What were the
key findings?

This section outlines the key findings based on the data collected by the installed property sensors across the sample portfolio of 121 properties. The average measurement period for data collection was 21.8 days. Sensors were installed between January and July 2025.^h

EPCs can provide an inaccurate measure of a building's thermal performance

The Heat Transfer Coefficient (HTC) is a key metric used to assess the overall thermal performance of a building. It measures the rate at which a building loses heat through its external envelope (fabric, air infiltration and ventilation), based on the rate of heat loss per degree of temperature difference between inside and outside.

Across the sample of properties included in the pilot, the average thermal performance gap between modelled and measured HTC was 25%.
This means that, on average, the properties included in the sample under-perform their modelled predictions by 25% - they are losing heat, on average, 25% more quickly than their EPCs predict. Also, there were some notable outliers – two properties, located on the same street and with very similar characteristics, had a thermal performance gap of more than 300%.

Figure 3: Measured vs modelled heat transfer coefficient

_{^h Senze's Live Thermal Measurement Algorithm (LTMA) can estimate thermal performance to a high degree of accuracy over a short measurement window and is minimally affected by occupancy patterns, seasonal variation, or resident behaviour (as noted in the findings of Birmingham City University's verification). However, the relatively short measurement window does have implications relating to the observed property conditions during the pilot, as explained on p.11 of this chapter.}

Analysis of thermal performance by property age also revealed notable trends. Older properties in the pilot sample generally outperformed modelled predictions, or at least aligned more closely with expected performance, compared with newer homes. While the sample size, particularly for newer properties, was limited, the finding warrants further investigation. If confirmed at scale, this could suggest that older buildings may require fewer interventions than anticipated, while newer properties may merit greater attention.

Figure 4: Thermal performance gap vs age of property

A measured data approach could enable more effective and efficient retrofit decisions

Senze has utilised measured data such as HTC to arrive at a set of ‘measured’ EPC for each of the homes, by comparing these against the ‘modelled’ EPC, Housing Associations can make better-informed decisions on:

Which properties require retrofit works to improve their performance and should be prioritised?
What specific interventions will be most effective to improve the performance of each property?

Decision optimisation: Prioritising the right properties for retrofit

As part of its consultation on Minimum Energy Efficiency Standards (MEES), the UK Government has proposed a deadline for all social housing properties to achieve an EPC rating of C or above - according to reformed EPC metrics - by 2030.

Given this policy landscape, Figure 5 shows how a measured approach to retrofit could be used to inform a decision-making process to help housing associations decide what action to take regarding each segment of the sample portfolio, and in what order. The logic behind this approach and its application to the pilot portfolio is described in more detail within this chapter.

Figure 5: Decision tree

The chart below shows how modelled SAP scores and estimated SAP scores (based on measured data) compare across the sample portfolio.

Figure 6: A decision optimisation approach

In this video, Tom Fenton, CEO of Senze, reflects on the pilot’s findings and highlights the critical need for more robust measurement and verification in driving smarter retrofit strategies.

We used this analysis to divide the sample portfolio into four groups for the purposes of retrofit planning:

Core priority

Retrofit is required based on both EPC rating and measured data (i.e. these are properties with an EPC rating of D or worse, and confirmed by measured data).

Compliance priority

Retrofit works are required based on EPC rating, but not measured data (i.e. these are properties with an EPC rating of D or worse, where measured data indicates performance of EPC C or better).

Home Health priority

Retrofit works are required based on measured data, but not EPC rating (i.e. these are properties with an EPC rating of C or better, where measured data indicates performance of EPC D or worse).

No action required

Retrofit works are not required – these are properties with an EPC rating of C or better, and measured data confirms that they perform as such. Most of these are properties that have already been or are already in the process of going through retrofit works, funded through Bromford Flagship's original programme of works (pre-dating this pilot project).

Decision optimisation: Identifying the most effective interventions

Once the data has identified which properties to prioritise for retrofit works, the next stage is to identify which interventions will be most effective for each individual property. Senze makes these targeted retrofit recommendations based on room-by-room measured thermal performance data and 3D digital twins. This is demonstrated through the case studies in this chapter.ⁱ

_{ⁱAll costs stated in these case studies relate to the cost comparison based on Bromford Flagship's provided data on unit costs for retrofit interventions, not the Green Buildings Tool's unit costs. Therefore, any cost differences stated are driven entirely through the more targeted nature of the interventions, rather than any differences relating to the cost of the interventions themselves.}

Group 1: Core Priority

59% of pilot sample

Case Study:

Semi-Detached House

1930-1949

EPC D modelled

EPC D measured

Bromford Flagship allocated over £58,000 for a retrofit package including an air source heat pump, external wall insulation, loft insulation, and replacement windows and doors. However, analysis suggests that this level of intervention is not required to achieve EPC C.

Measured heat loss was significantly lower than modelled estimates, and room-by-room assessment identified only targeted improvements as necessary. The recommended measures are limited to improving airtightness and installing solar panels, at an estimated cost of under £15,000, representing a potential saving of over £43,000 compared with the initial budget.

The property's peak heat load of 3.8-Kilowatt Peak (kWp) also means that a 4-5 kW air source heat pump would be sufficient if installed. This avoids the need for an oversized 10 kW unit, which has often been used in previous retrofits.

Implications

For property owners – undertaking more targeted interventions could lead to significant cost-savings, in this case more than £43,000, as well as lowering embodied carbon emissions, for example through right-sizing heat pumps.
For residents – more effective scoping of interventions to deliver the improvements needed for the specific property, as well as potentially reducing disruption through avoiding unnecessary works.

Group 2: Compliance Priority

7% of pilot sample

Case Study:

Terraced House

1900-1929

EPC D modelled

EPC C measured

Bromford Flagship has earmarked nearly £36,000 to be spent retrofitting this property, including a range of measures to improve the building fabric, such as external wall insulation, loft insulation and replacing windows and doors.

Yet measured data shows the property already performs to an EPC C equivalent, comfortably outperforming modelled expectations. As a result, the property does not require retrofit works where measured EPC C equivalence is the benchmark. Asset owners may choose to prioritise other dwellings where measured data indicates higher levels of heat loss. Given the ongoing policy consultation, EPC may consider using measured data to improve accuracy.

Implications

For housing associations – avoiding unnecessary works could lead to significant cost-savings, in this case £36,000, relative to the £1,500 Senze assessment cost.
For residents – reduced or avoided disruption and avoidance of unintended consequences, such as over-insulation leading to moisture build-up, mould growth, and poor indoor air quality.
For policy and practice – this property's solid wall construction is likely the reason for its stronger-than-modelled thermal performance. Research by UCL¹⁹and BRE²⁰has shown that measured U-values for solid walls are often lower than modelled SAP estimates, indicating better thermal performance.^jThis has potentially significant implications, including overestimation of heating demand in energy modelling, potentially affecting policy and investment decisions.

_{^jU-values show how well a building element, e.g. a wall or a roof, prevents heat from escaping.}

Group 3: Home Health Priority

13% of pilot sample

Case Study:

Semi-Detached House

1930-1949

EPC C modelled

EPC E measured

A property that has already undergone a major programme of retrofit works under Bromford's earlier retrofit strategy (pre-dating this pilot project and pre-dating its merger with Flagship). This was funded through a mix of its own funding and government grant. More than £42,000 has been invested to date.

Despite the significant investment, measured data shows the property performs at an EPC E equivalent, with a substantial thermal performance gap indicating high levels of heat loss. No baseline performance data was available, so improvements achieved since retrofit cannot be quantified. However, current measured performance demonstrates that interventions to date have not been effective in achieving EPC C, notwithstanding the property's modelled EPC C rating.

Previous retrofit measures focused primarily on decarbonisation, including installing a heat pump, as well as some measures to improve building fabric such as loft insulation, and new windows and doors. Yet measured data indicates that the property's performance would be improved more effectively through targeted fabric and airtightness upgrades, such as wall and/or floor insulation.

Implications

For residents – ongoing exposure to high energy costs and poor thermal comfort despite substantial retrofit works.
For policy and practice – risk of misdirected retrofit investment when works are not aligned to actual property performance.
For funders – raises concerns about the effectiveness of funding already deployed, underscoring the potential role that measured approaches could play to ensure value for money.

Group 4: No Retrofit Required

21% of pilot sample

Case Study:

Bungalow

1950-1966

EPC C modelled

EPC C measured

A property which has gone through the retrofit process based on Bromford's original retrofit programme (pre-dating this pilot project and its merger with Flagship). It is now rated EPC C, and measured data confirms it performs as such.

In terms of retrofit measures, this property has had solar panels, loft insulation and mechanical ventilation installed, as well as having windows and doors replaced. This was carried out at a cost of around £23,000.

We do not have baseline measured performance to compare against, so we cannot comment on the effectiveness or efficiency of this funding in improving thermal performance. However, we can confirm, based on the measured data, that the property is now performing at the level of an EPC C.

Measured data reveals signs of internal living conditions which may be impacting resident health and wellbeing

Within the sample properties in the pilot where internal property conditions were measured:

26 homes (22% of the sample) exhibited an average internal property temperature which would see them defined as 'cold homes' by both the World Health Organisation (WHO) and Public Health England. Homes maintained at such temperatures are likely to affect the health and wellbeing of inhabitants.²¹ It is likely that a high proportion of these homes are experiencing fuel poverty based on their measured energy use during the measurement period.
18 homes (15%) had average humidity levels above 60%. This can lead to damp, mould and mildew growth, and potential respiratory issues, and could have implications for Bromford Flagship as a landlord in relation to Awaab's Law, which will mandate that social landlords must address health hazards such as damp and mould within specific timeframes.²²

As no occupancy data was collected for this project^k, and the sensors were not installed over a full 12-month period, it is not possible to determine with certainty whether observed property conditions were due to under- or over-occupancy, or seasonal differences. Nonetheless, these types of findings show that measured data can provide housing associations with a useful high-level screening tool to identify properties that may require further investigation for potential health and safety actions as well as potential retrofit measures.

Further details on the internal property conditions observed during the measurement window can be found in the Appendices.^l

^k Although no occupancy data was provided, all properties in the pilot sample were tenanted. What cannot be determined is the level of occupancy during the measurement period – for example, whether residents were present throughout or absent for extended periods.
^lSee 'Appendix 6: Key Findings' in The Science Behind the Insight for further details.'