Latest / The Operator's Manual with Fexingo: Daily Decisions, Tactical Wins, and Business Execution / How a Single Algorithm Saves UPS Millions Each Year
Transcript
- Lucas: You know what number keeps me up at night? Twenty million. Luna: Twenty million what? Dollars? Miles? Lucas: Minutes. Every single day, UPS drivers collectively waste twenty million minutes idling at traffic lights. And that's before we even get to the left turns. Luna: Left turns. I've heard about this — UPS famously tries to avoid them. But is that actually real, or is it just a fun fact people repeat? Lucas: It's real. And it's not a quirky preference — it's the visible tip of a massive operational system called ORION. That stands for 'On-Road Integrated Optimization and Navigation.' And it might be the most ambitious route-optimization algorithm ever deployed outside the military. Luna: So ORION is basically a giant GPS that tells drivers which way to go? Lucas: That's what it looks like from the outside, but that's underselling it. ORION isn't just finding the shortest path. It's solving what's called the 'vehicle routing problem' — a notoriously hard computational problem — for fifty-five thousand routes every single night. Each route can involve up to two hundred stops, with variables like package volume, time windows, traffic patterns, and, yes, the number of left turns. Luna: And left turns are bad because… you sit waiting for oncoming traffic, burning fuel, blocking the flow. Lucas: Exactly. A left turn at a busy intersection can take two to three times longer than a right turn. Multiply that by hundreds of stops per driver, per day, across a fleet of nearly a hundred thousand vehicles, and the numbers get enormous. UPS calculated that eliminating just one left turn per driver per day saves about ten million gallons of fuel annually. Luna: That's absurd. One turn per driver. Lucas: The scale is hard to wrap your head around. But here's the thing — ORION didn't start as a left-turn avoidance system. It started as a pure math problem. UPS spends roughly three billion dollars a year on transportation costs, so even tiny improvements compound. The project kicked off in the early 2000s, and the first pilot was in 2008. Full deployment took until 2016. Luna: Eight years to roll out a routing algorithm? That's an eternity in tech. Lucas: It is. But the challenge wasn't computational — it was behavioral. The drivers. UPS drivers have a famously detailed set of methods — they call them '340 Methods' — that dictate everything from how to hold the key to how to exit the truck. These are people who've been doing the job their way for twenty years. And then along comes an algorithm telling them to do it differently. Luna: I can imagine the reaction. 'I know my route better than some computer.' Lucas: Exactly. Initial compliance was terrible. Drivers would follow ORION for a few stops, then revert to their own routes. They'd claim the algorithm was sending them into unsafe areas or making them miss time commitments. Some of that was legitimate — early versions had data issues, like missing one-way streets. Luna: So UPS had to win over the drivers. How? Lucas: They did two things. First, they made ORION adaptive — it learned from driver preferences and started incorporating local knowledge. Second, they changed the incentive structure. Drivers were measured on something called 'plan adherence' — how closely they followed the recommended route. And they started sharing real-time fuel-economy data per driver, so the savings became visible. Luna: So the behavioral change was as important as the algorithm itself. Lucas: That's the real lesson. The math was ready in 2008. The human system took until 2016. And by the time it was fully deployed, UPS was saving about four hundred million dollars a year in fuel, maintenance, and labor costs. Luna: And those left turns — did they actually drop? Lucas: Significantly. ORION routes now contain about twenty percent fewer left turns than the routes drivers would naturally choose. But here's the counterintuitive part — the algorithm doesn't eliminate all left turns. Sometimes a left turn is optimal, especially if it avoids a much longer detour. ORION just asks the question: is this left turn worth the cost? Luna: So it's not dogma. It's math. Lucas: Right. And that's why I think ORION is such a great case study in operations. It's not about a single brilliant insight — it's about the unglamorous work of implementation. The algorithm itself is based on academic research from the 1960s. The breakthrough was UPS's willingness to spend years and billions of dollars to make it work at scale. Luna: And the willingness to let drivers push back. That's rare in a top-down organization. Lucas: It's rare, and it's essential. If UPS had just mandated ORION without listening, the whole thing would have failed. Instead, they turned resistance into input. The system got better because drivers said, 'Hey, that street is one-way,' or 'That road is closed for construction.' Luna: That iterative loop — algorithm plus human feedback — that's the sweet spot for a lot of operational improvements, isn't it? Lucas: Absolutely. And it's not just about routing. There's a bigger principle here: the best operational systems aren't the ones that replace human judgment. They're the ones that augment it. Luna: You know, speaking of systems that augment judgment — if these conversations have nudged you to think differently about how you run your own operations, we'd be grateful if you considered supporting the show. It's listener-supported, no ads, no sponsors. You can find us at buy me a coffee dot com slash fexingo. We keep it simple. Lucas: Yeah, and it genuinely makes a difference. Helps us keep digging into stories like this. So thanks for that. Luna: Anyway — back to ORION. One detail I love is that UPS actually patented the algorithm. But they don't license it. They treat it as a competitive advantage. Lucas: That's right. And you can see why. If FedEx had ORION, the competitive landscape would shift. UPS is essentially saying, 'We have a proprietary operational edge that's very hard to copy.' Because even if someone else wrote the same algorithm, they'd still have to go through the same decade-long cultural change. Luna: So the algorithm is the moat, but the implementation is the real barrier. Lucas: Exactly. And that's a lesson for any business thinking about operational AI. The code is the easy part. Changing how people work is the hard part. Luna: Let's talk about the numbers again. Four hundred million dollars a year in savings. What does that break down to? Lucas: Roughly one hundred million fewer miles driven per year, which means about ten million gallons of fuel saved, which translates to about a hundred thousand metric tons of CO2 avoided. So it's also an environmental win. Luna: And that's just the direct savings. What about softer benefits? Lucas: Driver retention improved. Because routes are more consistent — drivers know roughly what their day will look like, rather than having to adapt to ad-hoc dispatches. And customer satisfaction went up, because delivery windows became more precise. Luna: So ORION touches everything: costs, environment, people, customers. Lucas: It's a classic example of a 'keystone' operational change — one that ripples through the entire system. But I want to be clear: ORION is not magic. There are still routes that it optimizes poorly, especially in dense urban areas with high variability. And the system requires constant updates. It's not a set-it-and-forget-it solution. Luna: Is UPS still improving it? Making the next version? Lucas: They are. They're now integrating real-time traffic data from third-party providers, and they're experimenting with machine learning to predict delivery windows even more accurately. The goal is to move from 'planned route in the morning' to 'dynamic rerouting throughout the day.' That's the holy grail. Luna: That would be huge. Imagine a driver already on the road getting rerouted because a package was added or a customer changed their time window. Lucas: Exactly. And that requires an even deeper integration between the algorithm and the driver's handheld device. UPS is working on it. But they're moving carefully — because every change has to be tested at scale before it goes live. Luna: So what's the takeaway for someone running a smaller operation — a local delivery company or even a restaurant with delivery drivers? Lucas: You don't need a billion-dollar algorithm. You need to track the right thing. For UPS, it was left turns. For your business, it might be something else — wait times, route deviations, fuel consumption per stop. Start with the data. Then experiment with small changes. And most importantly, involve the people who actually do the work. Luna: Bring the drivers into the room. Lucas: Exactly. Because the algorithm doesn't know that the bridge is out or that Mrs. Johnson likes her package left by the side door. That local knowledge is invaluable. And if you build a system that respects it, you get the best of both worlds. Luna: So ORION is really a story about respect. Respect for the math, respect for the drivers, respect for the scale of the problem. Lucas: I think that's exactly right. And it's why, after all these years, UPS still hasn't fully 'solved' routing. They've just gotten a lot better at asking the right questions.