Article 02: Why EPCs Do Not Tell You How Much Energy a Building Uses

A series mining the PhD thesis "London and UK Office Buildings: Investigating Energy Use and Landlord-Tenant Influences" (Azhari, 2025).

Key takeaway. A statistical analysis of every Greater London office with an EPC (2,654 buildings) finds no significant relationship between EPC band and measured energy use. That is uncomfortable for MEES, for ESOS and for any due-diligence process that treats the EPC as a proxy for the bill.

Chart

EUI distributions overlap across every EPC band

Medians (dark bar) cluster between 206 and 258 kWh/m²·yr. IQR boxes and whiskers overlap heavily across all bands. Only D vs E is statistically distinguishable after Bonferroni correction (p = 0.03).

The puzzle in one sentence

If two offices sit side by side, identical in shape, fabric and plant, and one is rated EPC B while the other is rated EPC E, common sense says the B should use less energy. That is what the EPC promises. The Greater London empirical data disagrees.

Across 2,654 office Self-Contained Units (SCUs) in Greater London with both a valid EPC and metered energy data for 2017, median energy use intensity (EUI) varies very little between bands. Each band contains the full range of operating performance, from very low EUI to very high. A Kruskal-Wallis test (the non-parametric equivalent of one-way ANOVA) finds the only statistically significant pairwise difference is between band D and band E, and even that gap is modest. Within and across all other bands the distributions overlap so much that knowing a building EPC tells you almost nothing useful about its actual bill.

Where the gap comes from

The gap is not a bug in EPC data. It is a feature of the methodology, and a deliberate one. EPCs in the UK are produced via the Simplified Building Energy Model (SBEM) for non-domestic buildings and equivalent calculation tools for the domestic sector. The point of the methodology is to compare buildings on the basis of their envelope and services, holding occupancy patterns and operating hours constant. In practice this means that two notional buildings with identical fabric and plant will receive the same rating regardless of whether they are run lights-on for twenty-four hours a day or tightly controlled to weekday working hours. That standardisation is what makes the EPC a useful design-stage and compliance instrument. It is also what makes it a poor predictor of actual energy use.

Put another way, an EPC asks how efficient is this building, assuming standardised use? Operational rating schemes (Display Energy Certificates in the UK, NABERS in Australia) ask how much energy did this building actually consume? These are not the same question, and they cannot be expected to have the same answer.

What the empirical evidence says

The Greater London sample makes the gap concrete. Most rated offices sit in band D (881 buildings), followed by band E (571) and band C (508). The total spread is even more revealing than the medians. Figure 4-17 in the thesis shows each individual office as a single bar of energy intensity, grouped by EPC band, with horizontal lines at the median of each group. A reader expects the medians to step down cleanly from band G to band A. Instead, they sit in a narrow range from roughly 200 to 258 kWh per square metre per annum across bands B to G. Band A holds only four offices in the sample, too few for a meaningful median.

The Kruskal-Wallis result reports a p-value of 0.02 across all bands, which is significant only in the sense that the null hypothesis (all groups equal) is rejected. The post-hoc Dunn pairwise tests with Bonferroni correction then locate the difference precisely. Bands D and E differ. No other pair does. That is a striking outcome on a sample of more than 2,600 buildings. With samples this large, even tiny real differences should surface as statistically significant. The fact that they do not is the empirical finding.

Chart

Within every EPC band, energy use spans from near-zero to 900+ kWh/m²

Each band's offices are sorted left to right by measured EUI. Dashed lines mark the median for each band — all clustering between 170 and 258 kWh/m²·yr. No band separates cleanly from the others.

ABCDEFG

A control group of 775 offices with no EPC at all has a distribution of EUIs that looks remarkably similar to the EPC-rated stock. This is a further sign that the EPC band carries little information about operational performance.

Figure

One significant pair out of 28: D vs E

Kruskal-Wallis across all bands finds p = 0.04. Dunn pairwise tests with Bonferroni correction locate the difference precisely: only one of the 28 possible band pairs is statistically distinguishable.

D vs E

p_Bonferroni-adj = 0.03

The only band pair with a significant difference in measured EUI. Median gap: 39 kWh/m²·yr (band D at 206, band E at 245). The EPC correctly ranks D above E in this one comparison.

27 other pairs

Not significant

No significant pairwise difference across any other band combination. With 1,828 buildings, even a 10 kWh/m²·yr real gap would register. The null result is the finding.

Effect size: 0.8%

ε²_ordinal = 0.008

EPC band accounts for less than 1 per cent of the variation in measured energy use intensity across the Greater London large-office stock.

No EPC: 222 kWh/m²·yr

Median, n = 775

Offices with no certificate have a distribution nearly identical to the rated stock, reinforcing that the band carries little operational signal.

The 2026 National Buildings Database (DESNZ), which the author contributed to, gives the national picture. Only 15 per cent of office premises in Great Britain have an EPC at all, and the mean lodgement date is June 2019. The thesis 44 per cent EPC coverage in Greater London turns out to be the better-covered end of the distribution. Nationally, the EPC is a still less reliable basis for policy, because for 85 per cent of office premises it simply does not exist as a measured data point.

Why this matters for policy

The implications follow quickly and run deeper than they look.

Minimum Energy Efficiency Standards (MEES) regulations in England and Wales are pegged to EPC bands. Properties below EPC E cannot be let, and the policy aim is to push the floor of the rented commercial stock toward band B by 2030. The hope is that pushing landlords up the EPC ladder will deliver real reductions in operational energy and carbon. The Greater London evidence suggests that the ladder is not pointing in the right direction. Upgrading a building from EPC E to EPC C through cost-effective fabric measures (insulation, double-glazing, more efficient boilers) does not, on average, deliver the operational energy savings the band change implies. The compliance has shifted. The bill may not have.

Energy Savings Opportunity Scheme (ESOS) is in a slightly different position. It mandates an energy audit every four years for qualifying large undertakings, with no obligation to act on the findings. The audit content includes EPC data among many other inputs, so the limitations are diluted. Still, in any portfolio-level reporting that leans on EPC averages, the implication is the same. Average band tells you almost nothing about average bill.

Lender due diligence and acquisition underwriting often use EPC as a screening filter for green-loan eligibility. If the EPC and operational performance are weakly correlated in the empirical data, then the screening is making a category error: it can label buildings as low-risk that are not, while penalising buildings that may already be performing well.

Tenant decision-making is the last casualty. Occupiers signing a lease in a B-rated building reasonably expect a smaller energy bill than they would get in a D-rated one. The empirical evidence does not support that expectation strongly. The implications for green-lease drafting and rent-versus-bill negotiations are uncomfortable.

A pragmatic alternative

This is not a case for abandoning the EPC. It is a case for not asking the EPC to do a job it was never designed to do. The two complementary moves are easy to state and harder to implement.

First, pair the EPC with an operational rating. The UK already has the Display Energy Certificate (DEC) for public-sector buildings. NABERS-UK, launched by the Better Buildings Partnership and BRE in 2020 and modelled on the Australian NABERS scheme, is the more sophisticated emerging option for commercial offices. NABERS rates buildings on what they actually use, with a clear distinction between base-building and whole-building ratings. Article 9 examines the Australian experience in detail, including the conditions under which NABERS drove market change that the UK has not yet replicated.

Second, refresh the benchmark scaffolding. CIBSE TM46 and the older ECG-19 office benchmarks underpin the DEC scheme and have not had a meaningful empirical update in three decades. The Greater London empirical data offers a basis for refresh, and the more recent Real Estate Environmental Benchmark (REEB) from the BBP is already a closer reflection of operational reality. Article 10 treats the benchmark question in its own right.

Neither move displaces the EPC. They surround it with the operational evidence it cannot produce by itself.

What the data does not see

A few caveats are worth keeping in view. EPC coverage in the Greater London office sample is incomplete (around 44 per cent of office SCUs), so the comparison group is a subset that may not be perfectly random. EPCs in the sample span multiple vintages of SBEM, which means lodgement years and calculation conventions vary. The upper tail of the band distribution is thin. Band A holds only four buildings, so the sample cannot say much about top performers. The metered energy data is for 2017 alone, so weather, occupancy and operational anomalies for that single year colour the EUI distributions. And finally, absence of a strong correlation does not mean the EPC has no value at all. It is a useful instrument for the design-stage and compliance work it was built for. The argument here is narrower than that: do not use it for a job it was not designed to do.

The next article moves on from one variable to many. If EPC band does not predict operational energy use, how well do the other physical characteristics of a building (its size, its height, its age, its heating fuel) collectively explain how much energy it uses? The answer turns out to be less than you would expect.

Limitations

EPC coverage in the London office stock is limited to around 44 per cent of SCUs, so the comparison is built on a subset that may not be random. EPCs in the sample span multiple vintages of the SBEM methodology, which limits like-for-like comparison. Band A holds only four buildings in the sample, making the upper tail statistically thin. The metered energy data is for a single year (2017), so weather and operational anomalies can colour the distributions. EPC band is one input among many. Absence of correlation with operational use is not absence of value for the original compliance purpose.

References

• Azhari, R. (2025) London and UK Office Buildings Investigating energy use and landlord/tenant influences. Doctoral thesis (Ph.D), UCL (University College London). URL: https://discovery.ucl.ac.uk/id/eprint/10204821/

• Better Buildings Partnership (2019). Design for Performance: a new approach to deliver energy efficient offices. Available at: https://www.betterbuildingspartnership.co.uk/design-performance-new-approach-delivering-energy-efficient-offices-uk

• Better Buildings Partnership (2020). Real Estate Environmental Benchmark: 2019 energy snapshot. Available at: https://www.betterbuildingspartnership.co.uk/our-priorities/measuring-reporting/real-estate-environmental-benchmark

• BEIS (2019). Non-domestic private rented property minimum standard (MEES): landlord guidance. Available at: https://www.gov.uk/government/publications/the-non-domestic-private-rented-property-minimum-standard-landlord-guidance-documents

• BEIS (2021). Energy Performance of Buildings: evaluation reports. Available at: https://www.gov.uk/government/publications/energy-performance-of-buildings-regulations-evaluation

• Cohen, R., Bordass, B., and Field, J. (2018). Achieving operational energy performance ratings: prediction-in-use and the design-performance gap.

• Bordass, B., and Field, J. (2007). Energy benchmarks for non-domestic buildings.

• Evans, S., Fennell, P., Humphrey, D., Liddiard, R., Oraiopoulos, A., Palmer, J., Ruyssevelt, P., Shamsi, H., Amrith, S., and Steadman, P. (2026). National Buildings Database Phase 2. Department for Energy Security and Net Zero. Available at: https://www.gov.uk/government/publications/national-buildings-database-phase-2-understanding-great-britains-buildings

Read next

• Article 9: NABERS for Britain? Lessons from Australia on Operational Rating.

• Article 10: Time to Retire ECG-19? Updating Office Benchmarks With Evidence From 6,000 Greater London Buildings.

About this series

This article is part of a fifteen-piece series adapting the 2025 PhD thesis "London and UK Office Buildings: Investigating Energy Use and Landlord-Tenant Influences" (Azhari, 2025) for a mixed academic and industry readership. The empirical findings draw on the 3DStock model of 6,038 office Self-Contained Units in Greater London with metered energy data for 2017, supplied by BEIS under a data-sharing agreement, alongside the Better Buildings Partnership Real Estate Environmental Benchmark. The qualitative findings draw on semi-structured interviews with seven major UK property organisations, conducted during the 2021 lockdown. Interviewees and their organisations are anonymised by role and organisation type. Please cite the original thesis for academic use.

Author. Rayan Azhari completed his PhD at the UCL Bartlett School of Environment, Energy and Resources in 2025, supervised by Paul Ruyssevelt and Kathryn Janda. The research was supported by the EPSRC Centre for Doctoral Training in Energy Demand (LoLo) and UK Research and Innovation through the Centre for Research into Energy Demand Solutions.

Other articles in the series. Article 1 The 30/85/89 Problem; Article 2 Why EPCs Do Not Tell You How Much Energy a Building Uses; Article 3 Eighteen Per Cent; Article 4 Mapping the Stock; Article 5 Height, Age and the Fuel Question; Article 6 The Split-Incentive Problem; Article 7 Green Leases and Service Charges; Article 8 From 38 to 73 Per Cent Energy Savings; Article 9 NABERS for Britain; Article 10 Time to Retire ECG-19; Article 11 Can London Speak for England and Wales; Article 12 The Hybrid-Work Footprint; Article 13 Why I Used Linear Regression Over Random Forest; Article 14 Vertical Postcodes; Article 15 What Is a Building?