The document discusses technical lessons learned from selling optimization solutions to business users. It provides tips for understanding the needs of different stakeholders in a business, including ensuring the optimization project solves the right problem and delivers return on investment. It also offers ways to improve optimization performance, such as moving to the latest solver release, changing the mathematical model, using multithreading, and tuning the solver.
Technical Lessons Learned While Selling Optimization To Business Users
1. Technical Lessons Learned While
Selling Optimization To Business Users
Jean-François Puget
Industrial Solutions Analytics and Optimization
IBM, France
SponsorSession IBM1
Monday July 1 – 14:30-16:00
Optimization does not start with data. Very often we start working on the optimization model without any data. The lack of data isn’t an issue per. It becomes an issue during the tuning phase, but not during the elaboration phase of a modeL Here are two examples with the Empty Container Repositioning (ECR) asset The optimization model is quite stable, what differs is how data is collected by the customer. Forecast of where empty containers will be needed in particular vary a lot from customers to customers. Last customer we visited do not have any relevant data. It is once they understood the potential ROI enable by the optimization model that they started to think about how they could collect the required data. That data does not exist prior the ECR discussion. The previous customer, once interested, asked us to validate the ROI duing a POC. The POC lasted 8 weeks, and all the work in the POC was about getting the right data in the ECR asset. Once this was done we proved a 7.5% redution in transportation cost.