Helping citizens with predictive analytics for the public sector

Advanced predictive analytics have been successfully integrated across a wide range of sectors in recent years, particularly those with an academic focus such as science, technology, medicine and engineering.

These tools have led to significant improvements in the collection and use of data. However, in the public sector, at both central and local government level, there has been some hesitancy and fear around incorporating predictive modelling into their operations.

Due to lack of trust and misunderstandings about the way data is consumed and analytics is performed, many public sector organisations are reluctant to rely on algorithms to serve citizens. But if analytics are used in the right way, they can provide an opportunity for councils to improve the services they offer and better protect vulnerable citizens.

In this article we discuss several perceived barriers to success that can be overcome with the right consideration and processes.

Improving customer journeys

Since the onset of the pandemic in early 2020, demand for council services has increased and many people have found their circumstances have changed. Job losses and difficulties reaching social care services due to Covid-19 restrictions have led to increased financial pressures and challenges for citizens.

Council tax arrears have risen to £4.4bn as of March 2021, while a quarter of adults have now been identified as financially at risk. Councils are under pressure to reduce the revenue gap, while supporting the needs of the most vulnerable residents.

High-quality analytics can help local councils to identify residents at financial risk early on, preventing some of these issues. For example, credit history data can be used to determine which residents have missed payments in the past, putting them at greater risk of problems.

Once identified, councils can intervene and support people with payment plans suited to their needs, helping them to avoid any late payment penalties.

Personalised customer journeys will help people to feel more confident about the payment processes, knowing that help is at hand if they find themselves in difficulty. In turn, local authorities will know they are better meeting the needs of the most vulnerable residents.

Some considerations for use

Organisations do not always have the right data to make predictive analytics accurate and, therefore, valuable. However, there are external data sources available for organisations who are keen to increase the quality of their services via innovative data techniques.

Whichever sources are chosen, councils need to ensure that the analytics they’re using are accurate and meaningful. To do this they will need to work closely with data scientists and specialists who are able to develop models that are sensitive to the needs and concerns of individuals, while delivering the information that council employees need to provide the very best services.

The data used must have a strong scientific base and model features should be carefully curated using expert knowledge. For example, in healthcare, factors like obesity and smoking are used to determine someone’s heart disease risk, because it’s proven that these issues are related to the likelihood of a person developing illness. The same model can be applied to financial risk, but must be carefully assessed to ensure the right factors are taken into consideration.

In addition to ensuring that the technology and data is flowing efficiently, local authorities need to ensure that employees are equipped with the soft skills they need to support citizens directly, always remaining empathetic to their situations. Moreover, and essential to return on investment (ROI), the output of the analytics should be integrated with IT infrastructure to build end-to-end automated solutions. For example, predicting the correct contact method for payment nudges should be connected with communications platforms.

An ethical code

Additional scrutiny is always placed on analytics being used in the public sector, with questions being raised about the ethics of data use. While governments have made U-turns in the past after algorithms have delivered divisive output, it’s important to keep in mind how rapidly these technologies are developing and how quickly our knowledge in these areas is growing.

There are also several methods that can be used to ensure that analytics are used to improve public services and help to meet people’s needs, whilst maintaining an ethical approach.

Predictive analytics as a technology alone does not have the ability to be unethical. But when they’re in use, human bias and error is a risk. In the first instance it’s vital to understand what analytics are being used for, and that the public knows the reasoning behind this. Secondly we need a thorough understanding of the data selection for the model at the design stage and, finally, to ensure that the output is used in the right way.

Clean and complete data

Data selection is paramount for building predictive models. A mix of internal variables and external data sources may be necessary, but the data should also be of sufficient quality and cleanliness. The handling of missing data is a scientific discipline in itself and should be carefully managed to make sure the data remains accurate, and unbiased.

Model selection is the second crucial part. Local authorities will need to trial multiple model types to achieve maximum efficiency, as well as running follow-up tests to check for bias or variation. Consideration should be given to whether algorithms used should be auditable, so that they can be manually validated at different points and called upon to answer for their output.

Governance

Governance should focus on what the data is being used for. Frameworks on ethical guidelines should be established. If there could be any negative effects, for example in cases where there might be changes to people’s benefits, there will need to be extra scrutiny to ensure that analytics don’t lead to inadequate decision making that could impact individuals.

To prevent poor outcomes, councils should embed frameworks and introduce ethics boards to critically discuss and analyse new data models before they’re signed off. A certain level of academic level scrutiny should be placed on data projects to make sure they’re being conducted to the correct ethical standards.

Conclusion

Predictive analytics are underused, and it’s clear that we need to tackle some of the barriers around this. We know that if used correctly, data and AI has the power to significantly improve services, better supporting both councils and their local residents. Through good governance and clear, ethical communication around their use, we can create more efficient systems that truly benefit people.