Trust, Data and Years of Notice

At a glance 

• Sovereign Network Group's development, repairs, planned works and forward maintenance functions each held their own version of property data, and the records did not always agree. 

• Spotr scaled from an initial 700 homes to SNG's full 65,000 home general needs portfolio once the new system's outputs matched existing processes. 

• AI is used to model what if scenarios for retrofit and divestment decisions worth tens of millions of pounds, but the final call stays with people, not the system. 

A single view across disconnected systems 

Sovereign Network Group manages around 85,000 homes across development, repairs, planned works and forward maintenance. Until recently, each of those functions held its own version of the truth about a property. A team could arrive to fit an air source heat pump and find a new gas boiler already installed, because responsive repairs had worked from a different system entirely. 

Richard Young, Technical Director for the Built Environment at SNG, and Dirk Huibers, co-founder and CEO of the technology platform Spotr, have spent the past year working through that problem together. The result, a single asset view, brought SNG's separate records into one place. A retrofit or maintenance plan that once took weeks to compile can now be produced in minutes, and conflicts between systems surface before they cause a wasted site visit or misdirected capital. 

Before building the system, Spotr's team spent time working alongside SNG's staff to establish what the underlying problem actually was, rather than building around an existing product and asking SNG to fit its business into it. 

Requirements that stack rather than replace 

Wave 2.1, then Wave 3, then a series of changes to Minimum Energy Efficiency Standards, have each added requirements on top of what came before, rather than replacing them. A system built to reflect one set of rules at a single point in time is out of date within days. 

A recent change to the MEES exemption clause means providers must now carry out interventions to secure an exemption at all, rather than deferring compliance to 2039. For SNG, that shift affects planning across every property archetype in the portfolio, and the interventions required differ from one archetype to the next. 

Data quality and where investment lands 

Many housing associations have merged over time, and each merger tends to bring together organisations that kept records in different systems, to different standards. That legacy affects decisions made today. An association with better quality data in one area than another risks weighting investment toward the area it can see most clearly, regardless of where the need is greatest. 

Improving data quality across a merged portfolio is a precondition for fair investment decisions, not a separate technical exercise. Cleaning and reconciling records from legacy systems determines whether resources reach the homes that need them, rather than the ones that happen to have the clearest records. 

Building trust before extending scope 

The single asset view was tested by checking that it reproduced the same answers SNG's existing spreadsheet-based processes already gave, rather than attempting to improve on them straight away. Only once that match was established did the scope extend, moving from an initial 700 homes to the full 65,000 home general needs portfolio. 

The same sequencing carries through to the supply chain. Where SNG can show ten years of expiring EPCs that will need loft insulation, a contractor can plan capacity and price against a firm pipeline, rather than an annual order that might run to two hundred homes one year and three hundred the next. 

Where AI stops and judgement starts 

The system is used to build what if investment scenarios: what a given level of compliance would cost, what selling more homes would save, and what the difference between those routes would be. SNG's investment decisions run into tens of millions of pounds across a portfolio managed over multiple years, and social housing decisions do not always follow a purely commercial logic. 

A property can make financial sense to sell and still be the wrong outcome for a resident or a community. The modelling is used to make the reasoning behind a decision visible to a board, a regulator or a tenant, while the decision itself remains with people. 

The distinction turns on how well defined the problem is. A well-specified, repeatable task can be handled by a fixed workflow with a clear point at which it is finished. A decision with several competing factors, and no single correct answer, is better served by keeping a person at the centre of the process throughout, checking the system's output rather than acting on it automatically. Social housing decisions, where a commercially sound answer and a socially sound one can differ, fall into the second category. 

Ongoing research with a university partner is exploring whether a system could represent the different parties in an investment decision, such as a tenant, a financial director and a sustainability manager, to produce a preliminary plan for a person to check. The aim is to give tenants a more consistent voice in that process, including tenants who might otherwise struggle to take part, without replacing their voice with the system's own. This work is still some years from practical use. 

Longer notice for residents 

A longer planning horizon changes what a housing provider can tell a resident. Rather than a single letter arriving once a repair or replacement becomes urgent, a household can be told years in advance that a kitchen or boiler is due for renewal, and kept informed as that date approaches. 

For SNG, that shift sits inside a wider corporate plan built around the customer and the home as linked priorities, with the single asset view treated as a requirement for delivering both. 

Practical steps for housing providers 

• Map which internal systems, development, repairs, forward maintenance and planned works, hold conflicting records for the same properties, and where those conflicts have caused delay or rework. 

• Before trusting a new system with complex decisions, test whether it reproduces the outcomes your current processes already produce. 

• Set out clearly where human sign off sits in any AI supported investment decision, particularly where social and commercial outcomes may not align. 

• Review whether stacking regulatory requirements, such as MEES alongside existing retrofit programmes, are being planned for together rather than addressed as separate exercises. 

• Consider what a longer planning horizon would allow you to tell residents about work planned for their home, and how far in advance that notice could realistically be given. 

Image credit: Adobe Stock