The scoring system your council quietly switched on
A TED talk, or a TEDx one in Grantham, explains its subject to an audience that has chosen to be there and can ask questions afterwards; the tenant this software has just ranked gets neither the choice nor the questions. Somewhere in South Kesteven's housing office, a piece of software is already deciding which tenants a council officer calls first.
In August 2026, South Kesteven District Council — the local authority covering Grantham, Bourne, Stamford, and Market Deeping — approved the procurement of Mobysoft RentSense, a predictive analytics tool built for social housing providers. The council is rolling it out within its Housing Income Management service during 2026 and 2027.
The system works by ingesting payment history and other housing management data, then applying machine learning to spot patterns that suggest a tenant may fall into arrears. It can flag potential risk up to three months before any debt arises. From that analysis, it produces a ranked list: these tenants first, those tenants later. Housing officers use that ranking to structure their caseload. In effect, an algorithm decides who gets a call and when.
South Kesteven framed the tool as a tenancy-sustainment measure — designed to catch problems early and keep people in their homes, not to penalise them. Mobysoft, the software's developer, reports that RentSense customers achieved a 26.8% reduction in former tenant arrears, compared to an 11.1% increase among non-users. Those figures are the vendor's own and should be read accordingly.
What is less clear is what tenants know about any of this. Whether a resident can find out they have been ranked, query the scoring, or seek any remedy if the system has assessed them incorrectly — those questions remain open, and they matter.
What tenants are unlikely to know
If you are a South Kesteven council tenant, there is no publicly available evidence that you have been told you are being scored. SKDC has published no tenant notification procedure, no explanation of how the RentSense criteria are weighted, and no specific appeals route for residents who believe their risk ranking is inaccurate.
That gap matters because the scoring shapes outcomes in both directions. RentSense is positioned as a supportive tool — early contact framed as help rather than pressure. But the algorithm still determines who receives that contact and when. A tenant incorrectly flagged as high-risk may receive early intervention that feels intrusive or stigmatising; one who has slipped through a scoring error may receive nothing until the debt is already serious. The Ada Lovelace Institute's global review of algorithmic accountability found that such systems are 'often illegible to those whose lives they shape' and 'hard to contest'.
None of this is confirmed as SKDC's failure in particular. This is a sector-wide pattern — most councils using algorithmic tools have not built tenant-facing transparency into the deployment. The absence of any published notification or redress mechanism may reflect ordinary gaps in how councils communicate about operational software purchases. But the transparency gap is real, and the potential for harm is structurally embedded in it.
Why public law was not built for this
The legal machinery that exists to hold councils to account was not designed with machine learning in mind.
In September 2025, the Law Commission made this explicit. Its 14th Programme of Law Reform named public-sector automated decision-making as a priority reform project, observing that 'there is no specific legal framework governing the use of ADM by the state' and that public law 'developed to ensure the accountability of human officials and not automated systems'. The principal tool available to citizens who believe a public body has wronged them — judicial review — is, in the Commission's words, 'not well-suited to scrutinising decisions made using ADM'. It works by examining whether a human official followed a proper process and applied the law correctly. When the decision emerges from a scoring model, the reasoning chain is rarely legible enough for that kind of scrutiny.
The scale of the issue came into sharp focus during a December 2024 House of Lords debate on the Public Authority Algorithmic and Automated Decision-Making Systems Bill. Peers were told that approximately 540,000 benefits applicants are secretly assigned fraud risk scores by councils' algorithms before they can even access housing benefit or council tax support. That figure illustrates a structural condition, not an edge case: algorithmic profiling is already shaping access to welfare services without residents knowing it is happening.
The Bill proposes transparency registers, pre-deployment impact assessments, and independent dispute resolution — mechanisms that would begin to fill the gap the Law Commission has identified. As of mid-2026, it has not been enacted.
Three ways a council can get this legally wrong
Three distinct legal pressure points face any council that deploys a tool like RentSense.
Fettering discretion is the most straightforward. Officers must use the algorithm's risk rankings to inform their judgement, not replace it. If a housing officer routes their caseload entirely by whatever the software outputs — without independently weighing individual circumstances — the council has effectively delegated a public decision to a machine. That is unlawful.
Indirect discrimination does not require bad intent. A model trained on historical arrears data will reflect the social patterns embedded in that history. If certain groups appear disproportionately in past arrears records, the algorithm may systematically elevate their risk scores — potentially breaching the Public Sector Equality Duty even when no one set out to discriminate.
Opacity compounds both risks. As section 2 noted, residents may not know they have been scored. But the problem extends further: if a score cannot be meaningfully explained to a tenant or an appeal body, a fair challenge becomes almost impossible — 'computer says no' as a structural barrier rather than a catch-phrase.
The legal landscape is live. What is thought to be the UK's first quasi-successful challenge to a public-sector algorithmic system ended not with a ruling but with a council withdrawing its tool before trial — showing both the vulnerability of opaque systems and how difficult mounting such challenges remains in practice. None of these risks have been found to apply to SKDC specifically; they attach to the model of deployment itself.
The trust a TED talk builds is not the same as being able to contest a decision
Research offers a more precise diagnosis than the legal frameworks do. A 2025 cross-cultural replication of Grimmelikhuijsen's work found that when people are shown how an automated decision was reached — not merely told that it was — trust in the outcome rises measurably. Crucially, the effect is not uniform: it is strongest in high-discretion decisions, ones where a human judgement call is expected and the personal stakes are real. Housing income management sits squarely in that category. An early arrears contact from a council officer carries more legitimacy if the tenant understands what prompted it; without that, the call is simply unexplained pressure from a public body. A TED talk earns exactly that kind of trust: it explains a mechanism clearly enough that an audience nods along, without ever exposing itself to being checked. Housing income management cannot settle for the same one-way transaction — a tenant needs not just an explanation that reads as plausible, but one they can push back on.
The Ada Lovelace Institute's global review of algorithmic accountability found that residents frequently cannot engage with decisions that directly shape their lives because the systems behind those decisions offer nothing to engage with — no stated reason, no clear route for question or challenge. That same opacity also keeps alive the risk identified by the Centre for Data Ethics and Innovation's 2020 review: that biased patterns in historical data persist and compound, precisely because no-one can see the data doing the work. Illegibility and inequality reinforce each other.
None of this requires publishing model weights or exposing proprietary code. Explainability, in practice, means telling a tenant in plain language what information prompted their case to be prioritised, and ensuring a clear route exists if that account turns out to be wrong.
The questions South Kesteven has yet to answer publicly
Three questions remain unanswered in any public document available at the time of writing.
Has South Kesteven's Overview and Scrutiny Committee reviewed the RentSense procurement — its methodology, its equality implications, or its operation in practice? There is no published record that it has.
Was an equality impact assessment completed before or after deployment? Nothing in SKDC's publicly accessible decision-making record confirms one either way.
If a tenant believes their risk ranking is wrong, what is the process for saying so? No tenant-facing explanation of how RentSense is used, or how to challenge its outputs, appears to be publicly available.
These are not accusations. They are gaps in the public record that the council is in a position to fill. The question for residents is not whether predictive analytics should feature in social housing management — that debate has, for now, been settled locally. The question is whether the institution that made this decision can be held to account for how it runs.
A scrutiny committee review, a published equality impact assessment, and a clear written explanation for tenants of how RentSense shapes casework decisions would each represent a specific, achievable step. None requires new legislation. All would require SKDC to ask itself, plainly, who answers for the algorithm — and to make the answer visible.
- [1] A Replication of 'Explaining Why the Computer Says No: Algorithmic Transparency Affects the Perceived Trustworthiness of Automated Decision-Making'. (2025). https://doi.org/10.1111/padm.70015 https://doi.org/10.1111/padm.70015