electric strips. Without it, the heat pump starts at wake time and takes longer, about forty minutes on a morning like this, and uses much less power doing it." The customer chose the slower, cheaper recovery and moved the wake time forty minutes earlier instead.

Confirmation came from the same two records, read again three days later over the phone:

• Auxiliary heat minutes: 0, 0 and 4 on the three mornings. The 4 minutes came on a morning at 28 F, below the 30 F lockout, where strips are allowed and expected.
• Daily usage: 37, 39 and 40 kWh, back in the band of the week before the update.
• No hot-dust smell reported.

The tech then updated the install record: new firmware version, date read, the lockout restored, the new feature noted as present and off, and photographs of every settings screen attached. The record now shows two versions and the exact difference between them.

The conversation about blame

The customer asked, reasonably, whether the shop should have known. The honest answer was in three parts. The update came from the maker, on a schedule nobody in the house or the shop controls on this model. The setting the shop had chosen was correct and was undone by the update, which the record proves with a date and a value. And the shop would now photograph every settings screen at install, because a list of changed values cannot see a feature that arrives later.

That answer worked because it was specific. "The thermostat updated itself" sounds like an excuse when it is said without evidence. Said with the log line at 02:14 and the record showing 30 F, it is a finding.

The office then took one more step that the case made possible. It searched its records for the same thermostat model on version 4.2.1 and found 11 other installs. The tech who closed this call drafted a two-sentence message to those customers explaining the update might have reset the lockout and switched on early recovery, and how to check. See related: Firmware Updates: The Silent Change to Installed Equipment, which owns how to run that kind of fleet check.

What carries to the next case

The method transfers to any connected device: before you diagnose hardware on a "nobody touched it" call, ask for the version and the event history, and compare them against your record of what you left. If the version changed on or just before the day the problem started, read the settings screen against the record line by line before you pick up a tool. The device usually wrote down what happened to it. Your job is to have written down what you left, so the two can be compared.

References

• NFPA 70E-2021, 120.5, establishing an electrically safe work condition
• 29 CFR 1910.333(b)(2), deenergizing and lockout and tagging of electrical equipment
• Manufacturer thermostat and air handler installation manuals, auxiliary heat lockout and staging settings
• See related: Firmware Updates: The Silent Change to Installed Equipment
• See related: Leaving a Device Record the Next Tech Can Use