Case studiesCase study — a household

The house that audited its own heating.

Is it cheaper to keep underfloor heating ticking over all day, or to run it on a schedule? This household had argued the point for years. Two winters of measured data settled it, cut the gas used per degree day by 10–12%, and put a number on exactly how much heat the house leaks.

Verdict

10–12%

extra gas per degree day when run constant

Heat loss

450 W/K

the house itself, measured over two years

Readings

5 min

room temperatures, round the clock

Build

1,500 lines

of Python, running unattended

The situation

Every house with underfloor heating hosts the same debate. One side says the screed holds its heat, so keeping the system ticking over around the clock must be cheaper than reheating it from cold. The other side wants it off overnight and suspects the always-on habit is what’s inflating the gas bill. Each position sounded perfectly reasonable, and neither could prove a thing.

The evidence already existed. A Heatmiser Neo system was measuring every room, a smart meter was recording gas and electricity every half hour, and the tariff priced each of those half hours differently. The readings simply lived in three separate apps, none of which would answer the actual question.

What we did

A small Python app now gathers the lot into one database on a home server. Room-by-room temperatures arrive every five minutes from the heating hub. Gas and electricity come in half hour by half hour from the smart meter, matched against the tariff so every reading carries a price in pounds. Outside temperature comes from a free weather service, because you can’t judge a heating system without knowing what it was up against.

A dashboard charts it all, and a companion command-line tool adjusts room setpoints and heating schedules. That tool has guard rails: it refuses any temperature outside 5–24 °C and writes every change to an audit log. The whole build is about 1,500 lines of Python in ten small modules. It has run unattended since the day it went in.

As with all our work, AI development tools did the heavy lifting. Each feature went from a plain-English description to reviewed, working code in the same sitting, tested against the live heating hub before moving on.

What the data settled

The headline: running the heating near-constant uses 10–12% more gas per degree day than running it on a schedule. (Degree days are the fair way to compare heating use across mild and bitter weeks.) There is a wrinkle, and it matters: deep overnight setbacks let the screed cool far enough to risk cracking it. So the winning strategy is a gentle schedule with modest setbacks, which keeps the saving and spares the floor.

The same two years of readings measured the house itself. It loses about 450 watts for every degree the outside temperature drops, a figure no spec sheet or EPC estimate can give you. The data also ranked the rooms by how much heat they leak, which turned “we should probably do something about the windows” into a shortlist with the worst offender at the top. When money next goes on glazing or curtains, it goes where the evidence points.

The honest column

Full disclosure: the client on this one is the consultant’s wife, and the house is the one we live in. She got the same treatment a paying client would, including the fixed scope agreed up front, and the guard rails on the heating controls exist for exactly the reason you’d guess.

The other honesty: 10–12% is a share of the gas the heating burns, and every house will differ. A modern smart thermostat gives you scheduling out of the box. Where this build earns its keep is the pricing of every half hour against the actual tariff, and the measured heat-loss figure no thermostat app will ever show you.

Got a question your bills can’t answer?

The £295 audit goes through your subscriptions, retainers and licences. If we find no savings, it’s free.