Should’ve, Could’ve, Would’ve – Making the Optimal Decision.

In a previous blog I talked about the value of machine learning and how it could help organizations by making smarter predictions by continually learning and adapting models as it consumed new interactions, transaction and data.  I compared that to how my son embraced learning about the world around him to become gradually smarter, more knowledgeable.  But that doesn’t always mean he is going to make the best decision because he may not have all the information or foresee or correlate past events.

So, I guess there are times when we wonder – or get asked by others – whether the decision we just made was the best possible one.  Think of when you made your last hi-tech or car purchase.  It’s often difficult to judge at the time if it was the best choice as there are many parameters involved including (but not limited to) logic, price, best value, best meets our needs and wants, emotion, political.  How many times have you sought to justify that purchase immediately after by doing even more research on reviews by others?  Don’t worry. It’s a natural human emotion known as post purchase cognitive dissonance.

The collective name for this capability is called prescriptive analytics and it relates to predictive analytics as shown in figure 1 below.


Figure 1: Descriptive, predictive and prescriptive analytics relationship.

Making that optimal decision at a business or boardroom level becomes even more exacerbating and complex, involving many people and large sums of money. Every decision needs to balance risk with cost and with benefit.  Quite often there are conflicts of interest, bias, power struggles involving political and emotional agendas that can result in suboptimal decisions being made for the business.  Oh the frailties and flaws of humankind!   The crux of the matter is there are far too many parameters, data points, correlations and patterns for us humans to be aware of to make the optimal decision for every transaction and interaction.  So there are times we need these capabilities to augment and balance our own judgements.  These decisions can range from a simple purchase or where best to distribute warehouse stock, where to place emergency services in the event of a pending disaster or be financial focused or risk based through to decisions that ultimately help preserve or save lives in the medical field.

Decision Optimization at every transaction.

IBM has been providing decision optimization for many years as part of the business logic within applications with products such as IBM Decision Optimization and its market leading CPLEX Optimizer engines.  Together these offerings are providing a collaborative environment for key personas to deliver a powerful solution application to line of business users to make better and faster decisions in planning, scheduling and resource assignment. In short, the aim of Decision Optimization is to remove the guess work from making a decision by “prescribing” and automating the best decision for you. Just to put your mind at rest – the business user / planner still has the ability to change that decision, interact with it – it is something which is recommended, but the end-user can still own that choice of whether the decision will be operationalized or not.

There are three key personas that are involved in building a decision optimization solution as part of the optimization application development cycle (figure #2) namely the business analyst, operation research (OR) expert, application developer – and the LOB consumer of course.


Figure 2: Optimization Application Development Cycle.

Together these personas have a broad range of responsibilities and needs such as:

  • Create optimization models / algorithms.
  • Use APIs to embed in decision making applications.
  • Solve models with the IBM CPLEX Optimizers.
  • Analyze results.
  • Use Rest APIs to embed in decision making applications
  • A Collaborative development environment to build & deploy enterprise-scale optimization based applications
  • Visualize trade-offs across multiple plans, scenarios, KPIs.
  • Optimize costs vs robustness.
  • Recommend decisions hedged against data uncertainty.

You don’t have to be an Operations Research Expert to use IBM Decision Optimization.

The goal is for decision optimization to be consumable making this powerful technology available to decision makers who are not necessarily trained in mathematical modeling.  Business managers need to be able to use their own business language to define a decision model.  To support this, the IBM Decision Optimization R&D team created the proof-of-concept for “Cognitive Optimization”.  The goal of Cognitive Optimization is to allow business users to directly create and work with Decision Optimization models, without requiring the intervention of a mathematical whiz.    It uses a combination of the business data and the business user input, in natural language, to figure the business intent, to suggest potential decision models, and then to use those models for “what-if” and trade-off analysis.”

Similar with what was announced for the Watson Machine Learning service, Optimization as-a-Service will be also available as part of the data science experience (DSX), in addition to being a standalone cloud, hybrid, and on-premise offering, for the personas mentioned above to experience decision optimization as part of a prescriptive analytics solution as a stand-alone service.   If you consider the high level flow of optimization as-a-service it looks very similar to machine learning as-a-service as shown in figure #3 and, as such, they share many commonalities which makes DSX an ideal collaborative environment for machine learning and operations research practitioners.


Figure 3: Optimization-as-a-Service high level workflow

Clear business benefits reducing time, risks and costs.

In summary the combination of IBM Watson Machine Learning and the Optimization as-a-service capability can help organizations across every industry make progressively smarter cognitive insights and act on optimized decisions while removing the human frailties of emotional, political and personal bias.

Organizations are already experiencing the benefits. For example, a global tire manufacturer uses IBM decision optimization solutions to help:

  • Optimize long-term production planning, considering up to 10 million constraints across all products and plants
  • Predict when major machine bottlenecks are likely to occur, enabling staff to take corrective action early on
  • Drive smarter decision-making with 30 times more what-if scenarios evaluated
  • Save up to 30% of planners’ time, allowing them to focus on value-add initiatives

Your Optimal Decision

Over the next few months this is going to be an exciting space to watch as decision optimization moves forward with three key design principals of simplicity, collaboration and convergence of the technologies helping more organizations complete their cognitive journey. So what do YOU do next? What’s YOUR next best decision?  Sign up for Optimization as-a-service as part of the Data Science Experience here.

Dinesh Nirmal,  

Vice President, Analytics Development

Follow me on Twitter @DineshNirmalIBM

One thought on “Should’ve, Could’ve, Would’ve – Making the Optimal Decision.

  1. Dhiraj Kumar November 28, 2016 / 7:15 pm

    Hi , its wonderful blog and Optimal Decision through Analysis to bring business at right place. We develop/use multiple tools/applications for Analytic in days. I would love to understand, The IBM Bluemix and recent blog , how comparable in high efficiency for business impact.

    Liked by 1 person

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s