In our optimisation Insight series, we’ve uncovered how optimisation can be key to managing your business challenges, how to stay competitive in the On-Demand Economy and how optimisation can unlock unprecedented value for your business.

So, how do you leverage the power of optimisation to achieve your business objectives?

Solid foundations: four key components for success

Optimisation is a powerful way to turn data into solutions to difficult business challenges, like making the best decisions amid complexity, improving the efficiency of critical processes or balancing competing demands.

To unlock the potential of optimisation to meet your objectives, your business needs strength in four key areas: mathematical modelling, data, algorithms and user-focussed deployment. Below we discuss each of these capabilities, using the analogy of a car to help tie them together.

optimisation deployment
How can you solve your business problems with optimisation?

Mathematical modelling

Put simply, mathematical modelling – or ‘modelling’ for short – is the process of describing real-world situations using the language of mathematics.

Through this translation, complex business problems can be analysed and solved. The best mathematical models are targeted: they encode the vital processes at work but cut away unnecessary, distracting detail. For example, suppose you wanted to improve the efficiency of production within a factory. A good model relating schedules to productivity would include the rule that part B needs part A, say, but perhaps wouldn’t include highly detailed measurements of how long it takes a factory worker to walk from one workstation to another. Experienced mathematical modellers are able to draw upon a wide range of techniques, in order to find the best mathematical representation of a business problem and produce reliable results.

In our analogy, the mathematical model is the car that can take you to new places. But without fuel, a good engine and the means to operate it, you are just left stuck in the garage…

Data

To apply your mathematical model to your business challenges, you need data.

For example, a delivery company working on a large scale may have a model to describe the different ways to deliver stock to its customers, but it will only be useful for identifying the most efficient routes if they also have data on the travel times or distances along the road network. Data offers the greatest value when it provides the information that the mathematical model requires to function. If a mathematical model is the car to take you to new places, then data is the fuel you put into it – and you get best results when you use the right fuel for the car.

Algorithms

If the model is the car and data is the fuel, then what is the engine that uses the fuel to get the car moving? The answer is algorithms.

Algorithms are the methods used to push the data through the mathematical model and get the answers you need. Not all algorithms are equal, however. Poor algorithms can lag, stutter or even stall completely. Good algorithms, on the other hand, are tailor-made to get you the right answer at the right time without demanding the earth in return (such as unreasonably long time scales or excessive computing power). The combination of a good model, quality data and great algorithms provides that X-factor that gives you the edge over the competition.

Deployment

What good is an answer if it can’t take you anywhere?

The fourth piece of the puzzle is to deploy the answers from the model, data and algorithms in a way that delivers insight to decision makers. It is analogous to the car’s accelerator, which enables a person to drive themselves to new places. Often this is best done through deploying a software tool. Bespoke solutions have the advantage that they can be tailored to the end user’s needs, whether that be producing a glossy, ‘boss-ready’ dashboard or a command-line interface for use by technicians who want the nitty-gritty details.

Iteration: the spark that brings the foundations to life

Getting the know-how and data in place for mathematical modelling, algorithms and deployment plants the seeds for success. But how do you best capitalise on this potential to deliver successful business solutions?

The answer is to view this process not as a line from start to finish, but as a cycle that you go around again and again. There is a natural flow through the four components we described above:

  • The mathematical modelling leads to a specification of the data that should be sought.
  • The model and data inform the algorithms that are used to generate answers.
  • The whole approach is then deployed to bring answers to the end user in a useful way.

The leap is to realise that you don’t have to proceed once through this process, in an attempt to achieve all your goals in one go. Instead, the whole modelling-data-algorithms-deployment development cycle can be performed multiple times, with each new iteration delivering something that is closer to what you ultimately require. The end of each iteration provides a valuable opportunity to gather feedback from end users on the latest version of your product or analysis, as well as a chance to consolidate lessons learned from the development process. The iterative approach thus offers a basis for frequent correction, a bit like the way a sat-nav constantly recalculates the best route to your destination.

There are two fundamental ways this cycle is beneficial to a project. Firstly, it helps to clarify the requirements of a project and align them with what is achievable. Innovative projects often start with requirements that are hazy or only partially understood, even if the high-level goals are clear – this is only to be expected when breaking new ground. By quickly bringing prototypes through the full development cycle, customers and teams have something tangible to experiment with and critique. This provides a means to ascertain quick success and sort out the ‘must-have’ requirements from the ‘could-haves’.

Secondly, problems which could derail a start-to-finish approach will be found and fixed sooner. To return to our example of improving efficiencies within a factory, an early iteration might involve a scheduling and resourcing model for only a single production line. Having brought this through to a prototype software tool, it may be discovered that solving the problem already takes several hours, so that including many production lines using the same methods would be impractically slow. Discovering this early means a new approach (such as improved algorithms or a simplified model) can be developed while there is the time and flexibility to do so. Early experimentation illuminates the approach that will best tackle the challenges you face and, consequently, reduces the risk of a project that overruns or fails to deliver.

Conclusion

Successful solutions to critical business challenges are founded upon expertise in mathematical modelling and algorithm design, together with quality, relevant data and a deployment that has the decision maker’s needs at its heart. Frequent iteration through these four components is key to clarifying requirements, providing early insight and maximising the chance of delivering results with lasting impact.