Predicting which businesses will survive: Marrying industry support measures with advanced machine learning - COVID-19 Blog | Deloitte Australia has been saved
Limited functionality available
The COVID-19 pandemic has upended daily life for many Australian businesses, and even as the country takes its first tentative steps out of lockdown, there remains significant uncertainty about the future.
Quickly establishing the delivery capabilities required to distribute industry-wide stimulus in an evidence-based manner is a crucial focus area for many public sector agencies over the coming months. And it must be done in a transparent and repeatable way that ensures the shape and scale of industry support meets the needs of businesses and workers alike. It must also ensure that adequate mechanisms are put in place to support the long-term monitoring and evaluation requirements expected when providing stimulus at this scale.
Sadly, we are already seeing media reports of some dubious claims across a number of these packages, and while not unexpected – they serve to underscore the importance of making the right decisions about what businesses receive support and how much is acquitted. The potential for sandbagging (putting aside money for future purposes), non-compliance with reporting obligations or the mismanagement of funds all serve to highlight the risks associated with implementing these packages too quickly or without enough supporting evidence when assessing applications.
The past few months have highlighted just how closely linked the financial sustainability of businesses is to broader economic conditions, and these relationships can be reflected in models that help to measure solvency and access to capital. How governments administer these programs will dictate both the benefits, and indeed costs, that influence productivity; using the right tools will be crucial for maximising the former and minimising the latter. And herein lies the challenge – traditional approaches to undertaking detailed financial assessments pivot on the availability of skilled resources to complete this work and unfortunately, the nature of the COVID-19 crisis has upended what we know as normal.
To this end, it has been pleasing to see a renewed focus by governments on the use of advanced analytics – particularly credit risk modelling – in supporting efforts to assess business applications under these packages. Yet we see both opportunity and necessity for government to be even bolder in its use of models. Indeed, to adequately and urgently scale and address the volume of support claims being made requires a model-led approach, one that decouples capacity from traditional human capability, lest we risk seeing businesses suffer even further with unnecessary delays.
And herein lies the opportunity, by augmenting traditional credit risk modelling approaches – which are very linear by nature – with advanced machine learning techniques, we can enable a more direct consideration of the economic context in which these application decisions are being made, and allow them to be tracked over time to understand whether they were correct. Enter the Deloitte Model to Assess Financial Sustainability (MAFS), a machine learning model specifically developed as a decision support tool for dispensing industry stimulus.
The Deloitte MAFS model provides this functionality through its ability to not only assess and categorise the financial sustainability of individual businesses, but to extrapolate this to create sector or region-based views that provide governments with the ability to assess individualised risk at scale and that generates the insights required to inform targeted response measures.
In this way, the models don’t just take account of the financial position and credit history of the individual business applying for support but consider the relative financial health of other similar companies in assessing risk so that any economy or industry-wide headwinds are factored into the overall decisions.
Additionally, with the support of revenue agencies, the important task of monitoring the impacts of policy decisions on individual organisations can also be effectively measured. So too can this data be used to police the progress of individual organisations in keeping the commitments given when taking receipt of industry assistance.
More sophisticated machine learning models provide governments with opportunities to mitigate concerns of regulatory burden and reduce implementation costs, as well as minimise any unintended consequences for allocative efficiency through insights.
There will always be limits to what we can know, with experts now predicting that the short-term uncertainties of COVID-19 are fundamentally different to what may occur over the longer-term. Leveraging more sophisticated machine learning models in assessing business applications for support provides strong governance and oversight by drawing on data-driven insights to minimise risks and maximise outcomes – despite the uncertainty.
Will is a Partner within Deloitte's Analytics & AI Consulting Practice where he is focused on helping organisations mature their insight-driven decision-making capabilities. As a relentless integrator of capability, Will is passionate about leveraging the breadth and depth of Deloitte to deliver outstanding solutions to our clients’ most complex problems. He believes that processes for collecting, storing and managing information have no inherent value; information only becomes truly valuable when it is enriched, combined and applied in the generation of insights that directly enable targeted business decision-making.
Leighton is a Senior Manager in Deloitte Consulting’s Analytics & AI Practice where he works with public sector clients to deliver insight-driven tools that support policy development, implementation and evaluation. He is passionate about leveraging quantitative and qualitative frameworks, models and Artificial Intelligence technologies to help solve the complex public policy and service delivery challenges facing government today. At Deloitte, Leighton has supported public sector clients to design and implement analytics operating models, develop public-facing strategies, and embed Artificial Intelligence capabilities into policy and regulatory delivery functions.