Kris Van Marcke Senior Business Consultant +32 474 98 42 83

Optimization Model Development and Validation

A mathematical trick that reflects the best possible optimization solution

 Do you have a correct value function, measureable and with the right KPI’s, as basis for your optimization? You know what you will optimize thanks to the “value discovery” approach. Solving your optimization puzzle follows from this. We prepare a model for it. A mathematical model with predictive value, which reflects the process that you want to optimize and is sufficiently accurate. This accuracy is assessed by reference to an actual problem, a representative dataset. The model is, in a way, validated. Ordina can assist you in preparing different mathematical optimization models or can support you in validating models that you create.


Optimization algorithms work on a model that approximates reality. The more strongly the model reflects reality, the more credible the optimization results. The model has sufficiently accurate predictive values: if the model predicts a particular effect, it must be possible to corroborate it in reality by an objective observation.

On the other hand, driving the accuracy of the model to the top can be negative. Indeed, the optimization puzzle could become so complex that it can no longer be solved within an acceptable computing time.


Most optimization techniques work with mathematical models calibrated on reality. There are other techniques too. Ordina develops the right model for you. We guarantee that the model is logical and correct and will overcome all the hindrances inherent to the chosen optimization techniques. Our experts validate the accuracy of the model by subjecting it to a specific problem. Using a realistic dataset, we assess the model and the results presented.

Ordina conducts these validation checks both on models of its own development and on optimizers created by the customer.


Afterwards, you have a validated model, which assures you of:

- An accurate, yet not excessively accurate, model of the problem

- An objective, measureable target function

- A model that respects the limitations imposed by the chosen optimization technique

- A model tested using a relevant dataset

Optimization Model Development and Validation Contact Contact Kris Van Marcke