EPC ATLASProperty intelligence and EPC research
Home / Research / Methodology
Methodology - EPC Atlas UK Data Book 2026 v0.3

How EPC Atlas publishes large-scale EPC research

The 2026 research is deliberately conservative: every number is directly counted, arithmetically derived from those counts, or clearly labelled as an exploratory association.

17,624,989rows represented
1,357,124,153programmatic validation assertions represented
10supplied England-and-Wales regional datasets
0future inspection dates after 28 Aug 2026

Core publication rules

RuleReason
Say “records”, not automatically “homes”Administrative EPC evidence can be historical and duplicated across source certificates.
Preserve denominatorsKnown-status, applicable-property and all-record denominators answer different questions.
Explicit No is not missingCritical for mains gas, Solar PV, roof and loft.
Text evidence is not a verified installationCritical for heat pumps and other technology descriptions.
Do not globally dedupe on raw UPRN aloneThe ten-region register contains 26 cross-region UPRN collisions.
Treat 2025-2026 as a review boundaryAbrupt potential-rating and window-efficiency distributions require source-method verification before trend claims.
Separate finding from hypothesisCorrelations and odds ratios are descriptive unless an appropriate causal design exists.

Field completeness and applicability

Solar PV is known on 96.58% of the ten-region corpus, tenure on 99.94%, floor area on effectively 100%, mains-gas status on 74.95%, roof type on 69.30% and numerical loft depth on 50.51%. Lower completeness does not automatically mean poor data: some fields are naturally less applicable in flat-heavy areas.

Off-gas analysis uses explicit No only among records where mains-gas status is known. Unknown is never treated as off-gas. Low-loft analysis is restricted to applicable House/Bungalow records with a known numerical loft depth.

What the data cannot tell us

  • Current legal ownership or current occupancy.
  • Whether recorded solar or heat-pump equipment is still present, working or owned today.
  • Technology installation date.
  • Property-specific retrofit suitability or verified retrofit cost.
  • Household income, fuel poverty, conservation/planning constraints or present market value.
  • Actual metered energy use or live measured carbon emissions.

Known quality flags

  • North West Solar PV has 602,268 missing/unknown values.
  • North West Lead_Rank contains 1,162 duplicate values and 1,162 missing sequence values; it must not be used as a unique key until regenerated.
  • 26 provisional cross-region UPRN collisions are recorded.
  • 1,005 records have potential numeric score lower than current; 320 have a worse potential letter.
  • 55 current numeric scores and 81 potential numeric scores exceed 200 and are quarantined from ordinary averages.

Revision discipline

Combined national totals are regenerated from harmonized regional outputs rather than manually patched. Revision v0.3 incorporates the audited North West master and preserves v0.2 as a historical nine-region snapshot.

Download full methodology in the Data Book