construction · energy · family: the wrong ruler
insulationnobody checked
98% of government-subsidised external wall insulation jobs in Britain have major defects — installers could choose their own audit odds, employed their own retrofit coordinators, and only 1% of homes fully met the standard
Problem statement
Retrofit at scale is a workmanship problem before it is a technology problem, and Britain's flagship retrofit subsidy has just shown what happens when the quality-assurance layer is not built to the same scale as the installation layer. Under the Energy Company Obligation (ECO4, 2022–26) and the Great British Insulation Scheme (GBIS, 2023–26), 28,000 homes received external wall insulation (EWI) and 45,200 internal wall insulation (IWI). Statistically representative audits of 758 homes commissioned by DESNZ and Ofgem found that "98% of homes with external wall insulation have major issues requiring remediation (between 22,000 and 23,000 homes)" and 29% of IWI homes (9,000–13,000): 92% of EWI homes have major issues that "will affect the insulation's performance, often creating the risk of water ingress and mould," 6% have "health and safety risks that require immediate correction, such as inadequate ventilation" (elsewhere the report cites exposed live cabling and insufficient ventilation for gas boilers), and just 1% "fully met the PAS 2035 standard." The unsolved problem is not how to install insulation correctly — PAS 2030/2035 specify it — but how to design a subsidised, high-volume, many-small-installer retrofit market whose assurance system cannot be gamed and actually scales with volume.
Why this matters
Solid-wall homes are the hardest and most carbon-intensive part of the housing stock to decarbonise, and every country pursuing renovation-wave targets will need to push wall insulation through thousands of small contractors paid by public or obligated money. Here the failure produced tens of thousands of damp-prone homes occupied largely by low-income and vulnerable households, an Ofgem estimate that retrofit businesses "falsified claims for ECO installations in between 5,600 and 16,500 homes to potentially claim between £56 million and £165 million," and a remediation programme that by September 2025 had fixed only 8% of affected EWI homes and 10% of IWI homes. In the sample audit, around 40% of the households that auditors managed to reach refused them entry. The NAO's headline conclusion: "there have been clear failures in the design and set-up of ECO4 and GBIS and their consumer protection and quality assurance system." Because damp and mould damage compounds, the cost of the same defect rises the longer it goes unfound — and public trust in retrofit, the scarcest input to any renovation wave, erodes with each headline.
What’s been tried and why it hasn’t worked
The system was designed as an answer to earlier failures: after the 2016 "Each Home Counts" review, government built an arm's-length quality regime — installers certified to PAS 2030 by UKAS-accredited certification bodies, projects overseen by PAS 2035 retrofit coordinators, all lodged with TrustMark, energy suppliers accountable to Ofgem, DESNZ setting policy. The NAO found each layer failed in a specific, mechanistic way. The audit rate was set by standard, not risk: PAS 2031 required certification bodies to inspect only "between 3% and 7% of external wall insulation and between 4% and 10% of internal wall insulation installed by each installer," and "no-one asked if they were adequate to collectively show that most ECO measures were installed correctly." Installers could game the sample: they could be "certified by multiple certification bodies or transferring their certification between bodies," resetting their history to earn the lowest audit rate, and certification bodies' records "understated the number of installations compared to the installations lodged in TrustMark's data warehouse" — the auditors did not know how many jobs existed. The NAO adds that it is "also possible that it was commercially beneficial to installers to remediate the small proportion of projects subject to audit rather than meet the required PAS standards across all their projects." The independent check inside each project was captured: retrofit coordinators, "originally intended as a control against the incentives on the installer," were "often contracted or employed by the installer." TrustMark's funding model "did not allow it to employ enough trained staff who could audit against PAS 2035," its analytics were not operational until late 2024, and until March 2025 there was no common way even to categorise the severity of audit findings. Potential reasons suggested by DESNZ, Ofgem, TrustMark and UKAS: "poor workforce skills, including subcontracting work to others who are not competent or registered with TrustMark; uncertainty over how the different standards apply to different jobs; and shoddy work produced as retrofit businesses 'cut corners' in both the design and installation." Fixes so far — a 20% audit sample for ECO4 EWI, a ban on multiple certifications, data-sharing agreements — patch the sample without changing the incentive that made gaming pay.
What would unlock progress
The reframing is to treat installation quality as an adversarial verification problem rather than a compliance-paperwork problem: sampling that installers cannot predict or dilute (drawn from a single register of every job, with the rate set by risk and published), independence for the coordinator role (paid from the subsidy, not by the installer), payment structures that hold back a share of the subsidy until an independent post-installation check clears, and audit findings that are categorised consistently so trends surface within weeks rather than years. Adjacent precedents: financial-audit independence rules and rotation, food-safety and aviation-maintenance inspection regimes that scale audit intensity to installer history using a single mandatory registry, and photographic/geotagged evidence requirements used in agricultural subsidy verification. Cheaper verification (guided photo capture at defined stages, thermal imaging at handover, occupant-reported checklists) would let a higher fraction of jobs be checked at the same budget.
Entry points for student teams
A team could design and simulate a risk-based, gaming-resistant audit-sampling scheme for a market of many small installers — modelling installer strategies (multi-certification, under-reporting, remediate-only-the-audited) and finding the sampling and payment rules under which cutting corners no longer pays. A product team could prototype a stage-gated photo/thermal evidence app for EWI/IWI installs, aligned to PAS 2035 checkpoints, and test with a local installer or housing association. A policy team could produce a comparative design proposal for a successor scheme (retrofit-coordinator independence, retention payments, single job register), using the NAO evidence as the baseline. Relevant skills: mechanism design/game theory, building physics and retrofit practice, data systems, public-policy analysis.
Genes
Structural cousins: stuck the same way, in other fields
Sources
National Audit Office (2025), "Energy efficiency installations under the Energy Company Obligation," Report by the Comptroller and Auditor General, HC 1334, Session 2024–2026, ordered to be printed 13 October 2025 (published 14 October 2025), (landing page ), accessed 2026-08-18 go to source 1 ↗ go to source 2 ↗
Working notes
Notes made while writing and checking this brief, shown as written. They use working shorthand.
All figures and quotations (28,000 EWI / 45,200 IWI installs; 758-home representative audit; 98%/29% major issues; 92%/6%/1% breakdown for EWI; 22,000–23,000 and 9,000–13,000 homes; £56–165 million suspected false claims; 8%/10% remediated by September 2025; 40% refusal; PAS 2031 3–7% and 4–10% audit rates; multi-certification and under-reporting gaming; retrofit coordinators employed by installers; root-cause list) are from the NAO report PDF (64 pp.) read on 2026-08-18. `constraint:coordination` was considered because six organisations were involved, and rejected on filter (1)/(2): the actors did not share an agreed approach to assurance (no one defined the required level) and the binding constraints were the installer incentive structure and an under-funded, gameable audit design — hence `behavioral` + `economic` + `regulatory`. `failure:regulatory-mismatch` for the perverse-incentive sub-pattern (audit rate fixed by standard; coordinator paid by installer); `failure:ignored-context` for the deployment/institutional-fit sub-pattern (assurance capacity not modelled against the expected volume; NAO: DESNZ "found no evidence that it had modelled TrustMark's funding against the expected increase in installed measures"). `stakeholders:multi-institution` passes the three-criteria test: DESNZ (policy), Ofgem (administration), TrustMark (project register/PAS 2035 audit), UKAS and certification bodies (installer certification/PAS 2030 audit) and energy suppliers (obligation) each own a non-substitutable piece; the NAO attributes the failure to "unclear and fragmented roles, responsibilities and accountabilities" — the boundary is the barrier. `temporal:window` (deadline type): ECO4/GBIS were scheduled to close in March 2026 (the NAO notes DESNZ was consulting on a six-to-nine-month ECO4 extension) and DESNZ is designing successor schemes (Warm Homes Plan) now, while damp damage in ~30,000 unremediated homes compounds. `failure:success-caused` was considered (the scheme scaled installations, and scale outran assurance) and rejected because volume was the intended output, not a success mechanism structurally coupled to the harm. Related collection briefs: `energy-building-performance-prediction-gap` and `construction-panelized-retrofit-midrise-attachment-gap` address retrofit performance and technology; none addresses assurance-system design for subsidised retrofit markets.
Source type: Self-articulated (national audit body reporting on the scheme's own government owners)
Verified at intake 2026-08-18: gate (net) + adversarial source check + contested-tag second coding. Verifier re-read the NAO PDF (HC 1334); all figures and quotations confirmed (98%/92%/6%/1%; 29%; 758; £56–165m; 3–7%/4–10%; multi-certification gaming; coordinators employed by installers; 'clear failures').
Reconciliation 2026-09-29: Fact check after a site review read the whole write-up against its source. The National Audit Office report (HC 1334, 64 pages) was fetched in full on 2026-09-29 from https://www.nao.org.uk/wp-content/uploads/2025/10/energy-efficiency-installations.pdf, and its landing page from https://www.nao.org.uk/reports/energy-efficiency-installations-under-the-energy-company-obligation/. (1) The title said installers "Chose Their Own Audit Odds." The report says it was possible for installers to game the audit process and that they were able to game the system (paragraphs 3.13 and 3.18). It does not report a confirmed case. The title now says "Could Choose." (2) For the same reason, "Installers gamed the sample" now reads "Installers could game the sample." (3) The text quoted the report as saying "it was commercially beneficial to installers" to fix only the audited jobs. The report's sentence begins "It is also possible that" (paragraph 3.18). The quotation now includes those words. (4) The text said "40% of contacted households" refused auditors entry and tied this to the remediation programme. The report says auditors contacted 3,798 homes for the sample audit, established contact with 1,186, and around 40% of those (428) refused entry (paragraph 2.4). The sentence now says that. (5) The text introduced the list of causes as "Root causes named by" the four bodies. The report says they "suggested potential reasons" (paragraphs 20 and 3.2). The sentence now says "Potential reasons suggested by." (6) The text said the report cites "blocked boiler ventilation." The report's wording is "Insufficient ventilation for the gas boiler" (Figure 6), and the word "blocked" does not appear in it. The phrase now follows the report. Confirmed against the report and left as written: the Source line (title, HC 1334, Session 2024–2026, ordered to be printed 13 October 2025, published 14 October 2025); 28,000 external and 45,200 internal wall insulation installations; the sample of 758 homes; 98% and 29% with major issues; 22,000 to 23,000 and 9,000 to 13,000 homes; 92%, 6% and 1% for external wall insulation; 5,600 to 16,500 homes and £56 million to £165 million in suspected false claims; 8% and 10% remediated by September 2025; the PAS 2031 audit rates of 3% to 7% and 4% to 10%; the 20% sample asked for in December 2024; March 2025 for the common categories of audit findings; and every other quotation in the body. NOT checked: the suggestions in "What Would Unlock Progress" and "Suggested Entry Points for Student Teams," which are advice and do not come from the report. The earlier notes above still describe the gaming and the root-cause list in the original wording; they are left as they were. Genes untouched.