Why Your Energy Comparison Is Wrong (And How to Fix It)

Few business rituals inspire as much misplaced confidence as the monthly energy report.

Across the UK, thousands of energy managers will this month compare their February bills to January and declare victory. Consumption down 8 per cent. Costs falling. Progress.

Except February has three fewer days than January. That “saving” of 8 per cent is actually an increase in daily consumption. The numbers lied. Sweet dreams are made of this.

This calendar trap catches organisations every year. The difference in days between months creates apparent swings of up to 10 per cent before anyone has touched a thermostat or switched off a light. January to February looks like a win. February to March looks like disaster. Neither tells you anything useful about efficiency. Time after time, the same mistake gets made.

If your February consumption dropped 8 per cent compared to January, you haven’t saved anything. You’ve actually used more energy per day.

For commercial and industrial sites, the distortion runs deeper. What matters isn’t calendar days but working days. Everyone’s working for the weekend, after all, and those weekends don’t consume like Tuesdays do.

Consider March and April 2023. March had 23 working days. April, starting on a Saturday and losing two days to Easter, had just 18. That’s a five-day difference, or 21.7 per cent, between consecutive months created entirely by how the calendar fell.

April 2023 was the perfect storm. It started on a Saturday, losing weekend days at the start. Easter fell in April, losing two bank holidays. And there were only 30 days to begin with. Any comparison between March and April that year without adjustment is meaningless. The same pattern will repeat in 2028.

Easter itself presents a particular headache. The holiday moves by up to a month from year to year, and energy managers still haven’t found what they’re looking for: a consistent baseline.

Comparing April 2024 to April 2025? One has no Easter holiday, the other loses a long weekend. The “year on year” analysis beloved of boardroom presentations falls apart. Other shifting events create similar problems: school half-terms vary by region, and Chinese New Year moves for global operations.

Then there’s weather. Gas consumption for heating correlates so strongly with temperature that the relationship approaches mathematical certainty. Heating Degree Days, a measure of how cold conditions were, typically explain 85 per cent or more of the variation in gas bills. When the heat is on, the meter spins.

The relationship is nearly linear. A “bad” gas month might just be a cold month. A site using 45 MWh of gas in January and 34 MWh in March hasn’t necessarily improved anything. January was simply colder. Without weather normalisation, you cannot separate efficiency from climate. If your consumption doesn’t show at least an 85 per cent correlation with degree days, you may have genuine efficiency issues. Or data quality problems.

Manufacturing adds another layer of complexity. A factory’s July electricity bill will look impressively low. The summer shutdown explains why. December’s reduced consumption isn’t seasonal goodwill but fewer production days.

A site producing 15,000 units in June and consuming 102 MWh looks far less efficient than July’s 62 MWh, until you notice July only produced 8,000 units. Without normalising for output, the data deceives. The same principle applies across sectors. Offices should measure kWh per occupied desk or full-time equivalent. Retailers should track kWh per transaction or per pound of revenue. Warehouses might use kWh per pallet moved. The right denominator depends on what actually drives your consumption.

The solution is straightforward in principle if tedious in practice. Raw kilowatt hours per month must become kilowatt hours per working day, or per degree day, or per unit produced. Pick the metric that matches your business. Apply it consistently. Document your methodology for the auditors who will inevitably ask.

The February “saving” vanishes when you do the maths properly. So does the March “spike.” What emerges instead is the true picture: how your building or process actually performs, stripped of calendar noise and weather effects.

For ongoing monitoring, consider Cumulative Sum analysis, sometimes called CUSUM. This technique tracks deviations from expected consumption based on your normalisation model, separating real changes from statistical noise. A genuine efficiency improvement shows up as a sustained trend. A cold snap or a busy month washes out.

Life moves pretty fast. If you don’t stop and normalise your data once in a while, you could miss it.

The next time someone announces that energy consumption is up 5 per cent, the only sensible response is a question…

Compared to what?