Comparing this year against last needs the weather taken out

The comparison everybody makes and nobody adjusts

After insulating a loft, changing a boiler or installing an array, the natural test is to compare the year after against the year before. It is the right instinct and it produces an answer that is dominated by something other than the measure being tested.

Heating demand is roughly proportional to how cold it was and for how long, and that quantity varies substantially between consecutive years in most climates. A mild winter and a hard one can differ by a margin comfortably larger than the saving from a typical efficiency measure.

So a household that insulates before a cold winter may see consumption rise and conclude the work failed. A household that does nothing before a mild one may congratulate itself. Both are reading the weather.

Degree days are the standard correction

The method for removing weather from a comparison is old, simple and widely used in commercial energy management. Pick a base temperature, being the outside temperature above which a building needs no heating, and for each day take the difference between that base and the day’s average outside temperature. Sum those differences over the period.

The result is a heating degree day total, and it is a single number describing how demanding the weather was for heating. Divide energy used by degree days and you get consumption per unit of coldness, which is a property of the building and its systems rather than of the season.

Cooling has a mirror version, summing the amount by which temperature exceeded a base. In climates where both heating and cooling matter, both totals are needed, and they respond to different measures. Degree day figures for most regions are published by meteorological services and by various energy bodies, so the data does not have to be gathered locally.

Choosing a base temperature is the imprecise part

The base is not a physical constant. It represents the outside temperature at which the building’s internal gains, from occupants, appliances and sunlight, balance its heat losses, and it therefore depends on the building. A well-insulated house with high internal gains has a lower base temperature than a draughty one.

Published series usually offer a choice of bases, and conventions differ by country. What matters for a comparison is consistency rather than correctness: use the same base for both periods and the ratio remains meaningful even if the base is not perfectly matched to the building.

And here is the subtlety that catches people out. Insulating a house lowers its true base temperature, because it now needs heating at a colder outside temperature than before. So the correction that makes the comparison fair is itself changed by the work being assessed, and a rigorous analysis accounts for that. For a household comparison it is enough to know the effect exists and biases the result conservatively.

What the weather does not explain

Not all consumption is weather-dependent, and separating the parts is the other half of the exercise. Plot energy use against degree days for a series of periods and, in a heating-dominated building, the points tend to fall near a line. The slope of that line describes the building’s heat loss. The intercept describes everything that carries on regardless: hot water, cooking, lighting, appliances, standing loads.

That decomposition is genuinely informative. A measure that improves the fabric should reduce the slope and leave the intercept alone. A measure aimed at appliances or hot water should do the reverse. If the numbers move in the wrong place, something in the story is off.

It also identifies the more embarrassing failures. A slope that did not change after significant insulation work suggests the work is not performing, whether through a gap, a bypass or a detail not executed as designed.

Doing this without becoming an analyst

A household version needs three things: meter readings at consistent intervals, degree day totals for the nearest published location, and the discipline to keep both for long enough that a pattern emerges. Monthly is fine. Anything shorter is noisy, because heat stored in a building blurs the boundary between one period and the next.

A year is the minimum useful comparison and two is better, because a single year contains only one winter and one winter is a small sample. There is also a limit to the precision available, since occupancy changes, a new appliance or a different thermostat habit all shift the numbers and none of them is recorded anywhere.

Which is the honest caveat to end on. This method turns a meaningless comparison into a rough one, and rough is a considerable improvement. It will not resolve a small saving from a single measure, and anyone promising that level of certainty from household meter readings is overstating what the data can support.