Bill Franks, Tom Davenport and Bob Morison of the International Institute for Analytics recently published their 2019 Analytics Predictions & Priorities. They had some great predictions and suggested priorities around the ethics of analytics, the value of data and the use of AI in fraud and cybersecurity. Three predictions, however, touched on our work around Decision Management, including some of the research I have done as a faculty member for IIA such as this on Framing Requirements for Predictive Analytic Projects with Decision Modeling.
Merge AI and Analytics
First, they predicted and strongly urged companies to merge their analytics and AI teams into one organization. Our experience is the same – it’s really helpful to think about AI and Analytics together. It focuses AI on decision-making (where analytics is already focused) not just conversational AI and chatbots, which are otherwise generally the first AI project. It’s not that chatbots aren’t useful, they’re just REALLY different from decision-making AI and are better thought of as user experience technology. Plus most of the challenges with deploying AI are the same as those with analytics – check out this brief on How to Succeed with AI for our thoughts.
Focus on analytic deployment – the last mile
Second, they are concerned about the low rate of successful analytic model deployment – as are we. Too many analytic teams focus on building an analytic model and don’t worry enough about how to operationalize that analytic effectively. As they quote in the paper:
According to the Rexer Data Science Survey, barely 10 to 15% of companies “almost always” deploy results and another 50% only deploy “often.” That leaves 35% to 40% of companies that only occasionally or rarely successfully deploy analytical models. We have encountered some organizations that say their successful deployment rates are less than 10%. [Emphasis theirs]
They urge companies to make production deployment a top priority and have some great suggestions for incenting analytics teams to do so.
We spend a lot of our time helping organizations operationalize the analytics they build and we have found that three things are critical:
- Begin with decisions, not with data. Focusing on the day to day business decisions you need to improve not the data you happen to have gets you thinking about where and how you will need to operationalize your analytics right at the start.
- Begin with an agile analytic deployment platform, not with visualization. Mix and match decisioning technology such as predictive analytics, machine learning, AI and business rules to create a business-led automation strategy.
- Focus on continuous improvement and how you will learn and improve. Don’t try an immediate big-bang, swing-for-the-fences approach – instead focus on getting something deployed and working that you can improve and evolve. This lets business owners get used to analytics and data-driven decisioning and manages technology risk at the same time.
Citizen Data Scientists
Finally they talk about the rise of citizen data scientists and suggest that analytics teams accept and enable these users rather than fighting them. We have some clients who have invested heavily in empowering their existing staff to develop analytic models and become data scientists. One thing they have found really helps is giving those citizen or new data scientists better tools for defining their requirements. Check out this IIA case study on Bringing Clarity to Data Science Projects with Decision Modeling to see how one of them did this.
You can request a copy of the International Institute for Analytics 2019 Analytics Predictions & Priorities here and Tom just wrote a new book The AI Advantage that I recently reviewed. Plus we have a number of assets on our approach on our page for Chief Analytics Officers.