Prescriptive analytics is a new idea for many organisations, although large corporations have been optimising their operational activities for decades. Prescriptive analytics technology solves the basic problem of having lots of variables and constraints, and finding the best solution based on a set objective – among other things. Consider a factory for example with a hundred machines on its production lines, all of which can be configured for several different tasks. The products produced in this factory all generate different levels of profitability and require different machine settings, and consume variable amounts of time on each machine. What is the optimal configuration for the factory so it is producing the most profitable mix of products? Problems of this nature, which might manifest in marketing, purchasing, the supply chain and virtually every aspect of a business, can usually be solved using optimisation technology – a major component n prescriptive analytics.
While optimisation goes back a long way, it has often been overlooked because of the high levels of skill needed to utilise the associated software technologies. This however is starting to change, and just as easy-to-use business intelligence tools are now widely available, so user friendly optimisation technologies are starting to appear. I had a conversation with Kavi Singh of FICO, a leading provider of optimisation solutions, and he confirmed that ease of use is the primary driver of greater use of optimisation technologies. He also confirmed that some large corporations are using optimisation on a near real-time basis, modifying the configuration of resources as conditions change – and to great effect.
Optimisation not only applies in situations where the value of variables are known – as in the example of the factory given above. Many variables in business are unknown for certain – next month’s sales for example. Stochastic optimisation supports levels of unpredictability, or randomness in the inputs and will still find the best solution based on the most likely outcome.
Businesses can start exploring optimisation with the Solver in Excel, or several other options (listed below) which extend Excel’s optimisation capability. For large scale optimisation problems suppliers such as FICO and IBM offer sophisticated tools capable of addressing the largest optimisation problems.
More generally the discipline of operations research, of which optimisation is just a part, is set to make a come back, after skulking in the shadows for a couple of decades.
Suppliers of optimisation technology include:
Frontline Solvers provide a number of Excel Add-Ins, and it is claimed that Premium Solver Pro will solver larger optimization problems with much greater speed (between 2 and 50 times faster). Up to 2000 decision variables can be specified and users can specify their problems with an Excel Solver type utility or a Ribbon and Task Pane interface. Premium Solver Pro automatically chooses the best solver engine based on the model specification. A licence costs US$995.
SolverStudio is a free Add-In for Excel that supports the creation of optimization models using a variety of modelling languages and solvers, including PuLP (included with SolverStudio), AMPL, GMPL, GAMS, Gurobi, COOPR/Pyomo, and SimPy. The models can be created and edited without leaving excel, and the model is stored within the workbook.
This is a an Add-In provided by Lindo Systems and is targeted at large industrial size optimization problems. It addresses linear, nonlinear (convex and nonconvex/Global), quadratic, quadratically constrained, second order cone, stochastic, and integer optimization. Some of the largest problems formulated in What’sBest use over 100,000 variables and it is claimed that execution speeds are still acceptable. A variety of options are available with a starting price of US$495, rising to several thousand dollars if all options are included.
Used in many of the world’s largest corporations and addressing many complex optimisation problems, FICO leads the market in providing optimisation technology that integrates with analytics and business rule based systems. The technology is currently being made available to a broader audience via the FICO Analytic Cloud.
Since 2009 ILOG has been part of IBM and completes an impressive array of optimization capabilities. IBM ILOG CPLEX Optimization Studio provides an environment for model development via mathematical and constraint programming. IBM ILOG ODM Enterprise is a platform for building optimization based planning and scheduling applications. Supply Chain Optimization is also offered as a particular solution.
Enterprise Optimizer has several components. Workstation supports rapid development of decision-support applications. EO Server supports deployment of planning and analytics solutions. EO IBP Framework facilitates integrated business planning. Strong integration with Microsoft systems architectures.