Algorithms and segmentation. Reflections on the ethics and practical use in social housing work with tenants.

By Hannah Absalom @HousingHannah

Image of a blackboard with maths symbols on. By Chuk Yong – Pixabay

Algorithms, segmentation, big data, predictive analytics are buzzwords that sound mysterious and complicated; possible silver bullets to solve some of the long-standing challenges in our sector. This article focuses on the use of algorithms and customer segmentation, approaches used in the private sector to understand customer preferences.

This article will explore the ethics of using these approaches in work with social housing tenants (repairs and maintenance is a different topic and not the focus here), as I presume the social housing sector is different to ‘true’ private sector organisations so the ethical and practical issues may be different.

The article outlines what algorithms and segmentation are, discusses why they applied in social housing work, outlines the possible benefits and risks and discusses possible practical actions to ameliorate risks and maximise benefits. It closes by outlining more challenging areas that need further consideration.

Algorithms and Segmentation – What are they?

Algorithms are a process or set of rules used in calculations or problem solving by a computer. They are a way of reducing real world complexity to a set of procedures that can be used to manage and predict real world activity. Some are simple, in that the same rules are applied repeatedly until they are reviewed by a human. Others are complex and capable of ’self-learning’ based on how people interact with them. This process of self-learning influences the choices we make. If you have ever used Google, Netflix or Amazon, you have interacted with complex algorithms.

Outside of making sales in the private sector, there are other uses by national government and large private companies in flagging potential fraud and other crimes and in modelling behaviours in closed systems such as travel networks in cities. The approach we are starting to see used in social housing tends to be imported from the commercial private sector with the intent of obtaining a better understanding of tenants in order to intelligently target services and so improve service efficiency.

The process of segmentation usually occurs first, which is the sorting and categorising of customers into different types, so the best algorithm can be applied to that customer profile. While private sector organisations tend to segment and then apply algorithms, it is still very early days in social housing, so it is often human beings who decide on the best course of action based on the segmentation. For example, you could segment customers into ‘technology savvy’ and ‘technology unconfident’ and use that information to change how you communicate; sending texts and emails to the first and targeting letters and phone calls to the latter group.

Self-learning algorithms can be used to highly individualise services, so go to a deeper level than segmentation, we are not at this point in our sector and may never be, as such processes require constant interaction with the customer and we do not have enough ‘data points’ (interactions) for this to occur. It is relevant to note that algorithms are still initially designed by a human being even if they later ‘self’ learn based on interactions with customers.

Why are algorithms and segmentation getting used in housing?

There are various reasons why algorithms and segmentation are used in housing, but a key driver seems to be seeking to overcome the issue of tenant stigma. This has caused some organisations to reflect on how they understand and interact with tenants, and to want to do this better with a view to reducing stigma. In-depth segmentation offers an opportunity to engage with tenants on a more nuanced level when building the profiles. The profiles can be used to challenge some of the negative stereotypes surrounding social housing tenants in the UK.

Another factor can be that social housing providers, via mergers and acquisitions, have increased both their geographical and social distance from tenants. Larger organisations think they can rely less on developing and supporting human based knowledge of tenants and localities and are moving towards more technology and number-based approaches to fill these gaps. Algorithms and segmentation have particular appeal in appearing to be consistent, efficient and in maintaining the tenant and landlord relationship from a distance.

A further motivation is poor service and desire to improve service efficiency. Segmentation alone allows for more efficient and targeted communications and algorithms allow for further service refinement, especially of online services. The sector seems to be looking to the private sector for practices to import that will aid its service improvement. A key tension here though is social landlords are often seeking to reduce interaction with tenants in a drive to save money, while the private sector seeks to increase contact with its customers to drive sales.

There are also cultural drivers in a shift from an expectation of one size fits all to one of tailored services. This is in part driven by changes with social provision from the State, for decades there’s been a move away from monolithic and standardised welfare systems that operate at scale, to an expectation that people look after themselves and when needed support is provided by specialised local service providers. Consumer culture is increasingly geared towards individualisation which is shaping our consumer tastes and preferences and these expectations may be finding their way into the social housing sector. Catalysing this cultural shift to individualisation is the digital revolution, which provides both the means to, and desired endpoint of increasing individualisation.

How social housing is different and some thoughts on the benefits and risks of algorithms and segmentation for the sector

With new and innovative methods come new benefits and risks that require consideration. There are currently no ethical guidelines that are specific to our sector on how to apply these approaches.

We have some key differences to the commercial sector including our ‘product’ is rationed and we allocate it based on need, not want. Most housing providers are overwhelmed with applicants for properties, we are not competing to increase our market share in social renting. We are living with a legacy of quasi-marketisation which makes it difficult to locate our relationship with tenants and is reflected in the discussions of how we name them; are they tenants, customers, or service users?

The sectors ‘true’ commerciality lies in its appeal to the financial market to obtain loans and other investments and to private buyers of properties, as these relationships fund the sector’s social purpose of providing social housing.

We have a complex and changing relationship with national and local government which means our regulatory landscape can change quickly and in surprising ways, impacting heavily on how we structure our organisations.

Hopefully the above summary outlines how the social housing sector is assembled differently to ‘true’ commercial sectors such as finance and retail. I suggest these differences mean we need to think about how the discussed methods are applied and the following list of potential benefits and risks are intended to start conversations, not be treated as facts and applied without reflection.

Some of the benefits and risks can exist before, during and after the application of algorithms and segmentation and may also reflect wider complexities with a ‘computerised’ approach to managing people, so may also apply to organisations seeking to digitise their interactions with tenants and applicants.

Potential BenefitsPotential Risks
Better understanding of tenant variation so helping to challenge stigmaWidening and deepening of customer relationshipsMore efficient targeting of servicesAllow for smarter interventions such as, better timed support interventionsPredicting problems before they happen or before they are reported so they can be planned for rather than reacted toBring much needed new knowledge and skills to the sectorForces review of often awful data and a drive to make it accurateA cultural shift from business reporting to business intelligenceTenants feel better ‘seen’ by landlords; trust and quality of interactions improve.Can be combined with human and context centred approaches to better understand complex social problems that impact some communities, i.e. long-term unemployment, successful tenancy maintenance for care leavers.Provides an opportunity to reflect on organisation purpose and direction.Provides a ‘health check’ on an organisation’s operational and strategic ethics.Creation of employment and training roles for digitally savvy tenants. We know many work-based programmes lack aspiration, the more technically advanced our sector gets, the better we can support talented tenants realise more complex ambitions.  Creating new labels that stigmatise; rather than the binary deserving/undeserving divide, we have more nuanced ‘boxes’ that still locate tenants within this divide.The model becomes perceived as ‘the truth’ of tenant lives obscuring the alternative notion that there are multiple truths that constantly change .  Rather than being useful tools to inform work, they dictate how the work should be done – effectively a ‘computer says no’ culture and a disempowerment and erosion of trust in staff decision making.The approaches are based on reductive logic which means complexity and context are stripped out. Our work in housing is uniquely contextual due to the material reality of the stock and the historical legacy of our neighbourhoods. Approaches that account for world experiences in relation to context and local history and resources may have more coherence and value.Lack of understanding of ‘in the moment’ complexity.Shuts down other ways of knowing from fields outside of the private sector that may fit better. Risks vary across service areas, Allocations is a high risk area – as there’s a temptation to exclude tenants deemed risky or to impose untested and time consuming interventions such as ‘tenancy ready’ courses on more vulnerable applicants.The  asymmetric power relationship with tenants and applicants, many of whom feel unable to challenge decisions. Algorithms and segmentation influenced choices are harder to challenge than those made by humans and existing in policy and procedure.Obvious attempts to change behaviour can lead to avoidance strategies, so damaging tenant relationship.Hidden attempts to change behaviour are both ethically problematic and a risk to organisation reputation.Uneven spread of benefits and harms across individuals, groups and organisations.  


While the list of potential risks is longer than the potential benefits, this is partly to reflect that change and innovation is always complex. This section will discuss some possible practical approaches and also outline areas that require further consideration and pose bigger challenges to the social housing sector.

Some practical approaches to the challenges posed

Baked-in discrimination

Algorithms and segmentation can ‘build in’ discrimination. This is seen in the American justice system where black defendants received longer custodial sentences than white due to algorithm bias.

To challenge and monitor discrimination in algorithms there’s a need to collect sensitive personal data to assess the algorithm, which is of course a complexity of its own as collecting sensitive data is fraught with problems. For example, people may hide their sexual orientation from social landlords.

It’s suggested that equality impact assessments will help capture and reduce discrimination risks. As a former practitioner I have been involved in these assessments and know they are often treated as a tick box exercise to rubber stamp pre-determined decisions. Staff with little to no research skills often undertake the assessments so information pertinent to the process is often missed. A possible solution could be to carry out a properly researched equality impact assessment informed by some of the bias reducing findings of behavioural science to improve the quality of the process. This would need to be followed up with long-term evaluation that focuses on impact to the tenants’ day to day life and emotional wellbeing not just service-based measures. The long-term review process would help to identify both bias and also ‘inflection points’ – locations in a complex system where change can have the most impact.

Giving tenants the choice to opt out

Digital service portals give landlords a location to explain and provide assurances and tenants the ability to dynamically respond to permissions about data collection and to opt out of having such approaches applied to them. This second step goes well beyond GDPR requirements, but would resolve the problem of algorithms, segmentation and other innovative methods being used invisibly and encourage a conversation with tenants about the use of advanced methods.  

Views of the use of algorithms and segmentation is variable, with young people being more accepting than older people. There is little research into the impact of such approaches on vulnerable populations, but it is acknowledged that low income households tend to be less savvy with data privacy management than more well-off households. There are obviously large populations of tenants who experience complex social and digital exclusion and also plenty with mental bandwidth-consuming day-to-day pressures that may reduce their time and ability to engage with decisions on how data and innovative techniques are used. There is a responsibility on organisations to engage with excluded profiles and to utilise behavioural insights to make the information and decision making as easy as possible for tenants and applicants.

Authentic opportunities to challenge the process

There needs to be a freedom for both staff and tenants to challenge the pathways created. Real life complexity is stripped out by such approaches, and there will always be situations that arise that do not fit in any box. Staff need to maintain the freedom to deviate away from recommendations and tenants and applicants the freedom to challenge, bearing in mind they may find this especially difficult to do as they may fear a penalty for non-compliance.

There are other practical approaches to be developed to meet some of the challenges raised. It would be excellent to hear from tenants, landlords and invested third parties in how possible benefits could be ethically maximised and the risks reduced. There remain complex problems where practical solutions are elusive, and these are discussed next.

More complex challenges with no prescriptive suggestions

Complexity, networks and innovation

A key challenge is a cultural one. Organisations are complex networks that sit in an even more complicated web of relationships with other organisations and individuals. English organisations have a tendency not to see their networked and complex contextualised locations and to focus on methods and approaches that maintain the illusion of simplicity and control. This means that the ripple out effect of the application of innovative practices in seemingly discrete service areas can go unseen until the damage has already been done. This is not an excuse to stop innovating and trying new approaches, I am instead proposing that organisations improve their knowledge about networked theories and ethics, knowledge available from overseas practice and academic research. True innovation involves making mistakes, acknowledging sector complexity is a key step towards developing a culture of thoughtful innovation.

Tenant involvement in an age of complexity

The new approaches offer both potential benefits and risks for tenant involvement. For example, segmentation can help target service improvements where they are needed the most. However, as the new techniques tend to require highly personal types of data, there is a need to have high levels of trust with tenants in order to obtain the required data. Even if sensitive data is collected accurately and carefully, there is a need for tenants to see that the information is being used for their benefit, otherwise trust can be lost. Effectively algorithms and segmentation pose new challenges to how we ‘do’ tenant involvement; the more complicated and personal the techniques get, the more challenging true accountability and scrutiny will become. Landlords and sector bodies need to ensure tenants are involved with shaping how new approaches are developed and applied.


Algorithms, segmentation and other related practices of the digital age offer social landlords a new range of tools with their work with tenants. To ensure the that the tools are fit for purpose and don’t compound or add to the challenges facing the sector, there needs to be clarity in the purpose of social housing at different scales; at the level of the customer, the organisation and also a national sense of purpose in social housing. This starts by asking questions about what our purpose is, questions such as ‘are we social organisations using private methods to achieve ‘efficiency’ or private organisations with a social purpose?’ so we can start to answer them coherently.

Without a clear and articulated sense of purpose there is a risk that the new ethical challenges posed by the new methods go unacknowledged, potentially causing unintended harm to tenants and damage to organisation reputation.  Engaging with and developing on the conversation points opened up by this article would be a great starting point to ensuring algorithms, segmentation and other innovative practices are given full thought and consideration before, during and after social housing work with tenants.

About the Author

Hannah Absalom is a co-founder of Social Housing Matters, was raised in council housing in the North of England and has worked in the sector for 18 years in a variety of roles. She is now studying an ESRC funded PhD into the use of Behavioural Insights in Social housing in England and the Netherlands.

You can contact her at and follow her on Twitter @HousingHannah


Leave a Reply