Portland hit -4°F for nine consecutive days in January. Six smart thermostats were already installed and logging.
What we were testing
The marketing claim every learning thermostat makes is adaptation: as weather changes, the thermostat adjusts. The implicit promise is that a polar vortex triggers pre-heating, runtime extension, and energy optimization automatically.
We had three specific questions: Does pre-heat scheduling shift when outdoor temps drop significantly below historical norms? Does the device continue to hold setpoint without manual override? Does energy usage scale appropriately or does the system hunt?
The results, briefly
Nest Learning Thermostat (4th Gen): Pre-heat extended by 34 minutes autonomously. Setpoint held within 0.4°F for all nine days. No manual override needed. This is what adaptation looks like.
Ecobee Premium: Pre-heat extended manually via eco+ suggestions, which we accepted. Required two manual interventions when auxiliary heat staging wasn’t triggered early enough on night three. Still the best Ecobee has shipped.
Honeywell T9: Held schedule without dynamic adjustment. No cold-snap learning. Setpoint held, but only because the furnace ran 68% longer than baseline — it didn’t adapt, it just worked harder.
Three budget models: All three held setpoint adequately. None adapted schedules. One required a manual reboot on day five after connectivity loss.
The takeaway
A $280 Nest or $249 Ecobee is not just a prettier thermostat. The adaptation logic is genuinely different from a $80 Wi-Fi thermostat. In a normal week, the difference is marginal. During a polar vortex, it’s the difference between a home that adapts and a home that fights you.
The full thermostat rankings are in the gift guide.