Business Intelligence or Business Analytics ?
The simplest decision support system, used in many businesses, is data visualization. An Excel spreadsheet does the trick, and in many cases it is the first “decision-support tool” used. But it offers little help. As the business grows, Excel spreadsheets are replaced by “cubes” aggregating performance indicators that make it simpler to compare products, customer segments, or business units, and facilitate decisions. This method goes under the pompous name of “Business Intelligence” (because it sounds impressive to customers), but in the end it merely presents the data from an angle that helps you make choices.
While visualizing indicators is sufficient for observing gaps and focusing attention, sometimes you need to go further in order to measure the impact of a decision and quantify your choice. This “simulation” of the decision’s consequences – this optimization of the choice – goes beyond Business Intelligence and enters the realm of Business Analytics.
While Business Intelligence helps view the past and previous activities, Business Analytics projects decision-makers into the future of their decisions and helps them with the process. It is augmented reality, similar to the GPS navigation system in your car that helps you find the way (often the shortest path) to your destination.
In our first example, in order to determine the impact of closing site A, the company may need to “optimally” re-route customers previously served by A to either B or C, and to route incoming streams from production areas. Determining the best “routing”, and applying it to three hypotheses (close A, close B, close C) in order to calculate the most efficient solution, is a combinatorial optimization problem for which “decision support system” are available – some of it generic – and applicable to logistics networks for applications ranging from yoghurt and milk to oil, cables, and sodas. “Scheduling systems” are used when decisions need to be scheduled over time (programmed site closings, investment decisions, delivery dates).
Deciding despite uncertainty
The vast majority of human or industrial activities is subject to uncertainty. For a manager, decisions usually entail some degree of uncertainty. Factors such as the price of oil, electricity, and raw materials are also subject to market tensions and volatility that decision-makers cannot ignore. The weather can also disrupt certain businesses. Scheduling systems cannot ignore factors such as the employee absence rate and turnover. Similarly, decision support tools for pricing of e-commerce sites, passenger transportation, hotels and recreational centres must consider the uncertainty of human behaviour.
Where uncertainty is involved, optimization algorithms need to apply more robust solutions involving a different type of optimization in order to reflect the decision-maker’s strategy: how much risk to accept, if any.
For our consultants, decision support with uncertainty usually starts by representing the uncertainty via hypotheses, decision trees, future scenarios, and forecast ranges. These hypotheses, often backed by data from the past, add further complexity to both the decision-making process and the decision-support systems used. This is called stochastic optimization, or stochastic optimization with recourse when the choices of year N+1 may include some of the uncertainties observed during year N. Once the decision is made, it becomes new data and is added to the hypotheses used for making future decisions.