The science of making better decisions

Management science, or operations research, is a specialised discipline for business decision-making.  It involves solving problems that have complex structural, operational and investment dimensions involving the allocation and scheduling of resources.

Some problems are “one-off” while others need an on-going solution.  Where an on-going solution is required we are able to deliver a framework for solving the problem in the future and, if necessary, a computerised solution integrated into an organisation’s existing information infrastructure.  Some of the possible solution techniques include:

• Mathematical programming – model complex economic, technical and operational systems as a set of variables related by mathematical rules and find optimal “solutions” (e.g. resource allocations) using Linear or Non-Linear Programming;
  
• Network modelling and routing – reduce flow costs and travel times; improve efficiency and productivity;
  
• Simulation – construct a system model and then trace the outcomes over time to identify significant events and patterns;
  
• Stochastic Optimisation and Decision Trees – calculate an optimal strategy that is hedged against uncertainty;
  
• Real Options Theory – calculate the value of flexibility under uncertainty;
  
• Game Theory – construct a model (e.g. Cournot) of the interactions between individuals or organisations and evaluate their respective incentives;
  
• Heuristic or approximation methods – to solve otherwise impossible problems and give an answer that is good enough, but not necessarily perfect;
  
• Decision Support Systems – identify critical parameters and evaluate for management focus and decision making; and
  
• Statistics and Econometrics – analyse data to identify key components, determine “best fit” mathematical relationships and produce forecasts.

Orbit Systems has extensive experience in all facets of Operations Research to deliver the right solutions for your business needs.

In many situations, the same type of decision is made repeatedly as part of a planning cycle. Planners potentially spend a lot of time coming up with the proposed plan, and even a small improvement in each result will make model development worthwhile.

Average behaviour is the key with repeated decisions. The optimal outcome for each decision is not necessary as long as we routinely make at least ‘good’ decisions.  An example of this is eating.  No one gets healthy by eating the right food for one day.  Happily, people won’t get unhealthy by eating junk food for one day, or several days in a row; as long as they make good choices most of the time.

Having a model to achieve this can save planning time as well as money.  If it takes a planner until 4:30pm to calculate the daily plan, a company may have been operating poorly most of the day even though they had information to enable them to do better at 9am.  And what happens to the quality of those solutions when the planner, who knows what they’re doing, is away?

If even a small incremental improvement in the quality of the multiple decisions justifies the investment, then you are likely to need us.

A very big decision must be made.  Whatever it is, it’s going to cost a lot of money.  So it’s worth the effort to spend some time figuring out if the best possible decision is being made. Even confirmation that the correct course is being taken might be worth the development of a model.

To know whether a model is needed, start by deciding the “size of the prize”.  Ask yourself “if we made the best possible decision, how much would that gain?” and conversely, “if a poor decision were made, how much would that cost?”

The best solution may be apparent, even if we’re not sure how it’s achieved.  For example, everyone knows the correct answer to the Rubik’s Cube, even if they can’t solve it themselves.  So knowing where you are and how far from a good solution you might be is useful information.  Sometimes it’s a question of asking whether an answer that is 1% better than the current one justifies further study.  If it does, you likely need us – and we generally save a lot more than 1%.

How do you know this is a problem for Orbit Systems?

The chances are, you have one of these two types of problem:

1. A very big decision must be made. Whatever it is, it’s going to cost a lot of money and spending a certain amount of time figuring out if the best possible decision is being made is likely to be worth the effort. Even confirmation that the correct course is being taken might be worth the development of a model.
2. The same type of decision is made repeatedly. In this case, people potentially spend a lot of time coming up with the answer, and even a small improvement in each result will make model development worthwhile.

Some simple questions will signal if an Operations Research-based solution is likely required:

• How difficult is this problem? Is getting any feasible solution going to be difficult?
• If the problem is solved without a model of some kind (e.g. by hand), how far will the solution be from the best possible solution?
• If the answer to the previous question is ‘I’m not sure’, what method would be used to gain a better idea about what the best possible solution is?

If you’re not happy about the answers to these questions, it might be time to talk to us.

When problems are difficult to solve, any feasible solution can seem acceptable.  But dig a little deeper and it often becomes clear that problem solvers have no way of knowing how close to the best possible answer they’re coming.  And very often the person charged with solving problems is ill-equipped to improve on the quality of the solutions.

This can be because:

• The problem really is a difficult one to solve.
• There’s no time to work on a better technique because the current solution process takes all available time.
• The tools and techniques for creating a better solution are not available, or not well understood by the problem solver.