Imagine you are the CEO of a major corporation. You’re sitting in a conference room, surrounded by bright people, facing down a strategic decision. You need to know whether your company should make a big investment in an emerging sector or not. So, you turn to your best data analysts, and you ask them what you should do?
If you are a layman, you might imagine that this is a reasonable request. After decades of amassing data on every conceivable aspect of your business and marrying that with the seemingly endless statistics compounded by governments and third parties, an analyst should be able to give you a reasonable answer to a pressing business question. But it doesn’t work that way. While data tends to be extremely good at telling us what might happen in the future, it is largely powerless to tell us what will happen when we make a large, strategic decision.
A recent survey of business leaders, for example, found that although 99% of Fortune 1000 companies are investing in data and AI, only 30% feel they have a well-articulated data strategy.
In data terms, this is the difference between predictive and prescriptive analytics. The former, which is used by businesses every day, tell you what is likely to happen. It’s important to note that it’s not a perfect crystal ball, but even knowing the future within a range of certainty is incredibly useful. It can tell us which prospects are likely customers, who are likely to buy something, or when your operations might become overloaded. All of that can inform decisions, especially in limited, tactical ways.
However, it doesn’t address the larger questions that many would like answered. Those questions fall into the domain of prescriptive analytics, which seeks to understand the outcome of a particular action. Prescriptive analytics doesn’t merely stop at the likely future, it tries to identify the best actions you can undertake to affect that future. According to Gartner, it answers questions like “what can we do to have this happen” or “What should we do?”
On a small scale, prescriptive analytics is seeing increasingly widespread use. One of the better-known examples is Amazon’s pre-positioning of products based on predictions of demands. You can also find it working well in industrial maintenance. Companies are using it to make decisions about what to do and where to send workers before they are needed. Such anticipation has also been scaled into systems that greatly increase efficiency and reliability.
With anything like prescriptive analytics, however, it’s extremely important to make a distinction between point solutions that resolve specific situations, and strategic solutions, which can answer any unanticipated decisions you have to make. Prescriptive analytics solutions today tend to “ride on rails.” They work very well on a particular task but cannot generalize beyond it.
Of course, the real promise of prescriptive analytics is not to resolve limited difficulties but to be generally useful to the business as a whole. To get there, we will need to resolve to solve at least four outstanding issues:
Relevant data. Right now, prescriptive analytics tends to be effective when it has large amounts of highly relevant and useful data are readily available. For example, geophysicists are currently using it to find the optimal location for dating oil and gas wells. To do so, they employ models that incorporate ocean surveys, seismographic data, capital cost information, as well as unstructured data sources, such as images taken inside test wells. To make a single decision they analyze literally terabytes of data. Of course, such decisions may also save millions of dollars, so it’s important to get them right.
A similar amount of data is also needed for more generalized business decision-making. We need to be able to incorporate everything from demographic databases to videos and social metrics into the pool for analysis. Of course, this is possible today, but it can typically only be done when compiled by a skilled analyst in products like Excel or more sophisticated tools like SAS. Such ad hoc solutions can deliver value, but they require a considerable investment that isn’t available or reasonable for most decisions.
Collections of ready-made models. While we can create prescriptive models for specific purposes, businesses will need to have ready access to a wide range of models that address the wide range of decisions a business needs to make. While no one could ever anticipate all potential use cases, the world has plenty of examples of shared resources used to overcome analogous problems. Today, it is possible to collect a database of relevant, working models that can cover most likely economic conditions and circumstances — especially if a marketplace could be developed for sharing and selling them.
Modeling tools. Something, for the course, will need to bring this together: an analytics platform that can understand both the inputs of investment, the ongoing financial activities of the company, and other inputs needed to understand bottom-line results. Such a platform should standardize and categorize models — as well as make them available in libraries for analysts.
Analytics. Obviously, we would also need a suite of superior analytics tools that could be used for all kinds of simulations that might be required. This is the smallest of the hurdles we need to overcome, as plenty of sophisticated analytics platforms already exist that is able to undertake these tasks.
The simple fact is that if we take the right approach, we are on the brink of a much wider application for prescriptive analytics than currently exists, and one that would greatly facilitate data-driven decision-making. While AI-driven predictive insight has aided businesses in innumerable ways, we are running up against its limitations every day. Perhaps, with the application of new approaches, CEOs will finally get the value they’ve been seeking from the massive amounts of data they’ve been amassing. In that way, when it comes time to make big strategic decisions around the conference table, analysts will be leading the way.
Andrew Usacheff is a serial entrepreneur with extensive experience in the fuel and energy complex. He is the CEO of AssetData Group - an international consulting group and software developer for business assets analytics. He is an expert in supply chain management (Accenture Academy, Energy Delta Institute), project management (Adam Smith Training), investment project evaluation (E&Y), and energy conservation.