Energy Analytics & Benchmarking

The first move is a metric you can compare. Raw annual kilowatt-hours tell you almost nothing - a big building should use more than a small one. Energy use intensity (EUI) fixes this by dividing annual energy by floor area, giving energy per square metre per year (or per square foot in the US). Now a school and a shopping mall, a large office and a small one, become comparable on the thing that matters: how intensely each uses energy for its size.

EUI is powerful but must be used honestly. It should be normalised for the things a building cannot help: climate (a hot-humid city needs more cooling than a temperate one - hence weather-normalised or degree-day-adjusted EUI), operating hours (a 24/7 hospital versus a 9-to-5 office), and use type (a data-dense trading floor is not a warehouse). Compare like with like, or the number lies. Used well, EUI is the single most useful energy figure a building has: one number, fairly normalised, that instantly says whether there is a problem worth chasing - and against which every efficiency effort can be tracked over time.

A subtlety worth carrying: there are different flavours of EUI, and mixing them causes endless confusion. Site EUI counts the energy delivered to the building; source EUI also counts the energy lost generating and delivering it, so a mostly-electric building and a mostly-gas one compare very differently on the two measures. There is also the question of whether you count only the base building or tenant loads too. None of this is academic: a benchmarking claim is meaningful only if everyone is using the same definition. When you see an EUI, the disciplined reflex is to ask site or source, and what is inside the boundary? - because two honest analysts can otherwise report very different numbers for the same building and both be right.

An EUI on its own is just a number; benchmarking gives it meaning by comparing it against a reference. There are two kinds. External benchmarking compares your building to peers - similar buildings in a similar climate and use - to see where you sit in the pack. The best-known tool is the US EPA's ENERGY STAR Portfolio Manager, which scores a building from 1 to 100 against a national dataset of comparable buildings, so a score of 75 means you outperform 75 percent of your peers. Many jurisdictions now mandate exactly this kind of benchmarking and public disclosure for larger buildings. Internal benchmarking compares a building to itself over time, or one building in a portfolio to another - often the fastest way to find the outlier worth investigating.

Benchmarking's value is triage at scale: across a portfolio of a hundred buildings you cannot audit them all, but a benchmark instantly ranks them so you attack the worst performers first. Its danger is unfair comparison - benchmark a lab against offices and you will chase a problem that is not there, or miss one that is. Always ask what peer group you are being measured against, and whether it truly resembles your building. A benchmark is a starting question - why are we worse than our peers? - not a final verdict.

A single annual number cannot tell you when energy is wasted; for that you read a load profile - energy plotted against time, typically over a day, a week and a year. The shape is enormously revealing. The overnight base load is the clearest tell: what is a building drawing at 3 a.m. when it is empty? A base load that is a high fraction of the daytime peak usually means plant, lighting or equipment left running when nothing needs it - among the most common and cheapest waste to fix. The morning ramp shows how early plant starts and whether it is optimised. Weekend and holiday profiles that look like weekdays reveal schedules that ignore the calendar. And the peak matters for cost, since demand charges often price the single highest half-hour.

Reading load profiles is a skill worth practising because it turns an abstract efficiency problem into a concrete, findable one. You are not told the building is inefficient; you see the flat overnight line that should drop, the ramp that starts three hours too early, the Sunday that looks like a Monday. Each shape points to a specific, usually cheap intervention - and each is invisible in the annual total. Load-profile analysis is where energy analytics stops being a scorecard and becomes a treasure map.

The hardest and most important question in energy analytics is: did that efficiency measure actually work? The naive answer - last year's bill minus this year's - is almost always wrong, because everything else changed too: the weather, the occupancy, the hours, the tariff. A mild winter can masquerade as a brilliant retrofit; a heatwave can bury a real saving. Measurement and verification (M&V) is the discipline that separates genuine savings from noise.

The core idea is a counterfactual baseline. You build a model - often a simple regression of energy against weather and occupancy from before the change - that predicts what the building would have used had you done nothing. Then the saving is the gap between that projected baseline and what you actually metered, over the same conditions. This is what internationally recognised M&V frameworks formalise, and it is the basis of credible energy-performance contracts, where a contractor is paid from verified savings. The lesson for anyone making an efficiency claim is discipline: a saving you cannot verify against a fair baseline is a story, not a result. Honest M&V is what lets analytics claims survive scrutiny - from a finance director, an auditor, or your own future self checking whether the money was well spent.

None of the four moves works without data of the right shape, and that is a metering decision - one usually made at design time and expensive to retrofit. Two dimensions matter. The first is granularity in time. A single annual figure gives you an EUI and nothing more; you cannot read a load profile from it. To see the overnight base load, the morning ramp and the weekend that looks like a weekday, you need interval data - typically half-hourly or finer, which modern smart meters provide as standard. The finer the interval, the more waste becomes visible: fifteen-minute data reveals patterns that monthly bills bury completely.

The second dimension is granularity in space. A whole-building meter tells you the building is heavy but not where the energy goes. Sub-metering - breaking consumption out by system (HVAC, lighting, plug loads, lifts), by floor, or by tenant - is what turns a vague we use too much into a findable the third-floor plug load never drops at night. Sub-metering also makes M&V far cleaner, because you can baseline and verify the specific system you changed instead of hunting a small saving inside a large, noisy whole-building total.

There is a real trade-off: more meters mean more cost, more data to manage and more points to keep calibrated and correctly named - the data-quality foundation from lesson one returns here with force, because a mislabelled sub-meter is worse than none. The craft is to meter where the decisions are: enough granularity to find and verify the waste that matters, not a sensor on every socket for its own sake. For an architect or engineer, designing sensible sub-metering into a building is one of the quiet, high-leverage moves that makes every future energy analysis possible - and defensible.