The Operations Podcast with Fexingo

The episode packs substantial operational insights: slot-based auctions with dynamic pricing, segmented stacking with ML-driven dwell prediction, behavioral incentives via rebates/surcharges, and iterative constraint identification. Most claims are concrete and non-obvious (e.g., 40% of dwell time from four-hour truck arrival windows, 78% prediction accuracy, 35% reduction in reshuffling). Minor padding around ad read and closing pleasantries, but core content is dense.

“They introduced a system where trucking companies bid for time slots in thirty-minute windows. A slot during the 10 AM to 11 AM window might cost fifteen euros more than a slot at 2 PM.”

“They use machine learning to predict dwell time based on the vessel schedule, the consignee's historical behavior, even the commodity type.”

The application of Erlang C queuing theory to port operations is fresh, and the specific combo of dynamic slot auctions + ML-based segmented stacking + behavioral incentives is not a recycled framework. However, the underlying concepts (queuing theory, price signals, dwell fees, predictive modeling) are individually well-known. The originality lies in orchestration and context-specific tailoring rather than novel theory.

“The port's operations team explicitly used a variant of the Erlang C formula - that's the same math call centers use to predict how many agents they need.”

“They introduced something called 'segmented stacking' - basically grouping containers by expected dwell time.”

Lucas appears knowledgeable and speaks with specificity about port operations, but the episode does not clearly establish his title, affiliation, or hands-on operational role at Rotterdam or elsewhere. He demonstrates deep familiarity with the case but reads more as a well-informed analyst or consultant than a practicing port operator who led this work. No introduction of guest credentials.

“The Port of Rotterdam is Europe's largest container port. In 2023, the average container sat in the yard for 5.2 days before someone picked it up. By early 2026, that number was down to 3.1 days - a 40 percent cut in dwell time.”

“Each intervention had a measurable impact. And they used the data from the previous intervention to fine-tune the next one.”

Exceptionally strong on specifics: named port (Rotterdam), baseline (5.2 days), end-state (3.1 days), percentage (40%), slot pricing (15 euros differential), prediction accuracy (78%), reshuffling reduction (35%), dwell-fee impact (34% to 19% in four months), named carriers (Maersk, MSC, CMA CGM), timeframes (six-month ramp, eighteen-month rollout), and exact percentages of problem sources (40% trucks, 25% stacking, 15% fee-addressable). Few vague claims.

“In 2023, the average container sat in the yard for 5.2 days before someone picked it up. By early 2026, that number was down to 3.1 days - a 40 percent cut in dwell time.”

“The port published a paper showing that the model predicted dwell time within one day for about 78 percent of containers... It reduced reshuffling moves by about 35 percent.”

Luna asks follow-up questions that push into mechanics ('how did truckers react?', 'how accurate were predictions?', 'unintended consequences?') and surfaces real trade-offs (customs delays, replicability, power dynamics at LA). However, few moments of genuine disagreement or skeptical probing; the conversation largely affirms the case study rather than testing assumptions or challenging claims. Questions are solid but not incisive.

“So instead of policing behavior, they used price signals to smooth out the flow. That's interesting - how did the truckers react?”

“Did the fee cause problems for importers with customs delays?”