Since the dawn of the computer age, the promise of technology (at least in the eyes of the end user, and frequently in the eyes of the corporate buyer too) has never quite matched the reality of the experience. Over the years we’ve been heavily sold on features and functionality but have often then struggled to drive expected value from our technology investment. Nowhere has this been more true than in the public sector where a number of high profile technology investments have failed to deliver required value or even been completely abandoned.
Having recently left the public sector, I well understand the nervousness of those tasked with investing in technology about being truly certain of what they are going to get in return for their investment. But my significant fear is that this reticence will be a direct barrier to the UK Public Sector’s ability to make sure that the latest data analytics approaches can be used in what is critical work for the country.
Recently, more and more reliance has been placed on data and analytics and with it, the expectation that analytics will deliver phenomenal insight and return for the public sector has grown. At the same time, investment is a challenge and some high-profile issues with analysis have undermined the belief in the reliability of analytics. Recent surveys (run by KPMG in 2018 and SAS early in 2019) have shown that senior leaders have a significant lack of trust in their analytics, with only 30 per cent willing to say that they trust their data and analytics. If we are to expect leaders to make data-driven decisions then this trust gap must be closed and the ability to set expectations and explain the likely value outcome play a large role in this.
Can you explain it simply?
I explored this theme recently at the 2019 GSS conference as a speaker, focusing on how to ensure that confidence in data analytics is, at worst, maintained and, at best, improved. In establishing trust for data analytics projects, my keynote hinged on a central tenet; to be able to trust a solution it must be understood technically, explained simply and linked to a business question. Let me explore each of those points here in a bit more detail.
Firstly, when talking about the need for technical understanding, I’m of course not suggesting that any one receiving insights from their data has to actually fully understand the technical method to analyse the data to create those insights! Instead, I believe that technical understanding is only really needed by the data analyst or data scientist but that it is fundamental to then subsequently achieving confidence from the stakeholders. Any decent data analyst or scientist has to know their data sources and critically, their limitations. I’d argue also that they need to use the simplest approach to gain the outcome and accuracy needed, rather than simply default to their favoured analytics model.
Taking this approach is invaluable in then moving to the second key element in driving success with analytics; being able to explain outcomes in simple terms. Trust in a solution will only come for those who are being asked to invest in it or use it, if the explanation of what is going to happen is a clear and cogent one. Key to achieving this is ‘explainability’ (why and how the expected outcome is going to happen) and ‘interpretability’ (how effectively you can predict the outcome). Both must be factored into any solution scoping exercise, to ensure clarity of thought in preparing the solution and precise expectation setting for the likely outcome. After all, as Albert Einstein said; ‘if you can’t explain it simply, you don’t understand it well enough.’
Exploiting the obvious power of data analytics
The last key element for establishing trust in analytics projects is, perhaps the most crucial; what business problem is this project seeking to solve? Defining this question is sometimes the hardest but is always the most important part of a project, and it needs to do three things. First, it must be an obvious follow on from any stakeholders situation and need. Second, while there will always be multiple potential questions that could be asked with data, delivering what the stakeholder actually wants is critical! And thirdly, the desired outcome must be attainable; I’d always advocate the narrowing of scope to ensure that an ambitious desired outcome becomes a realistic one.
There is no reason why the Public Sector should not be able to exploit the obvious power of data analytics both today and in the future, as ever more powerful techniques come on stream. But in a world where manipulating data to drive key insights to inform ever better decision making is critical, the Public Sector must have confidence that it can derive value from investing in data analytics. By using some of the approaches outlined above, I’m certain that confidence in data analytics can only increase and drive more and more value for the UK’s Public Sector.
Dean Wood, Deputy Director, Data Science, Mango Solutions