LinkedIn·Friday, 17 July 2026·17 Jul 2026

"How do you validate the quality of your scheduling algorithms?" It's a common question, in industry and academia. By benchmarking them, of…

Geoffrey De Smet

CTO Timefold - APIs for scheduling and routing optimization, powered by our open source solver.

"How do you validate the quality of your scheduling algorithms?" It's a common question, in industry and academia. By benchmarking them, of course. But for us, that's changed dramatically over the years. Originally, we benchmarked a new algorithm (or an algorithm change) on our laptop on a number of datasets. If it's statistically better, release it. Like academic papers, basically. But as our solver become used across the globe, that practice became a liability: while a change might improve in one use case, it could regress in another. We actively avoid regressions. So we started benchmarking across use cases. Still on our laptops... which is impractical, unreliable and irrelevant. Impractical, because we couldn't use our laptop while benchmarking. Unreliable, because background processes impacted CPU cycles. Irrelevant, because production environments don't run on M-series Macbooks. These days, we benchmark our solver algorithms in the cloud. And it's a game changer. Far more realistic. Far less hassle. Even with real-world maps data for vehicle routing variants. Every algorithm idea that crosses our mind, is now validated quickly and thoroughly, across datasets and use cases. That unlocks the next level in innovation. The impact is already tangible in Timefold Solver 2, with an increase in scheduling quality and solving speed. This is just the beginning... What if we open this up for your custom scheduling models too?