In supply chain in particular, optimization models are seen as a black-box. The perception is that the results cannot be understood and explained.

This is not true.

This is exactly the point of mathematical models. Everything can be explained.

It is the nature of mathematics to be precise. You might not like the results but it is precise.

When I validate a model, I play with the parameters and the inputs. Because it is mathematics I know exactly what the output should be.

If the result is not as expected it means my equations are not correct. Or there is an issue in the data.

These are the only 2️⃣ reasons a model gives strange results: data and the formulation. Both of them we can investigate and fix.

Sometimes it is a bit of detective work and it can take some time. There is always a logical explanation, which is most of the times related to the business context, hence the users are able to follow it.

Moreover

Mathematical Solution + Planner Expertise = Business Solution

How can a planner create the business solution if they do not understand how solvers take decisions?

With the right training, users can learn this detective work. They can and should learn how to tweak the data, the parameters, even the constraints and objective function.

At the end of the day, the planners Excel formulae are the ones configured in the software. If they understand it in Excel they can also follow it in the software.

There is a learning curve and this is normal. All that is missing is patience to explain using a friendly and non-technical language.

This is one of the reasons for the hands-on optimization masterclass and recently optimization4all.com. Open the black-box, show how solver decisions can be 100% traced back and tweaked.

Users should be offered training about how to explain the results, take control of the tool and use it in the best way for their industry.

🎙️ Thoughts on this?