How to compare an EPC score, step by step

What to have ready

1. 1The official certificateHave the property's own EPC in front of you, from the official register. Every figure you enter comes from it.
2. 2The outward postcode district onlyThe first part of the postcode, such as AL1 or M11. Never a full postcode, a street address, a UPRN or a certificate number: none of those are used or needed.
3. 3The recorded property typeThe property type as recorded on the certificate, for example house or flat.
4. 4The construction age bandThe construction age band recorded on the certificate, for example 1930–1949.
5. 5The current score, and potential if you have itThe current energy efficiency score from 1 to 100 is required. The potential score is optional and is also taken from the certificate.

If you do not have the current certificate, first use our guide to finding your official EPC certificate. On the comparison page you enter those details, select Show comparison, and read the four published views that appear. Everything runs in your browser: nothing is submitted or stored. If you are unsure which band a score falls in, the A to G band reference lists every score range.

Worked example with published figures

Postcode district

AL1

Property type

House

Construction age band

1930–1949

Current score entered

60 (band D)

These four counts describe four separate published groups. They overlap, so they must never be added together, and they are not a narrowing hierarchy ending in one matched cohort. No published cell combines district, property type and construction age.

Every figure on this site comes from the published aggregate files in this release: national totals from national.json, district views from the matching postcode-district file, and national type-and-age groups from archetypes.json. Each page draws only on the files relevant to the figures it shows.

Reading the figures without over-reading them

• Median and mean are different measures

  The median is the middle score of a group: at least half the properties score at or below it and at least half score at or above it. The mean is the arithmetic average, so a small number of very low or very high scores pulls it away from the middle. EPC Benchmark reports both. The median is less sensitive to extreme values than the mean, which is a property of the statistic rather than a rule about which comparison suits a given question.

• Why the four views can disagree

  The four views can disagree because each holds something different constant. A district contains a mixture of ages and types, a national type-and-age group contains a mixture of areas, and each group has its own distribution. A score can therefore sit below one median and above another without any contradiction.

• Band shares are not percentiles

  Band shares show how many properties fall in each band. They do not place a score at an exact percentile within its group, because the published figures report bands, counts, medians and means rather than the full score distribution.

• When a view is unavailable

  A view reads as unavailable when the published group holds fewer than five properties and was suppressed, or when no figures were published for it. The remaining views are unaffected and stay readable on their own.

• How to read recommendation figures

  Recommendation figures count how often a measure was recorded across a national type-and-age group, so their order is a frequency ranking within that group. An individual official certificate lists measures in the sequence the assessment set out for that one building, which is a different thing and takes precedence for that property. The average figure alongside each measure is the total current-to-potential gap on the certificates carrying it, which belongs to the whole certificate rather than to the single measure, so it is an association and never the expected uplift of that item.

Methodology in brief

A summary of how the published aggregates are produced. The methodology page holds the detail.

1. 1Start from the official domestic certificate historyThe analysis starts from 27,654,564 domestic certificate-history rows for England and Wales, with a source snapshot ending 31 July 2026.
2. 2Reduce to one current property per valid UPRNCertificates are grouped by valid UPRN and only the latest certificate for each property is kept, giving 19,324,312 current properties. 485,734 raw rows have an invalid or null UPRN and are excluded.
3. 3Attach a postcode district and sector19,172,788 current properties join to a postcode district and sector using the ONS UPRN lookup, a join success of 99.21588929013359%.
4. 4Aggregate into published viewsProperties are aggregated into whole-district, district-by-property-type, district-by-construction-age and national type-and-age views. Only aggregate rows are published: counts, medians, means and band shares.
5. 5Apply suppressionAny group with fewer than 5 properties is withheld, so no small group can be identified from the published figures.
6. 6Summarise recorded recommendationsRecommendation rows (102,824,532 domestic rows) are counted by national property type and construction age. Figures describe association with the certificate's total potential gap, never an item-specific uplift.

Data notes

Limitations

• An EPC is a modelled assessment under standard occupancy assumptions, not a measurement of the energy a household actually uses.
• Comparability depends entirely on the fields recorded by the assessor, including property type and construction age band.
• This release publishes aggregate views only. It does not combine district, property type and construction age in one cell, so the views must be read separately.
• Groups smaller than 5 properties are suppressed, so some views are unavailable in smaller districts.
• Recommendation figures are associations with the certificate's total potential gap, not item-specific causal uplifts.
• No costs or savings are shown, because this release does not contain validated current costs.
• Not live data: the snapshot ends 31 July 2026 and the release was generated 23 August 2026.
• This is an independent service, not an official register and not a professional survey.