Companies are making frantic efforts to ensure that they build or procure systems infrastructure that has the aptitude to handle big data as they move into the future. These systems and related accessories ought to allow for complex algorithms and analytics that make more decisions about and on behalf of people. We have gotten to a stage where we have to ‘trust’ the systems to chart the course for our organisations. Yet, with so much now riding on the output of D&A, significant questions are starting to emerge about the trust that we place in the data, the analytics and the controls that underwrite a new way of making decisions.

This piece seeks to interrogate the ‘trust’ aspect that comes with analytics to establish how much if at all businesses are incorporating this fundamental consideration in their decision making. How do they ensure that they develop what KPMG has termed “Trusted Analytics”?

Trust is a rather difficult concept to pin down. It’s even more difficult to measure and quantify. What exactly then is the value of trusted analytics? Will organizations that have earned trustworthiness achieve better results, build better relationships or make better decisions?

In banking, trust has always been central to the relationship between a bank and its customers. And today, data and analytics offers banks an inherent opportunity to create value and build trust. Yet as analytics moves from the back office to the front line, banks will need to ensure that they trust their analytics to ‘do the right thing’ for customers, shareholders and regulators. Insurance, to the same degree, has always been about trust, risk mitigation and protection. But lately, the relationship between data, analytics and trust has changed, posing significant opportunities and challenges for the incumbents in the insurance sector. The relationship between insurers and their customers is fundamentally changing as both partners begin to gain more insight — and more power and uncertainty — through their ownership of rich data, which is granular, personal and streamed in real time.

At issue is that only a few organizations are currently thinking about ‘trust’ in their D&A strategies. For those that are, the concept tends to be quite narrowly focused on accuracy, reliability and security. But with complex ‘black box’ analytics, how do we know that a result is right or that automated decisions are doing the right thing? What does right mean? And to what extent does it matter, and who is the judge? Trust is a growing issue as analytics goes mainstream, not only across most sectors but also for regulators, policymakers and those who safeguard client/ customer rights.

A new, heightened focus on trust is emerging and the stakes are rising. Analytics have increasing financial power but are not guaranteed to be politically or morally neutral. As D&A becomes more opaque, the scope for poor design, errors, underperformance, misuse and unintended, even exploitative consequences increases. Those who can manage trusted analytics will be more confident, both in decision making and client relationships.

Most people have a similar instinct for what trusted data and analytics means. Are the data and the output correct? Is it used in a way that I understand, by people I trust, for a purpose I approve and believe is valuable to me? Is it used for only this purpose, and how would I know if it isn’t, or if something is going wrong? 

There is increasing need to explore some of the critical questions and challenges emerging around trust such as the customer view, trusted data science, policy and regulation and cyber security, among others. At KPMG, we believe that trusted analytics is founded on four key anchors.

The first relates to quality; are the fundamental building blocks of the analytics and data management processes good enough? This includes practices related to the accuracy, provenance and ‘freshness’ of data (nobody wants to receive an offer that is evaluated on personal data that is out of date). 

The second anchor relates to accepted use - is the intended analytical approach appropriate to the context in which it is being used because knowing when and how to appropriately apply data and analytics to various scenarios is key.

The third anchor of trusted analytics relates to whether or not predictions and insights are accurate and reflect reality. For example, the global financial crisis of 2008 was worsened by predictive risk models which were seen as technically correct and accepted by experts but failed hugely in their intended purpose. Finally and fundamentally, organizations will need to ensure that the way they are using the data and the ensuing predictions is ethical and managed with integrity. 

For more information please contact Trust Togarepi.

Trust Togarepi is a Manager in Audit and Assurance at KPMG Namibia. Trust is a Chartered Accountant and has over 8 years’ experience in Financial Services.