MeterBrief guides

What overnight baseload really costs, and how to cut it without disrupting operations

2026-09-11

Turn a client's half-hourly export into a branded report in about a minute.

Try it on your data

Most commercial sites are busy for around 50 hours a week and paid for 168. The gap is where overnight baseload lives: the flat or gently cycling demand that sits under the load curve when nobody is on site. It rarely looks dramatic on a bill, because it hides inside the same p/kWh as everything else. Across a year, though, it is often the largest single controllable line item a site has.

What follows sets out what causes overnight load in typical commercial buildings, how to attribute it from half-hourly (HH) data rather than guessing, and what savings are realistic once you respect safety, food hygiene and equipment protection. Rates, levy figures and industry programme dates move, so treat the linked source pages as the authority on current numbers rather than any figure quoted second hand.

Why the money is bigger than people expect

Take a site open 08:00 to 18:00, Monday to Friday. Occupied hours come to roughly 2,600 a year, which leaves about 6,160 closed hours. A 20 kW overnight baseload therefore consumes around 123,000 kWh a year. At a delivered cost of 25 p/kWh, including standing and network charges as well as levies, that is close to £31,000. Trim it by a third and you have found roughly £10,000 with no change to trading hours and no capital project.

A second effect compounds it. Closed-period consumption still attracts Climate Change Levy on every kWh, at the rates published by HMRC in its Climate Change Levy rates guidance, and most business supplies carry VAT at the standard rate unless they fall under the low-usage de minimis rules in VAT Notice 701/19. Overnight kWh are not cheap kWh for a business customer unless the contract has a genuine night rate, and even then network and levy costs still apply.

Typical causes in commercial buildings

The usual suspects repeat across sectors:

  • Air handling units and fan coil systems left on a 24/7 schedule, or on a time clock that drifted after a power cut or a BMS battery failure
  • Heating and cooling setpoints with wide dead bands, frost protection set far higher than needed, and simultaneous heating and cooling in the same zone
  • Refrigeration: cellar cooling, display cases running with lights on overnight, walk-ins with failed door seals or defrost heaters cycling too often
  • Compressed air, where a single 3 mm leak on a loaded compressor can run to several kW continuously; night running with no production is almost always pure loss
  • Comms rooms and server cupboards, plus their dedicated cooling, often oversized and running against each other
  • External, canopy and car park lighting on photocell only, internal lighting on manual switching, and signage running to 6am in an empty retail park
  • Domestic hot water: secondary circulation pumps and immersions running all night for a building with no occupants until morning
  • Catering kitchens: bains-marie, coffee machines, gantries, extract fans and pass-through warmers left energised after close
  • Process machinery in standby, EV chargers idling, vending, CCTV and access control, lift and escalator standby, trace heating

None of this is exotic. What makes it persistent is that nobody is on site to notice, and the daytime peak gets all the management attention.

Attributing overnight load from half-hourly data

Start with the data. Sites settled half-hourly have 48 readings a day available from the supplier or data collector. Smaller non-half-hourly sites with a smart or advanced meter can usually get half-hourly consumption too, subject to the data access rules Ofgem sets out in its smart meter guidance, and the industry-wide move to half-hourly settlement under the MHHS Programme is steadily making granular data the default rather than a favour. Ask for 12 months as a CSV, not a PDF summary.

Then work through a fixed sequence.

Convert units first. A half-hourly figure in kWh is multiplied by two to give average kW in that period. Everything below is easier in kW because it maps onto nameplate ratings.

Baseload also needs an honest definition. Rather than eyeballing the lowest night, take the fifth percentile of the periods between 01:00 and 05:00 for each calendar month. That gives a monthly baseload in kW and strips out one-off events. Plot the twelve values. A flat line across the year means constant equipment. A summer and winter hump means weather-driven plant, so heating or refrigeration reacting to ambient temperature.

The calendar gives you a free experiment. Compare weeknights with Sunday nights, bank holidays and the Christmas shutdown. Whatever still runs when the site has been closed for four days is the irreducible load, and the difference between that and a normal Tuesday night is scheduling, not equipment.

Shape matters as much as size. A clean step up at 22:00 that holds until 06:00 points to a timed circuit or a heater. Regular cycling of similar amplitude is refrigeration or compressor top-up. A slow ramp from 04:00 is optimum-start preheat, which may be entirely correct. Sawtooth patterns with rising frequency over months often mean a growing air leak or a fouled condenser.

Next comes a bottom-up reconciliation. List the assets you believe are live overnight, estimate kW as nameplate multiplied by a realistic duty cycle, and total them. Compare with the measured baseload. If your list explains only 12 kW of a measured 20 kW, you are missing something, and that gap is usually where the saving is. Close it with a night walk-round, plug-in loggers on suspect circuits, sub-meter reads before and after close, and the BMS trend logs. Half-hourly gas data, where available, tells you whether boilers are firing overnight as well.

Last, sanity check the commercial side. Look at the site's agreed capacity in kVA against actual maximum demand, and at how much of the bill sits in fixed network charges that overnight action will not touch. The framework for those charges follows Ofgem's Targeted Charging Review decision, and the detail sits in each distribution network operator's published charging statement.

What level of saving is realistic

For a building that has never had this exercise done, a 15 to 35 per cent reduction in overnight baseload is a fair expectation from scheduling, setpoints, leak repair and out-of-hours shutdown discipline. Sites with a well-tuned BMS and recent controls work will yield less, perhaps 5 to 10 per cent. Anyone promising to remove baseload entirely has not understood the building.

The cheap wins tend to be: correcting time schedules and holiday calendars; fixing compressed air leaks; turning off display and signage lighting after close; reducing hot water circulation to a timed profile; shutting down catering equipment on a written close-down checklist; and dropping unoccupied setpoints to sensible limits. Payback on most of these is measured in weeks.

Equally important is what must not be switched off. Frost protection needs to stay functional to avoid burst pipes. Food storage temperatures must hold, with records intact for environmental health purposes. Hot water temperature regimes are a legionella control, covered by HSE's Legionnaires' disease guidance, so cutting circulation pumps without a risk assessment is not acceptable. Emergency lighting, fire alarms, security systems, lab and medical refrigeration, IT cooling, trace heating and process holding loads all stay. There is also a physical trade-off: heavy overnight shutdown can create a longer morning preheat and a higher morning peak, which matters where capacity charging is a factor. Model the whole 24 hours, not just the quiet part.

Verify afterwards using the same percentile method on the following three months. If the monthly baseload has not moved, the change did not stick, which usually means a manual override or a schedule that reverted.

What a broker can do on a live account

  • Request 12 months of half-hourly data as a CSV at the point of tender, for every site, and keep it on file
  • Calculate the closed-hours share of annual consumption and put it on the front page of the quote pack alongside the unit rate
  • Price the overnight volume at the client's full delivered cost, including standing and network charges plus levies, so the saving is credible
  • Flag the cheapest interventions for that building type and name who on site owns each one
  • Put a re-check in the diary for 90 days after any change, using the same calculation method
  • Where consumption is falling materially, review the contracted volume and any take-or-pay tolerance before renewal, and check agreed capacity in kVA against actual demand

One caveat worth passing on to clients: efficiency work of this kind may also feed into wider compliance obligations such as the Energy Savings Opportunity Scheme, which is a separate topic with its own deadlines.

Turn a client's half-hourly export into a branded report

Baseload, capacity headroom, out-of-hours cost and a client-ready narrative, in about a minute. Edit anything, then send it under your name alone. Free preview on your own data.

Try it on your data