Article 04: Mapping the Stock: Inside the 3DStock Model and Why Buildings Resist Easy Counting

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

Key takeaway. Before you can measure energy use across thousands of offices, you have to decide what counts as an office. The 3DStock model stitches seven public datasets into a single geospatial picture, and exposes how unstable that question really is.

Figure

Seven public datasets, one geospatial picture

The 3DStock model is a controlled join across seven public datasets, each adding a piece the others lack, producing a Self-Contained Unit for every office in Greater London.

Ordnance Survey AddressBase Premium

Address points

OS Master Map Topography Layer

Footprints

HM Land Registry INSPIRE Polygons

Title boundaries

Environment Agency LiDAR

Heights

GIM International UK Buildings

Age, materials

VOA NDR + SMV

Activities, floor area

BEIS energy meter data

Operational layer

3DStock

Greater London

6,038 office SCUs

with energy data

Each dataset answers a different question; the engineering of the join is more than half of the methodology.

What is one building?

Walk down any street in central London and you can probably count the buildings without thinking. A façade, a doorway, a roofline: that is one. Try doing the same exercise with the data. The Ordnance Survey thinks in terms of address points. The Valuation Office Agency thinks in terms of premises. Land Registry thinks in terms of title boundaries. The Environment Agency thinks in terms of LiDAR returns. The Department for Energy thinks in terms of gas and electricity meters. Each of those answers a different question, and none of them maps cleanly onto the everyday idea of a building.

This sounds like a research-methodology nuance. It is actually a precondition for almost everything else in the thesis. Without a defensible answer to the question what is one office?, there is no way to compute floor area per office, energy per office, or even simply the number of offices in Greater London. The 3DStock model exists to provide that answer, and to do it for the whole city in one consistent way.

Seven datasets, one geospatial picture

The 3DStock model is best understood as a controlled join across seven public datasets. Each adds a piece the others lack.

The Ordnance Survey Address Base Premium contributes geolocated addresses, identifying which points are domestic and which are non-domestic. The OS Master Map Topography Layer provides building footprints and simplified heights. The Land Registry INSPIRE Index Polygons give property-title boundaries, which matter for ownership rather than physical extent. UK Environment Agency LiDAR data give geolocated, high-resolution height information, which lets the model resolve the difference between a four-storey building and a six-storey one without relying on coarser estimates. GIM International UK Buildings dataset adds approximate construction date and materials, both of which feed into age-band analysis. The Valuation Office Agency Non-Domestic Rating List (NDR) and the VOA summary valuation data (SMV) provide addresses, activities and floor areas for non-domestic premises. And finally the Department for Business, Energy, and Industrial Strategy (BEIS) provided annualised gas and electricity meter data under a strict data-sharing agreement, which is the operational layer that turns a stock model into an energy model.

None of these datasets is sufficient on its own. The rating list does not contain height. LiDAR does not know what a premise is used for. Land Registry does not know about meters. The art of 3DStock is the join, and the engineering of that join is more than half of the methodology chapter of the thesis.

The Self-Contained Unit and why it matters

The crucial conceptual move is the Self-Contained Unit, or SCU. An SCU is the smallest set of premises and physical footprints that share an energy boundary. In simple cases, an SCU is one premise inside one building with its own meters. In complex cases, an SCU spans several premises, several footprints and several meters. The model selects the smallest grouping such that every premise, every footprint and every meter inside the boundary belongs together, and no premise, footprint or meter crosses the boundary.

Why does this matter? Because the binary distinction between domestic and non-domestic buildings (or between commercial and residential) breaks down very quickly in a city like London. Many actual buildings contain a non-domestic ground floor (a shop, a café, a clinic), a non-domestic upper floor (an office) and domestic flats above, often with shared circulation, shared meters and shared services. Asking whether the whole thing is an office is the wrong question. Asking which subset of premises, footprints and meters constitutes the office part is the right one.

The SCU concept does that, and the Sankey diagrams in Chapter 3 of the thesis (Figures 3-9 and 3-12) show how often this matters in practice. A surprising share of non-domestic floor area in Greater London belongs to SCUs that mix office activity with other non-domestic activity, or with domestic activity, in ways the rating list alone cannot resolve.

Chart

Non-domestic premises floor area by CaRB2 class & activity

The flow of total floor area from all premises through broad classifications to specific activities.

Sankey diagram showing the subdivision of the non-domestic stock.

Chart

Complex relationship between premises and SCUs

Premises Floor Area (By CaRB2 Classes) mapped to SCUs Floor Area (By Dominant Activity).

Sankey diagram showing how floor area is reclassified across the energy boundary.

Data is messy: the filtering pipeline

A geospatial join across seven datasets at city scale produces a long tail of edge cases. The thesis describes a filtering pipeline that, broadly, does three things.

First, it reconciles floor areas. The VOA summary valuation data gives a recorded floor area for most non-domestic premises. LiDAR plus footprints give an independent estimate of floor area. Where the two agree closely, the SCU is taken at face value. Where the SMV and LiDAR figures diverge sharply, the case is filtered out as untrustworthy.

Second, the pipeline applies sanity checks on activity. Premises whose rating-list activity strongly disagrees with what the building physically looks like (for example, a warehouse use class in a high-rise residential tower) are flagged. Some are correctly labelled and survive. Others fall out.

Third, the pipeline matches energy meters. BEIS meter data is the operational input, but meters do not naturally come with a building name attached. The matching is done via the Unique Property Reference Number (UPRN) and through geospatial cross-checks. Where matches are weak or ambiguous, the SCU is dropped from energy analysis rather than carry forward a guess.

The result of this filtering is a clean sample of 6,038 office SCUs in Greater London, which is the dataset behind every quantitative finding in the thesis and every empirical article in this series. The number is large by historical standards. Armitage and colleagues (2015), one of the largest previous empirical studies of UK office energy, analysed a few hundred public-sector offices via Display Energy Certificates. The 3DStock sample is an order of magnitude larger and includes the private sector that DEC coverage misses.

What 3DStock still cannot see

The model is powerful but bounded. It is worth being explicit about what it does not capture, because almost every analytical caveat in the rest of the series traces back to one of these gaps.

3DStock captures geometry and geography. It does not capture operating hours, plant efficiency, internal loads, occupant density, fabric U-values or BMS set-points. It captures premise activity at the level the rating list records it, which is a coarse classification. It captures ownership only implicitly, through the rateable owner field, and not the more nuanced organisational structures that the qualitative interviews surface. It captures a single year of energy data (2017), not a longitudinal panel. And it captures Greater London, not all of England and Wales, although the thesis tests the extent to which London is representative of the rest of the country (Article 11).

Article 3 leans hard on these gaps to explain why a regression on 33 variables can only get to an R-squared of 0.18 on electricity EUI. Janda and colleagues (2022) have proposed a 4DStock extension that would attach organisational information to each SCU. That is the most interesting active research direction for stock-level energy work in the UK, and probably the most consequential.

Why this methodology deserves attention beyond the thesis

The 3DStock approach has applications well beyond academic energy research. Local authorities planning area-wide decarbonisation programmes need a defensible stock model. Investors writing transition-finance documents need credible counts of buildings by type, age and energy intensity. Regulators considering operational rating thresholds (see Article 1) need to be able to say with confidence how much floor area, how many premises and how much energy fall above and below any given line.

The model is a public-good asset built on public datasets, but its construction is non-trivial and depends on the data-sharing arrangements that BEIS maintain with researchers. The thesis argues, implicitly, for that infrastructure to be sustained and extended. Without it, the rest of the analysis in the series is impossible.

Since the thesis was completed, the 3DStock approach has been scaled to all of Great Britain in the 2026 National Buildings Database (DESNZ), which the author contributed to. NBD reports 163,131 office SCUs and 470,455 office premises across the country, using the same Self-Contained Unit concept at its core. The methodology has proved durable beyond the original Greater London application. NBD also extends the activity-class typology beyond a single office category, distinguishing General Office, Office (Local Authority and Central Government), Business Units, Studios, and Film Studios and Computer Centres. The finer typology is one of the most useful methodological refinements for the energy analyst, because median EUI varies substantially across these groups.

Limitations

The 3DStock model is a Greater London empirical model built from 2017 datasets. Coverage of LiDAR, EPCs and BEIS meter data is incomplete in places. Discrepancies between SMV (rateable value) floor areas and LiDAR-derived areas required filtering. Filtered cases are not random. Premise classification relies on VOA Non-Domestic Rating List categories, which are an administrative artefact and can mis-label mixed-use buildings. The SCU concept resolves many counting problems but creates its own boundary decisions that other modellers may make differently. Reproducibility outside the original research environment requires data access agreements with BEIS and the Better Buildings Partnership.

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/
• Evans, S., Liddiard, R., and Steadman, P. (2017). 3DStock: a new kind of three-dimensional model of the building stock of England and Wales. Environment and Planning B. Available at: https://doi.org/10.1177/2399808317725161
• Evans, S., Liddiard, R., and Steadman, P. (2019). Modelling a whole building stock: domestic, non-domestic and mixed use.
• Steadman, P., Bruhns, H., and Rickaby, P. (2000). An introduction to the national non-domestic building stock database. Available at: https://doi.org/10.1068/b2680
• Taylor, S., Lowe, R., and Steadman, P. (2014). Identifying buildings and properties in non-domestic energy datasets.
• Bruhns, H. (2008). Identifying determinants of energy use in the UK non-domestic stock.
• Armitage, P., Godoy-Shimizu, D., Steemers, K., and Chenvidyakarn, T. (2015). Using Display Energy Certificates to quantify public sector office energy consumption. Available at: https://doi.org/10.1080/09613218.2014.975416
• 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 1: The 30/85/89 Problem.
• Article 11: Can London Speak for England and Wales?

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?