One machine class, then another

A heat pump's efficiency falls as it gets colder outside, and below a 99% winter design temperature of 25°F, a standard ducted unit starts leaning hard on electric-resistance backup heat — which is only 100% efficient, not the 300%-plus a heat pump gets from moving rather than generating heat. Past that line, this site switches its recommendation to a variable-speed cold-climate unit (ccASHP), priced at a $2,500 premium, because it holds capacity down to roughly -13°F instead of falling apart well above zero.

What that looks like at the extremes

Fairbanks, AK's design temperature of -50.3°F is the coldest in this dataset; a cold-climate unit there still only reaches a seasonal COP near the model's floor. Bozeman, MT and Great Falls, MT sit in similar territory, both deep enough into cold-climate country that a standard unit was never a serious option. At the other end, Honolulu, HI's design temperature of 64.9°F and Miami, FL's 56.7°F both sit so far above the 25°F line that a standard unit runs near the top of its COP curve all year, and the cold-climate premium would be spent for nothing.

Why the model folds backup into one number

Rather than guessing how many hours a standard unit spends on resistance backup below its comfort point, this site folds that entirely into the seasonal COP figure shown on every city page. An explicit backup-hours estimate, without real temperature-bin data behind it, would be a guess dressed up as arithmetic; the folded-in number is honest about being an average rather than pretending to be more precise than the input data supports. See the full derivation in the methodology guide, and how this interacts with a city that fails to pay back at all in the no-payback guide.