All articles

Office occupancy and energy use: what kWh/m² does not show

Measure your offices’ energy use against occupancy: kWh per occupied desk, the baseload of empty floors and what UK rules (MEES, DEC, ESOS) do and don’t say.

Adrien Lafond
26 min read
Illustration: on the left a few silhouettes working under warm lamps, on the right empty dark desks while the ceiling air vents above the empty side still blow, marked by small red swirls

Since 1 April 2023, every let commercial building in England and Wales must reach at least an EPC E rating under the Minimum Energy Efficiency Standards (MEES)[1][1]RegulationGOV.UK, “Non-domestic private rented property: minimum energy efficiency standard – landlord guidance”, EPC E minimum, exemptions and PRS Exemptions RegisterSee in sources ↓. That rating models the building, not its use. Yet your floors empty out on Mondays and Fridays. The building’s energy curve, for its part, does not move.

In an office building, tracking energy use against occupancy, floor by floor, reveals what kWh/m² hides. It measures what the building consumes for nobody.

Office energy use and occupancy: benchmarks and measurement

What is the average energy use of an office building in kWh/m²/yr?

There is no single figure. CIBSE TM46 gives a general office a benchmark of 120 kWh/m² a year of fossil-thermal energy[2][2]Studies and reportsUK Government, EED regulations benchmark heat demand paper citing CIBSE TM46, general office at 120 kWh/m² a year fossil-thermal, 55% degree-day correctedSee in sources ↓, and the Better Buildings Partnership puts typical practice for an air-conditioned office between 107 kWh/m² (all-electric) and 178 kWh/m² (119 electricity + 59 fuel). Neither says how full the building is: average desk utilisation in the UK is now around 30%, against 44% before the pandemic[3][3]PDFStudies and reportsBritish Council for Offices, “Review of Post-Pandemic UK Office Utilisation”, average desk utilisation of about 30% (44% before the pandemic)See in sources ↓. Space heating, cooling and the split between electricity and gas explain most of the gaps.

This ratio says nothing about occupancy: a full office and an empty one show the same figure. That is why the ISO 50001 standard asks you to normalise consumption by the factors that make it vary, in offices as in every non-domestic building. Those factors are identified during the energy review.

107, 120 or 178 kWh/m²/yr: why the benchmarks diverge

These benchmarks do not contradict each other. They differ first in scope: which end uses, which fuels, which floor area (gross internal or net lettable). They then differ in sample and period: REEB’s 2025 benchmarks cover 2022/23 to 2024/25, years when offices were still filling back up, and building occupancy tracked by Remit Consulting only reached 44.2% in February 2026, its highest since May 2021[4][4]Studies and reportsCluttons, “Commercial Quarterly Examiner: office market update Q1 2026”, Remit Consulting building occupancy of 44.2% in February 2026See in sources ↓. The size and age of a sample weigh on its average.

Location matters too: BCO finds London desk utilisation about 7 points below the regions. Last gap, weather: TM46 treats only 55% of its fossil-thermal benchmark as degree-day dependent[2][2]Studies and reportsUK Government, EED regulations benchmark heat demand paper citing CIBSE TM46, general office at 120 kWh/m² a year fossil-thermal, 55% degree-day correctedSee in sources ↓, and the base temperature itself is a choice, best set building by building by regression[5][5]Studies and reportsDegree Days.net, degree-day data and guidance, choosing the base temperature building by building, by regressionSee in sources ↓. A year-on-year comparison without that correction mixes up climate and performance.

Practical conclusion: a national benchmark tells you where you stand, not how to run the building.

Heating, cooling, lighting, office equipment: how the energy splits

UK benchmarks split energy by fuel rather than by end use. For an air-conditioned office heated with fossil fuel, REEB gives typical practice of 119 kWh/m² of electricity and 59 kWh/m² of fuel, and good practice of 85 + 29[6][6]Organisations and institutionsBetter Buildings Partnership, “2025 Real Estate Environmental Benchmarks” (REEB), office benchmarks in kWh/m² GIA a yearSee in sources ↓. NABERS UK, for its part, rates the energy performance of an office in operation, from 1 to 6 stars[7][7]Studies and reportsSpecification Online, “CIBSE: NABERS UK Energy for Offices ratings recognised by the UK Net Zero Carbon Buildings Standard”, 1 to 6 star operational ratingsSee in sources ↓: the scope is not the same.

Domestic hot water remains marginal in an office. The energy balance is shared between electricity, natural gas and heat networks, depending on the plant. With heat pumps, electricity consumption absorbs space heating.

Remember this: heating and HVAC are the end uses that follow occupancy the least.

What the occupancy rate of an office building covers

In an office building, the occupancy rate covers two ratios. The spatial rate compares occupied desks with available desks. The time-based rate compares occupied hours with working hours.

A floor can show high daytime occupancy on Tuesday and none on Friday. Space efficiency and real-estate performance are read on these two ratios. The BMS knows neither of them.

How do you measure the real occupancy of a building?

No need to create a new system: most sites already have four usable sources.

  • access-control badge data
  • desk bookings for hot-desking
  • presence detectors reporting to the BMS
  • AHU CO2 sensor, read as a control indicator

Badges give the spatial occupancy rate, bookings anticipate it and detectors track daytime occupancy floor by floor. The finding on site keeps repeating: occupancy data already exists, it is never connected to the BMS. So the building management system keeps heating an empty office.

Is there an ideal occupancy rate for offices?

There is no ideal occupancy rate for offices. A high rate with hot-desking improves space efficiency. Its energy impact stays at zero if the BMS still heats the whole floor area.

The indicator that matters in office buildings is the gap between occupancy and consumption: two curves that should move together and that nobody overlays.

Does remote working lower office energy use?

Not automatically. On the day an employee works from home, their desk equipment switches off. Space heating, ventilation and corridor lighting still follow time schedules, not badges. Total consumption drops only slightly. Energy used per person present, on the other hand, climbs.

In other words, remote working does not lower the bill as long as the BMS runs on fixed schedules. The energy bill counts every kWh, useful or not. So how much do your floors consume for nobody?

Relating consumption to occupancy: the kWh-per-occupant method

Why relate energy consumption to occupancy?

kWh/m² is an owner’s unit: it stays stable when the building empties, because square metres do not work from home. It judges the energy performance of an envelope, not of a use.

Related to occupancy, the same meter shows what the building consumes for nothing. The logic is the one already applied to weather: heating is compared from one year to the next through degree days, calculated in the UK on a 15.5 °C base for most buildings[8][8]PDFStudies and reportsCarbon Trust, “Degree days for energy management”, 15.5 °C base for most buildingsSee in sources ↓. Occupancy deserves the same normalisation, in all non-domestic buildings (education, healthcare) as in industry working shifts.

The same building at 100, 60 and 40% occupancy: two readings of the same bill

Take one office building controlled on fixed time schedules, with unchanged annual consumption. The table below sets specific consumption per square metre against a consumption index per occupied desk: switch from one reading to the other to see which of the two reacts to empty desks.

One bill, two readings: per m² and per occupied desk Office building controlled on fixed schedules · annual consumption unchanged from one scenario to the next
Per occupied desk reading: the index rises from 100 to 250. Empty desks are carried by everyone present.
Occupied deskskWh/m²/yrIndex per occupied deskFoobot reading
100%Full floors 178 100base 100 The building works for the people in it.
60%2 desks out of 5 empty 178 167100 / 0.6 Two desks out of five are heated and ventilated for nobody.
40%3 desks out of 5 empty 178 250100 / 0.4 Each person present carries the energy of 2.5 desks.
Reading hypothesis, not a measurement: control on fixed schedules, identical annual consumption in all three scenarios. Benchmarks: 178 kWh/m²/yr is REEB typical practice for an air-conditioned office heated with fuel (119 electricity + 59 fuel) [6]; TM46 general office at 120 kWh/m²/yr fossil-thermal [2]; average UK desk utilisation of about 30% [3].

The per-m² ratio does not move, the per-occupied-desk ratio soars. This gap has a name, the unoccupied baseload: it shows on the electricity load curve, then is neutralised by normalisation.

Gathering the data: bills, half-hourly data, meters and BMS trend logs

Four sources, four traps. The energy bill gives the monthly spend, with no hourly profile. The supplier’s half-hourly load profile details electricity consumption, never natural gas or heat networks.

Sub-meters have rarely been checked since handover. On many sites, we find meters nobody can say are reliable. BMS trend logs, finally, have their gaps, disconnections and duplicates.

On the obligation side, the energy audit is set by the Energy Savings Opportunity Scheme (ESOS): large undertakings, with 250 or more employees, or turnover above £44m and a balance sheet above £38m, assessed at group level, carry out an audit every four years, with Phase 4 notification due by 5 December 2027. A certified ISO 50001 system covering total or significant energy use is an alternative compliance route[9][9]RegulationGOV.UK, “How to comply with the Energy Savings Opportunity Scheme (ESOS) Phase 4”, qualification criteria, deadlines and ISO 50001 routeSee in sources ↓.

How do you calculate a kWh-per-occupant indicator?

A kWh-per-occupant indicator divides the energy consumed over a period by an occupancy denominator. That choice determines the quality of your energy performance indicators, far more than specific consumption per m².

The occupied desk is the most telling, provided badge data is reliable. The occupied hour is finer, but requires an hourly energy history. Headcount stays stable but is blind to remote working; SECR reports ask for at least one intensity ratio[10][10]Organisations and institutionslegislation.gov.uk, Companies (Directors’ Report) and Limited Liability Partnerships (Energy and Carbon Report) Regulations 2018, SI 2018/1155 (SECR), intensity ratioSee in sources ↓, so choose it knowingly.

Our choice: occupied desk for reporting, occupied hour for controlling setpoints and optimal start.

Reading the unoccupied baseload on a load curve

The unoccupied baseload is the energy an office consumes when nobody is there. On an hourly electricity curve, it shows in four places:

  1. weeknights, once daytime occupancy is over
  2. weekends, from Friday evening to Monday morning
  3. low-attendance days, Monday and Friday
  4. the setback mode that is not one, on a floor set to setback whose equipment is still running

Why does the BMS heat empty floors, and how do you cut office energy use?

A typical week: the occupancy curve against HVAC consumption

Overlay occupant presence and HVAC energy consumption over the same week: that is what the diagram below does, for an office building controlled on fixed time schedules. Daytime occupancy changes from one day to the next, consumption hardly at all. Click each numbered marker to see where the gap sits.

Typical week · office building on fixed schedules

What the BMS heats while the floors empty out

Relative level 1 2 3 4 5 Schematic profile, not measured.

In one sentence: the occupancy curve moves, the HVAC curve does not. On Monday and Friday, space heating runs for half-empty floors at the same rate as on Wednesday. Four mechanisms explain this gap: global optimal start, constant-volume AHUs, frozen time schedules and forgotten manual overrides. None of them is inevitable.

Time schedules and setpoints set for a full building

Same setpoint on a sparse Monday as on a full Thursday, same time slots in August as in January. Time schedules never reviewed since handover still dictate space heating.

The temperature setpoint ends up aligned with the most demanding occupant (the director on the top floor who wants 24-25 °C in their office all year round, seen many times). As for domestic hot water, it is kept at temperature continuously, far from any low-consumption logic.

Manual overrides set one winter and never released

One January morning, a VIP complaint comes in: the executive floor is cold. The technician forces the AHU and the fan coil units on to put out the fire.

These manual overrides are never released. Nobody records who set them or why, and no review of the building services lists them.

Degraded mode becomes the norm. The energy bill records it silently, month after month, with no identifiable line of expenditure.

The AHU is not to blame. The real culprit is how it is run.

What remains is to tackle these mechanisms in the right order, starting with the one that costs almost nothing.

Matching heating and cooling to real occupancy

Once the BMS is reliable, it becomes possible to match HVAC to real occupancy. Floor-by-floor control, demand-controlled ventilation, setpoints differentiated by attendance day: each action aims to cut the consumption of empty areas without penalising those in use.

Office heating and cooling benefit directly. A heat pump whose loop temperatures follow the real need gains in efficiency. The energy savings then come from equipment already in place.

Lighting and office equipment: when a presence detector is enough

In a typical air-conditioned office heated with fuel, electricity accounts for 119 kWh/m² a year against 59 for fuel[6][6]Organisations and institutionsBetter Buildings Partnership, “2025 Real Estate Environmental Benchmarks” (REEB), office benchmarks in kWh/m² GIA a yearSee in sources ↓. Lighting and office equipment are part of that electricity and respond to simple levers: scheduling, switch-off at night and at weekends, involving every employee in an energy-saving plan.

We are sometimes asked whether we put AI on lighting. The answer is no: if a presence detector is enough, there is no point adding AI. To cut the electricity used for lighting, well-kept scheduling does most of the work. Here, the energy saving comes down to operational discipline.

What UK rules say (and don’t say) about occupancy

Asset rating or operational rating: how UK offices are measured

MEES applies to let non-domestic buildings in England and Wales[1][1]RegulationGOV.UK, “Non-domestic private rented property: minimum energy efficiency standard – landlord guidance”, EPC E minimum, exemptions and PRS Exemptions RegisterSee in sources ↓. The minimum is an EPC E rating, for new lettings and renewals since 1 April 2018 and for all let properties since 1 April 2023[1][1]RegulationGOV.UK, “Non-domestic private rented property: minimum energy efficiency standard – landlord guidance”, EPC E minimum, exemptions and PRS Exemptions RegisterSee in sources ↓. An EPC is an asset rating: it models the building and its systems, whatever the number of people using it.

The other approach is an operational rating, based on metered consumption. The only mandatory one is the Display Energy Certificate (DEC), for buildings over 250 m² occupied by a public authority and frequently visited by the public[12][12]Regulationlegislation.gov.uk, SI 2012/3118, regulation 14, Display Energy Certificates for buildings occupied by public authoritiesSee in sources ↓. It is valid 12 months above 1,000 m² and 10 years below, with a recommendation report[12][12]Regulationlegislation.gov.uk, SI 2012/3118, regulation 14, Display Energy Certificates for buildings occupied by public authoritiesSee in sources ↓. Neither sets a reduction target: the UK has no legal obligation to cut measured energy use in existing commercial buildings, with or without an ISO 50001 energy management system.

Two other texts complete the picture. Approved Document L requires a building automation and control system (BACS) in new non-domestic buildings whose heating or air-conditioning output exceeds 180 kW, and where a BACS is installed or replaced in an existing building above that output[13][13]RegulationHM Government, “Approved Document L, Volume 2”, 2021 edition incorporating 2023 amendments, §6.66-6.67, BACS above 180 kWSee in sources ↓; the 2026 edition, in force from 24 March 2027, keeps the 180 kW threshold[14][14]RegulationHM Government, “Approved Document L, Volume 2”, 2026 edition (Future Buildings Standard), §5.76-5.77, in force on 24 March 2027See in sources ↓. Neither requires existing buildings to be equipped by a deadline.

Does any UK rule take intensity of use into account?

No regulatory adjustment by occupancy exists in the UK. Operational ratings do reflect intensity of use, because they start from actual consumption: the DEC is based on metered energy[12][12]Regulationlegislation.gov.uk, SI 2012/3118, regulation 14, Display Energy Certificates for buildings occupied by public authoritiesSee in sources ↓, and NABERS UK rates offices in operation, on a voluntary basis. For offices, that is where use shows. Occupancy enters the picture through operational data, not through a regulatory target.

Any operational rating therefore relies on accurate occupancy data. The question is who provides it.

Six questions to ask your O&M contractor before committing to works

You do not need in-house BMS expertise to test a control strategy. Here are the six questions to ask your O&M contractor before signing a quote: tick those that remain without a precise answer, and the reading appears below the list.

Before signing a quote

Six questions to ask your O&M contractor

Tick each question that remains without a precise answer.
0 questions without a precise answer out of 6
Your reading will appear here Tick the questions left without a precise answer.

One vague answer does not condemn an office. Several ticked boxes, on the other hand, show that before any works, you need to cut the consumption of empty floors with what is already there. The last question, on the baseline year, determines any proof of energy savings and any return-on-investment calculation. These six answers are also our starting point.

What Foobot does in offices with variable occupancy

BMS advisory: getting the BMS in order before any optimisation

Our BMS advisory starts from the answers to these six questions. With the O&M contractor, we release the manual overrides one by one. We then make the meters reliable: disconnections, data gaps, duplicates. Sequences are reset floor by floor.

This foundation stands on its own, with no AI phase behind it. A building made reliable finally makes its energy balance readable. For many portfolios, it is the only useful energy-efficiency solution in the short term.

AI Optimal Start: a start time recalculated every day, zone by zone

Once the BMS is reliable, AI Optimal Start replaces global optimal start at a fixed time. Every day, it recalculates when each zone must start heating up. It takes into account thermal inertia, the weather and expected occupancy.

An office that is empty on Friday is no longer heated from 6 am for nobody. Space heating starts neither too early nor too late. The energy saving comes from that daily precision.

AI Control: regulating HVAC every 15 minutes on expected occupancy

AI Control only comes in if complexity justifies it: hot and chilled water loops, heat pump, several AHUs. A digital twin is trained on BMS trend logs, the weather and the occupancy of each office. The AI then regulates HVAC every 15 minutes, 24/7.

On the Valeo site, the case study shows -28.6% on heating and electricity. Savings are measured following the IPMVP protocol with degree-day correction[11][11]Standards and protocolsEVO, “International Performance Measurement and Verification Protocol (IPMVP)”, baseline adjustment and weather correctionSee in sources ↓. Avoided greenhouse gas emissions then enter carbon reporting on a verifiable basis.

Measuring what your offices consume when nobody is working there

Three habits are enough to see what kWh/m² hides. Relate consumption to real occupancy, per occupied desk or per occupied hour. Read the unoccupied baseload, starting with Sunday night. Keep to the order “make reliable, then optimise” before committing to any investment.

To compare two years that no longer look alike, our article on the baseline and degree days sets out the method. To start this tracking, the free Foobot EMS is available online. Before signing a quote, let’s look together at what your floors consume for nobody.

Let’s talk

FAQ - office energy use and occupancy

How much electricity does an office use?

It depends on how it is heated. For a typical air-conditioned office heated with gas or another fuel, REEB benchmarks give 119 kWh/m² of electricity a year, 85 in good practice[6][6]Organisations and institutionsBetter Buildings Partnership, “2025 Real Estate Environmental Benchmarks” (REEB), office benchmarks in kWh/m² GIA a yearSee in sources ↓. An all-electric office, with heat pumps for instance, is benchmarked at 107 kWh/m² in total, heating included.

What are the penalties under ESOS?

Failing to carry out an ESOS assessment can cost up to £50,000, plus £500 per working day for up to 80 working days; failing to notify, £5,000 plus £500 per working day[15][15]Regulationlegislation.gov.uk, Energy Savings Opportunity Scheme Regulations 2014 (SI 2014/1643), Part 8, civil penaltiesSee in sources ↓. These penalties concern the audit, not consumption: in unoccupied hours, it is still up to the BMS to switch to a setback mode that really is one.

What will MEES require of let offices after 2027?

The interim response of 18 June 2026 proposes EPC B from 2031 for let buildings over 1,000 m², where cost-effective; the EPC C milestone for 2027 will not be taken forward, and smaller buildings stay at EPC E[16][16]RegulationGOV.UK, “Minimum Energy Efficiency Standards (MEES) in the non-domestic private rented sector: interim response”, EPC B from 2031 above 1,000 m²See in sources ↓. Secondary legislation is still needed. An EPC upgrade says nothing about how empty floors are run.

Does ESOS require a company to carry out the measures identified?

No. ESOS requires an action plan and progress updates, not works. Since 22 July 2026, the amended regulations require an estimate of savings achieved, in kWh[17][17]Regulationlegislation.gov.uk, Energy Savings Opportunity Scheme (Amendment) Regulations 2026 (SI 2026/701), action plan reviewSee in sources ↓, and a review of the action plan measures not implemented, with the reasons[17][17]Regulationlegislation.gov.uk, Energy Savings Opportunity Scheme (Amendment) Regulations 2026 (SI 2026/701), action plan reviewSee in sources ↓. The audit identifies, the operator acts.

Are leased offices covered by MEES?

Yes. In England and Wales, any let non-domestic property must reach at least EPC E, for new lettings since 1 April 2018 and for all existing lettings since 1 April 2023[1][1]RegulationGOV.UK, “Non-domestic private rented property: minimum energy efficiency standard – landlord guidance”, EPC E minimum, exemptions and PRS Exemptions RegisterSee in sources ↓. The standard falls on the landlord and rests on a theoretical rating, which ignores occupancy.

Is a building management system mandatory in offices?

Only in new buildings, and when a system is installed or replaced: Approved Document L requires a BACS above 180 kW of heating or air-conditioning output[13][13]RegulationHM Government, “Approved Document L, Volume 2”, 2021 edition incorporating 2023 amendments, §6.66-6.67, BACS above 180 kWSee in sources ↓. The 2026 edition, in force on 24 March 2027, keeps the same threshold[14][14]RegulationHM Government, “Approved Document L, Volume 2”, 2026 edition (Future Buildings Standard), §5.76-5.77, in force on 24 March 2027See in sources ↓. The requirement sets functions (continuous monitoring, benchmarking, interoperability); the guidance notes that a Class A system would meet them.

How do you carry out an energy audit of your office buildings?

Under ESOS, large undertakings carry out an audit every four years covering at least 95% of their energy use, unless a certified ISO 50001 system covers total or significant consumption[9][9]RegulationGOV.UK, “How to comply with the Energy Savings Opportunity Scheme (ESOS) Phase 4”, qualification criteria, deadlines and ISO 50001 routeSee in sources ↓. The audit is carried out or reviewed by an ESOS Lead Assessor listed on an approved register[18][18]RegulationGOV.UK, “Energy Savings Opportunity Scheme (ESOS)”, approved registers of ESOS Lead AssessorsSee in sources ↓. In office buildings, make the meters reliable first.

What is the occupancy rate of UK offices?

Building occupancy measured by Remit Consulting reached 44.2% in February 2026, its highest since May 2021, against around 70% before the pandemic[4][4]Studies and reportsCluttons, “Commercial Quarterly Examiner: office market update Q1 2026”, Remit Consulting building occupancy of 44.2% in February 2026See in sources ↓. Desk utilisation is lower still: the BCO puts it at about 30% on average, 40% from Tuesday to Thursday.

What happens if a let office falls below EPC E?

The landlord must not let it unless an exemption applies: 7-year payback, third-party consent or a devaluation of more than 5%. Exemptions generally last five years and must be recorded on the PRS Exemptions Register[1][1]RegulationGOV.UK, “Non-domestic private rented property: minimum energy efficiency standard – landlord guidance”, EPC E minimum, exemptions and PRS Exemptions RegisterSee in sources ↓. MEES is not to be confused with the ESOS penalties for a missing audit.

How does hot-desking affect a building’s energy bill?

Hot-desking only lowers the energy bill if the floor area freed up stops being heated and ventilated. Grouping teams on fewer floors, then genuinely shutting down the floors released: that is where consumption falls. Otherwise, nothing moves.

Sources

Sources and references

18 sources.
Regulation9
  1. [1]GOV.UK, “Non-domestic private rented property: minimum energy efficiency standard – landlord guidance”, EPC E minimum, exemptions and PRS Exemptions Register05/05/2026↑ 1 2 3 4 5
  2. [9]GOV.UK, “How to comply with the Energy Savings Opportunity Scheme (ESOS) Phase 4”, qualification criteria, deadlines and ISO 50001 route30/07/2026↑ 1 2
  3. [12]legislation.gov.uk, SI 2012/3118, regulation 14, Display Energy Certificates for buildings occupied by public authorities2012↑ 1 2 3 4
  4. [13]HM Government, “Approved Document L, Volume 2”, 2021 edition incorporating 2023 amendments, §6.66-6.67, BACS above 180 kW2023↑ 1 2
Show the other 5: [14] to [18]
  1. [14]HM Government, “Approved Document L, Volume 2”, 2026 edition (Future Buildings Standard), §5.76-5.77, in force on 24 March 202724/03/2026↑ 1 2
  2. [15]legislation.gov.uk, Energy Savings Opportunity Scheme Regulations 2014 (SI 2014/1643), Part 8, civil penalties↑
  3. [16]GOV.UK, “Minimum Energy Efficiency Standards (MEES) in the non-domestic private rented sector: interim response”, EPC B from 2031 above 1,000 m²↑
  4. [17]legislation.gov.uk, Energy Savings Opportunity Scheme (Amendment) Regulations 2026 (SI 2026/701), action plan review↑ 1 2
  5. [18]GOV.UK, “Energy Savings Opportunity Scheme (ESOS)”, approved registers of ESOS Lead Assessors02/09/2026↑
Standards and protocols1
  1. [11]EVO, “International Performance Measurement and Verification Protocol (IPMVP)”, baseline adjustment and weather correction↑
Energy newsletter

Get our articles on AI-driven HVAC control

One case study a month, on fine-grained control and its impact on building energy performance.