Get DataMate free
Method

One week on, one week off

Yes. That is exactly what we do, one metered week with nothing controlled, against one metered week with the control running, read off your own meter. We are not going to talk you out of that description, because it is correct. What follows is why that shape is the strongest test available rather than a trick, and where it stops.

Measure it · Fix it · Scale it · Check how your site compares

How the saving is measured, in two phases

First we prove it on your own cabins. Then we keep measuring it for the rest of the project, a different way. Both animations use worked example figures.

Phase 1First we prove it on your own cabins: one week off, one week on

Seven days before the control goes on is your baseline. Seven days after is the test. Both read straight off the same meter on the same cabins.

Your meter, today
0kWh
Waiting to start
Mode
DataMate
Monitoring only, control off
The box is fitted and watching. It is not changing anything yet.
Before day 1
Monitoring runs until your consumption settles.
Day 8: DataMate becomes AutoMate, control on
Daily kWh, read off the meter
Baseline week, days 1 to 7
changeover
not counted
Controlled week, days 9 to 15
Baseline week
0
kWh, days 1 to 7, DataMate monitoring
Controlled week
0
kWh, days 9 to 15, AutoMate controlling
The difference
0
used with AutoMatesaved
kWh saved: baseline minus controlled, same cabins, same meter. Nothing scaled, nothing modelled.
You see this at the Savings Review and decide with a measured number in front of you.
Example figures, drawn to explain the method.

Phase 2Then AutoMate calculates the saving all project long, hour by hour

The proof week is done and the control stays on, so there is no week off to compare against. Instead, every hour of every circuit is compared with what the same kind of circuit used at the same outside temperature, in or out of working hours, on other Gaia sites without control. Those matching hours, averaged, are what yours would have used. The difference is the saving the portal reports, hour by hour, for the whole project.

Your hour, Tuesday 08:00
Drying room heater4°C outsideWorking hours
1.1kWh
AutoMate on
Read off your meter, this hour
Sites checked
0
Hours checked
0
Matching hours found
0hours, same three things
Same hour, elsewhere
Drying room heater4°C outsideWorking hours
3.6kWh
DataMate, no control
With AutoMate
0.0 kWh
Without
0.0 kWh
The difference, this hour
0.0kWh saved
used with AutoMatesaved
Same circuit. Same weather. Same hours.
Repeated for every hour of every circuit, all project. Nothing is switched off to prove it. Example figures, drawn to explain the method.

Phase 1 is free. Phase 2 is what you pay for.

Want to see what this would look like on your project?

Put DataMate on your site, free

or check how your site compares first

That is the method, and here is why

The proof of a saving is one metered week against one metered week. Same meter. Same cabins. Same circuits. Same site. Same tariff. Seven days each.

Both figures are read straight off the meter. Nothing is scaled and nothing is normalised, because two seven-day periods need no arithmetic to make them comparable; they already are.

That is the whole argument for doing it this way. Every step of arithmetic between the meter and the figure is a step somebody can argue with, and this comparison has none. A scaled baseline, a seasonal adjustment, a modelled “expected” consumption; each of those is a place a number can be nudged. There is nowhere to nudge two equal weeks off the same meter.

The dates are set by a rule, before anyone has seen a result

The changeover date is agreed and logged at the Wastage Review with you. The two periods then fall out of a fixed rule rather than a choice:

Baseline week: the seven days ending the day before the control goes on. Control week: the seven days beginning the day after. The changeover day itself counts towards neither.

The changeover day is excluded because the control goes on part-way through it, at no fixed time, so it is not a clean day either side of the line.

So the comparison covers 14 of 15 consecutive days. We state that rather than leave a day unexplained in the middle of the evidence.

The baseline week is a settled week, not a warm-up week

A cabin’s consumption while a site is still manning up tells you nothing useful, so we monitor from install until consumption has settled, in practice usually two to six weeks, depending on site patterns. Only then is a changeover date agreed.

The rest of that monitoring record is not wasted. It is what shows the baseline week was a normal week for your site rather than a convenient one.

What one week against one week cannot do

This is the limitation, and you should have it from us rather than find it.

Two weeks sitting next to each other share their season, their trade mix and roughly their occupancy. They do not share their weather. A colder control week, or a milder baseline week, moves the figure, and one week against one week is more exposed to that than a longer baseline would be.

So, plainly: no weather correction is applied to the comparison, and degree days are not supplied. We do not adjust the result for how the weather behaved.

We do capture outdoor temperature, currently from the local weather station. It is used to group like conditions together when we work out what a saving is worth, which is the next half of this page. It is never used to scale the result of the two weeks. We group; we do not correct.

If you want the longer view, that is what the case studies are for. Those are whole-period results across months or years of the same site’s meter data, and they are a different measurement from the two weeks described here. See the measured results.

Proving the control works is not the same as pricing it

The two weeks answer one question: did consumption fall. They do not answer the second question: what is an hour of switching something off actually worth. That is separate arithmetic, and it is where an energy saving usually gets inflated.

We value a saving at what the kit actually drew, never at its rated power

When one of our rules switches a load off during a period it would otherwise have been running, the saving we credit is:

the time it was off × the measured average draw for that kind of load in those conditions

Never its rated power multiplied by hours. A heater cycles. Its rating is what it pulls at full call, and its real average across a week is materially lower. Rated power multiplied by hours would hand us a far bigger number than the truth.

Like for like: three things have to match

An average is only worth anything if it came from comparable hours. Every average we hold is keyed on three things:

1. What kind of circuit: a convector heater, an air conditioning unit, a dehumidifier, a water heater.
2. The outside temperature, banded by temperature rather than by time of year; a 4 °C day in October and a 4 °C day in March make the same demand on a heater, and the calendar does not know that.
3. Whether it was inside or outside working hours, which we fix by the clock: weekdays, 6am to 6pm.

The third carries as much weight as the temperature does. A 9 °C Monday morning behaves nothing like a 9 °C Sunday evening, and temperature on its own cannot tell them apart.

We can tell you how many measured hours sit behind any figure we give you. Where a particular combination is thin, we say so rather than let the number stand unqualified.

The divisor is where the honesty is

That average is taken over every hour in the band, including the hours the load drew nothing. Not over the drawing hours only.

That is what accounts for the hours a thermostat would have had the load off anyway. Those idle hours sit in the divisor, so they pull the average down, and the saving down with it.

An illustration of the arithmetic. Not a result from any site.

Take 100 hours of one kind of load in one set of conditions. The heater drew 2 kW in 60 of those hours and nothing in the other 40. Divide by the 60 drawing hours and the average comes out at 2 kW. Divide by all 100 hours, which is what we do, and it comes out at 1.2 kW. Same reality. We credit ourselves at the lower one.

Dividing by the drawing hours alone would credit us for hours the load would have been off regardless. That is precisely the error this method exists to avoid.

It makes our number smaller on every load, every week, and that is deliberate. We would rather under-claim a saving than defend one we cannot evidence.

The proof is your site. The rate comes from everybody’s.

These are two different things and it is worth being exact about which is which.

The two weeks that decide whether the control worked are yours, your cabins, your meter, your circuits, nothing pooled. That is the half of this page above, and it is the part that decides whether you pay us anything.

What an hour of switching-off is worth is drawn from everything we monitor, matched on those three conditions. Your figures are measured against the largest record of welfare-cabin energy behaviour in UK construction, and it grows every week.

That is the stronger way round, not the weaker one. An average built from a single site is a small number of hours and a shaky figure. Built from every site we monitor, it is the most reliable estimate of what that kit really draws that anyone in this industry has.

One record, every customer. Splitting it per contractor sounds reassuring and produces a worse number: the three conditions already divide the record into many buckets, and adding the customer’s name as a fourth split thins every one of them. It would also measure the wrong thing; labour moves between contractors, and contractors hire cabins from the same small pool of suppliers. What drives consumption is already in the three conditions.

The averages refresh daily, from monitoring hours only. Hours when AutoMate was actively controlling a load never feed the rate a saving is valued at. The rate comes only from hours when nothing was being controlled.

And if we do not know what a load drew, we do not guess. We use that circuit’s own measured average. Failing that, the average for its band. Failing that, we credit zero. We never fall back on an assumed figure to fill a gap.

Two things we do not claim

We do not treat the benchmark as a snapshot taken before the automation goes on. The averages behind every credited saving are refreshed daily as every site we monitor adds to the record. It is neither a one-off measurement nor a static one.

We do not verify savings against your energy bills, and we would be wrong to claim we did. Welfare cabins rarely sit on separate metering, so there is usually no bill isolated to the cabins in the first place. Savings are measured at the circuit level, in kWh, against your own site’s meter.

Other ways to start

  1. Recommended first step

    Get DataMate free on your next project

    We monitor your cabins until consumption settles, and you get the wastage review.

    Your cost: coordination, and one engineer at the board.

    Get DataMate free

  2. Check how your site compares

    Answer a few questions about your welfare setup, and see whether it looks higher-risk than comparable monitored sites.

    Your cost: 20 seconds and an email address for the result.

    Check how your site compares

  3. Talk about full AutoMate

    For teams ready to standardise cabins across projects.

    Your cost: a 30-minute call.

    Book an AutoMate call

What we don’t publish, and why

The rule-level logic stays private: how each rule decides, its thresholds, the width of the condition bands, and how a saving is attributed to an individual load. Not because a customer should not see it, but because our competitors read this website too.

What we do commit to: if you are a customer, you can ask us to walk through any individual saving on your own site, and we will. The method above is what we hold ourselves to. The arithmetic on your site is available to you, line by line.

Get DataMate free Read the full methodology