A seasonal figure is an average, and averages hide the part of the winter that decides the bill. NRCan is explicit that its HSPF figures are calculated for AHRI Region V, a climate like Ottawa’s, and that actual seasonal efficiency varies with where the machine is installed. It is also explicit that a coefficient of performance between 2.0 and 5.4 at plus 8 degrees falls to between 1.1 and 3.7 at minus 8. The hours a Toronto heat pump spends below freezing are not the hours the seasonal rating was measured in.
Then there is the backup. Most air source installations shut the compressor off somewhere between minus 15 and minus 25, and something else covers the house below that. If the something else is electric resistance heat, those hours run at a coefficient of performance of one and cost roughly six times what the furnace would have. If it is the existing gas furnace, the house is on a dual fuel system and the honest comparison is a blend of the two columns rather than either of them.
What is left out on the other side is the equipment. This page prices fuel, not capital: no purchase price, no rebate, no financing, no maintenance, and nothing for the air conditioner a heat pump makes unnecessary. It is one line of a decision that has several, and the ones about capacity, ductwork and electrical service are on the HeatPumpIndex cost guide and comparison pages.