The Death of Static Logistics: How Quantum Optimization Keeps Fleets Moving

[00:00:00]So, with that, let's get into the web webinar. Next slide, please, Mike.

[00:00:11]So, Jim and I were recently talking, and Jim, I'm sure you'll remember this conversation.

[00:00:18]I was I was telling you about a recent logistics comfort conference that I was at, and I had the opportunity to talk to the head of transportation for this mid-size fleet operation, and I asked her to really, you know, walk me through the day-to-day, day in the life of what of what life actually looks like, right?

[00:00:41]And what she described to me really stuck with me ever since. Um and and the reason it stuck with me is because it wasn't unusual for her at all. It was just another Tuesday, right? She described the process of planning and executing logistics optimization and routing optimization as it really starts the day before, the night before. And her planning team kicks off a routine around 4:00 in the afternoon to plan their fleet fleet deployment and the different routes and stops that they have to make. So, it's numerous stops, dozens of trucks. They have to deal with delivery windows, driver availability and hours, as well as vehicle capacity, and all of this has to fit together. Um but what she told me that was interesting is the longer that they let their solution run, the better the plan gets. But the problem is the trucks roll out the next day whether or not the math or the solution is is finished. So, every night her team makes the same call. The same call every night. They stop the solution early and take whatever whatever they have.

[00:01:57]Or, in many cases, they don't know that the system has been set to stop at a certain time uh after a certain amount of time, so that the processing is complete in time for them to execute their plans. Again, not the best plan, uh but it is a good enough plan. And and she told me flatly that she knows and the team knows that they're leaving miles on the table every single day.

[00:02:24]Every single day they're leaving miles on the table.

[00:02:27]So, then the next day reality happens.

[00:02:30]Truck breaks down on the interstate and you know interstate, there's multiple stops still on the route. Um second truck, you know, hits black ice and is 100 hundreds of miles away.

[00:02:44]And her team then has the the her team has to rebuild or re-optimize those routes, um those same routes that they planned the night before, and continue to make trade-offs, right? She needs to plan fast. It takes, you know, 10 minutes, 11 minutes to load reload the data, another 20 minutes to run the application, and they just don't have that time. So, her planners do what they always do. They start moving stops around by hand. Uh once again, trading that precision for speed and hoping nobody gets stranded 200 miles away from home.

[00:03:21]So, later in that morning, um you know, they've already missed numerous delivery windows. They have angry customers.

[00:03:29]Uh they have 45 minutes of overtime uh from a driver standpoint that they hadn't budgeted for. And and and and she told me that this happens most days.

[00:03:40]It's not a crisis. It's a routine.

[00:03:43]And what struck me about that was how she framed the real problem. It wasn't the breakdown. It wasn't the weather. It was just disruptions are part of the job.

[00:03:54]And in in speed and quality were never both on the table. Uh for her anyone doing the job. And and the conversation is the reason we built this session today, right? Because it what she describes isn't only one company's bad luck. It's pretty much the standard that we experience across the industry. And it raises the simple question, what if?

[00:04:19]What if her team never had to make that trade-off at all? What if the plan was already the best plan? And what if the breakdown uh came in as as they got it, they were able to build a new high-quality route um in the time it takes to answer a phone call, let's say.

[00:04:38]And that's what dynamic real-time optimization really changes. And today, we're going to show you exactly that live, a real scenario running on a quantum computer.

[00:04:50]So, Jim, let's get into the demonstration. Let me turn it over to you, and let's show everybody this scenario, this situation, um and how we solve for it live using quantum optimization.

[00:05:04]>> All right. Can you see my screen, okay?

[00:05:07]>> Yes.

[00:05:08]>> So, I you know, and and Jason, this is a conversation you and I have uh many many times, and I I would even say that the scenario you just went through was a was a relatively advanced scenario, right? A a company that is doing um you know, they they've got a route optimization solution. They're they're doing something in that space. Um and they are doing some amount of recalculation during the day when they do have some challenge, you know, kind of against all odds. I would tell you more commonly what what we run into is, you know, if if folks whatever they have for planning, whatever they have for optimization, there's generally a cycle, and it's generally 24 hours ahead of time, you know, ahead of time, you know, and in a in a distribution world, you know, you hit a stock out time, things kind of go dark, the planning cycle begins, it runs as long as it runs, it's usually measured in hours. Um and to your point, when when they're done, now the plan's set, and then, you know, the first truck rolls out at 6:00 a.m. and the plan's busted by by 6:30.

[00:06:02]Um so, I you know, we we find that to be much more the case, and then and then people in general are living in this spreadsheet triage world um for the rest of the day. What what's the best that we can do with the information that we have and the resources that we have.

[00:06:15]Um and and to your point, you know, what what if it could be different? You know, the the industry seems to have generally just accepted that this thing works this way, and and what if it could be different? So, that that's the demonstration and the the the tip uh for this. And what I'm going to show you um is a a demonstration that's set up to simulate like uh a food distributor redistributor. Uh it's set in a in St. Louis, Missouri, metropolitan area. So, it's going to be a 200-mi circle around St. Louis.

[00:06:45]There's going to be 80 trucks, 80 drivers, uh 620 delivery points. And just because big numbers are kind of fun, and that's kind of the the challenge that we're dealing with, if if you look at the total problem space, like the total different ways you could assign those deliveries to those trucks, um it's 10 followed by, you know, 1,179 zeros, which which is a whole lot of zeros. Um that's a whole lot of calculation, and that's really, you know, just showing you that the breadth of the problem that we're going to solve.

[00:07:14]Let me flip over here.

[00:07:17]Let me set this.

[00:07:21]Okay.

[00:07:22]So, we're going to zoom in here to St. Louis. We've got our north, south, and central distribution centers.

[00:07:28]Um we've got 80 trucks. It's 5:00 in the morning, and we're going to start rolling.

[00:07:34]Um and you'll start to see them leaving, sorry, the distribution centers and kind of making their way out. And this is sped up, it's not in real time, because it's much more interesting to be able to actually see them moving as we're doing these things. And uh you know, this is going to be real life and uh real life is messy. And so what we have is this kind of disruption control panel where I can come in here and introduce challenges uh into their day. And what I'll show you is uh us introducing the challenge and then in real time um we're going to recalculate and reoptimize for the fleet under the this new set of challenges. And what's what's happening and you'll see these the pop-ups that come up and I'll talk through them as we're going to make live calls uh to the QP to the quantum computer. Uh the one that we're running off of right now is hosted in uh is hosted in Burnaby, British Columbia uh at at at D-Wave's location there. Um and our goal is to get back to optimal as quickly as possible. So we started the day optimal and now we're going to have traffic incident.

[00:08:33]Uh so we got five delayed, we're reassigning 20 stops across 2,530 variables and 140 constraints and we're actively computing right now.

[00:08:46]All stops are assigned and we're back on schedule, right? So that that took us a total elapsed time of you know, 1.1 seconds on the wall to recalculate and reconverge and get the planners the information they needed for what needs to go where now given that constraint.

[00:09:01]Um so now we can also introduce something like um you know, a dock closure.

[00:09:13]2,310 variables, we're actively computing right now. We're hitting the back end.

[00:09:22]And we're back.

[00:09:23]>> and Jim, this is you know, this is real life scenarios, right? As as we talked about and this is the type of information or the specific information organizations get on a daily basis, right? As the fleet is out there deployed, that reality has happened, the dock closures, the traffic accidents.

[00:09:42]So, in today's world, again, tying going back to the the story I talked about, um a lot of organizations have to do this manually, right? Cuz they're Otherwise, they have to reload data. Otherwise, they have to uh reprocess the data, build out the plans, and then re-optimize the plans, and they just don't have the time to do that. Where what you're showing is the the quantum application taking all that data in and then re-optimizing in that real-time example.

[00:10:14]>> That's right. And in in general, most most of these scenarios, if not all of these scenarios that I'm showing, generally result in that the plan that you calculated the night before is now broken. All right. Uh like I I just took a driver out of service because uh they were over on their hours of service. So, they're now not a bit They're not not available for routes. Well, what do I do with that?

[00:10:35]How do I plan that? And that normally means people are immediately back in spreadsheet mode. Uh you know, your local on-the-ground people are scrambling now to to manually re-optimize uh whatever they have to do.

[00:10:47]Uh versus if you have a uh you know, a a D-Wave quantum um system, that the opportunity is to, as you said, Jason, let's When we become suboptimal, let's recalculate and become optimal because we can do it so quickly. Um I can show you like we can take a vehicle out. So, this is literally going to remove uh you know, significant capacity in delivery.

[00:11:11]So, vehicle 11's now out of service. Uh we've got to We've got to reassign 48 uncovered stops.

[00:11:17]Um we're going to localize it to the local five, and we're going to start solving. And what you'll see here is we've solved two, and this was a This is a bunch of work. There's 48 to go. So, what's going to happen is it's going to continue to resolve and re-converge as it moves all the way through that stack and completely reassigns all the rest of the routes. Well, we just cut another We've got 39 left. And this will continue to iterate here until we've completely re- reassigned and re-converged and we're re-optimized again. So, the opportunity and the goal is to stay optimal in the face of reality with which happens all day, every day in in a metro, in a region, certainly in in a national or international fleet. And with a D-Wave's quantum annealing technology, your opportunity is to take into consideration all of these variables and very rapidly recalculate. Things that take hours can take minutes and that makes a huge business difference.

[00:12:13]So, just think about the business opportunity you have in a world where, you know, a truck, some trucks go offline, you can recalculate and replan your loads in minutes. What would that do for your business? Right? You know, if you have the opportunity to repack and recalculate because you've got driver no shows or you've got dock refusals and you're trying to manage to maybe not time you're doing time windows and or time deadlines. I mean, all everyone's business is different, but imagine the ability to recalculate and be re-optimized again in in minutes.

[00:12:53]That that's the real business opportunity that that we have now.

[00:12:58]>> Show you.

[00:12:59]>> Yeah, yeah, it's exciting, right? That we're able to take all those goals, those constraints and variables and let the system re-optimize that process and, you know, you and I you and I have talked to customers that the way they do that today is, you know, they may have sticky notes on the drivers' schedule and the availability and the different routes that they're going and they're literally looking at sticky notes to come up with a new plan.

[00:13:27]Clearly suboptimal, right? But come up with the with the new plan. And And I also think important to point out while we're showing real-life scenarios, right? The data that we're using in here is that synthetic data, but based upon real-life scenarios. So, we're not using any customer data per se, but they are real-life scenarios and masses amounts of data that that we're using as part of the demonstration.

[00:13:54]>> Yeah, it you know real-life real operational life is messy. And I think a lot of times um the the technology struggles to everything looks great on paper and it looks great in a kind of static environment, but the real world's messy. And that And that's really what I'm so excited about is the opportunity to me particularly in logistics for the first time like the opportunity to really react and live in that world and to stay optimal and to get out of the spreadsheet chaos, right? It just teams spend so much time and energy and money, frankly, working to get those solves excellent. You know, the things that they're doing today, whatever passes for their optimization strategy.

[00:14:36]And then it's literally kind of all for naught by you know, the the first opening hour of the day because real-life happens. So, this is a great example to me of technology being able to meet meet reality and navigate it successfully, right?

[00:14:52]Time and time again. So, um you know, again, we're we're in the first hour and 45 minutes of the simulated day here. We've had half a dozen different challenges and across the board we've continued to re-optimize, reallocate, and get us back on track making sure that we're getting where we need to get to on time in the most optimal way that we can, saving miles, saving hours, and increasing customer sat all along the way.

[00:15:21]>> Yeah, Jim, there there's a question in the chat. Maybe now's a good time for us to answer it. It says, "Can you explain what the quantum computer is doing differently to do this quickly versus feeding this into data into an AI model, right, which can kind of do the same same math?" What's your perspective on that?

[00:15:41]>> Yeah. Um I think that the difference being uh the quantum the quantum annealer the quantum computer um is like a fabulous uh purpose-built calculator uh that is unbelievably good at these problems, right? These these routing and constraining problems. So, given the given the input um it is incredibly fast uh because it gets to break the laws of conventional physics to move data faster than uh you know, it's possible on a conventional computer.

[00:16:15]Um and it's going to give you a um uh a specific known outcome.

[00:16:21]Um the the the AI models uh that uh I get you're probably talking about like a like a large language model. Um those are uh uh you're going to get a a non-deterministic outcome uh and uh counting on them to execute uh the math uh accurately uh in in real time is uh is not generally a a great operational uh uh uh a plan. So, what what what we find them to be really great at is providing supporting data and input to a model like this uh so that we definitely see it as complementary uh but not replacing the the calculation capability of a of a quantum annealer, if that makes sense.

[00:17:04]>> Yeah, yeah, and right, and I agree with that. And and fundamentally, right, what the quantum computer is is doing is using quantum effects, right? So, you know, you hear about things like entanglement and and tunneling um and and whatnot, right? So, we're using those quantum effects for speed and quality to the uh of the solution of of the answer. So, that's the technical difference between a quantum computer and uh traditional computers. And also, the problems like this aren't linear, which is what uh classical compute tends to be strong at.

[00:17:44]Um they are exponential. So, when using the quantum computer to optimize and using the quantum effects, that's where we see the speed and efficiency really of the outcome or the result of the quantum optimization.

[00:18:01]Another Another question that we had in there uh Jim is how how many vehi- vehicles were we routing in that scenario?

[00:18:09]>> [clears throat] >> And maybe we could just kind of go back through the variables of DCs, uh drivers, vehicles.

[00:18:17]>> In that example, it's um 80 drivers, three dis- three DCs in the metro, and uh 80 trucks, 620 delivery points.

[00:18:29]>> And I I believe as you showed in there on the front screen, you know, it's it's it's a very complex, big problem to solve for.

[00:18:38]>> Yeah, let me I can share this again, too.

[00:18:40]>> I think it was 10 to the power of something.

[00:18:43]>> Yeah. So, if you look at the like the total problem space 10 to the 1,180th, right?

[00:18:49]Um it's there it's a significantly large problem. The total These are the total stats on trucks, drivers, active routes, and delivery points, total fleet capacity, uh the number of distribution centers, right? So, the the goal is to to model something that's um recognizable, uh you know, in last last mile and businesses take all kinds of different forms, right? Like your your specific logistics set up and what you do and last mile, first mile, middle, uh white goods delivery, return logistics, you know, what do you do? And that's part of the part of the benefit of working with a D-Wave solution is us being able to take the customize and tailor it to your business because that's another thing that we commonly see with folks. There's there's lots of different There's lots of different things that people do from an optimization standpoint and you know, being able to tailor it to your business and its specific business rules will will get you the best outcome and that that's that's that's part of the process that we go through with customers is to help understand, you know, what what's unique about their business and making sure that we model those constraints into the math.

[00:19:54]>> Yeah, because at the end of the day, that's what it's all about, right? Is is delivering new different outcomes, but most importantly value, transformational value, right?

[00:20:02]The These aren't just AI initiatives or technology initiatives. It really is about a business initiative, right?

[00:20:10]Mike, if you don't mind, can you share your screen again and show the value proposition that that I'm going to tell another story about. Really, it's an extension of the story that I started off with that Jim showed as part of the demonstration.

[00:20:30]And here it is. And and and so you know, the conversation that I had with this logistics executive really did stay stay with me.

[00:20:41]So, afterwards, I spent some time looking into numbers and the industry as a whole and and the scale really stopped me in my tracks, right?

[00:20:51]So, the combination of trucks that the US drives is close to 200 billion miles per year.

[00:21:00]This is just in the US and this is just looking at trucks and trailer combination, right? And that's not per company, this is per year industry-wide.

[00:21:10]And And today the average cost to run a truck is about $2.34 for every mile driven.

[00:21:18]Um and that accounts for fuel, driver wages, maintenance, insurance, and all those other components. So, when you do the math, you're looking at something in the order of $450 billion a year just to keep the US trucks truck industry moving.

[00:21:34]Massive number. Now Now hold that number next to uh what that logistics executive told me.

[00:21:42]Her team was leaving miles on the table every single night because um because the plan that they dispatched wasn't the best possible plan.

[00:21:52]Um just one that they had the time to solve for, right? And compute. And that that that wasn't unusual. This was happening on a daily basis. And we think that it's the norm across the industry, right? We're making these tradeoffs all the time.

[00:22:09]So, let's ask the obvious question. What if a fleet could shave even 5% off the miles they drive? Not by cutting service, not by cutting routes, right?

[00:22:19]But by simply planning and replanning or optimizing and reoptimizing better. That 5% of 200 billion miles is about 10 billion miles, right? That never needed to be driven. And at today's cost per mile, that's roughly $20 billion a year the industry is spending on miles that that we didn't even have to drive, right? Even just the fuel portion uh at around 65 cents a mile is well over $6 billion of fuel alone, right? So, that's the macro number, but it's really about it's really uh not about that macro number.

[00:22:58]It's about when the math what that math looks like inside one fleet, one dispatch office, one planner's morning morning. The fewer wasted miles, the lower the fuel spend, the less unplanned overtime, fewer missed windows, right? The lighter the the the footprint, all from the same trucks, all from the same drivers, right? The same freight, just plan and adopt it better.

[00:23:25]And the business case underneath is really what we talked about, right? It's it's not it's not just about theory, right? It's about the efficiency gains, a real measurable share of that $450 billion that's out there in cost savings. Um and all of that can be captured by whoever plans smarter and adopts adapts faster, I should say.

[00:23:51]>> You know, another one that I think maybe [clears throat] folks don't think as much about, but I'm also really excited about and we're having a number of conversations with with customers and prospective customers on reduced time windows for service. So, thinking about like service delivery organizations that have got to set time windows for delivery or service provisioning for for customers.

[00:24:13]So, the opportunity to service more or the same customer load in the same time period, but increasing customer sat retention and in a world where folks are competing for you know, for for customers all the time, really being able to differentiate on even a higher level of service because you can give a tighter service window.

[00:24:31]All those types of things are possible here when you can plan better as well.

[00:24:37]>> Yeah, excellent.

[00:24:39]Um Mike, I know we are coming up on time.

[00:24:43]Were there any other questions that you were getting that you want to throw out to Jim and I?

[00:24:52]>> Yeah, and but first of all, thank you Jim and Jason for the really interesting conversation.

[00:25:00]Uh we do have some questions from the audience. Uh we'll try to get to as many of them as we can in the next couple of minutes. Um So, Jim, I'm going to ask this question to you.

[00:25:13]Um how often do you encounter client problems that require the processing power only a quantum computer can provide beyond what a traditional computer could handle?

[00:25:23]>> Yeah. Um I think I would I would I would reframe it a little bit and in general, we find that we can provide a better or more complete view to the business question that's being asked in that scenario uh because we can typically address um more variables in the same time period, right? So, you know, very often we'll have a conversation um where somebody's doing something currently and we're just able to have a slightly different conversation because we're able to consider some additional um you know, definitely um related business processes that they they can't or don't currently um consider in their formulation because it wouldn't be uh feasible to do so. So, that's a a lot of words to say uh fairly often. Uh that that that it's not an uncommon scenario for us.

[00:26:18]>> Okay, the next question um This could also be for you, Jim. Um what kind of disruptions is best suited for uh quantum computing or quantum optimization? Does it help with something simple like a truck breakdown or is it built for bigger stuff like weather closing down a >> Yeah, I I I think one of the um one of the things that's so exciting me about working in the space and and with D-Wave and with customers is is the the size of the problems that you can go after, right? So, um what is it best suited for? It's best suited for solving complex problems. It can definitely handle the complexity of a regional interruption, a local interruption, a single equipment interruption. So, I find that it it it it scales well, I guess, to to answer that question.

[00:27:09]>> Awesome. Well, I think that's the last question we have time for.

[00:27:14]I know we didn't get to answer all of the questions that the audience asked.

[00:27:18]So, we will reach out to you individually with a response. We promise you that. And Jim, Jason, again, thank you so much for for sharing your knowledge with this group today. And to the audience, thank you for spending your time, you know, listening in.

[00:27:35]Please don't forget to complete the post-event survey. It's going to help us improve our content.

[00:27:41]And yeah, so and you will get the recording to this session about 24 hours after we end the event. So, thank you everyone and have a great rest of your day or evening.

[00:27:52]>> Thanks, everyone. Take care. Thanks, Jim.

[00:27:54]>> Thanks. Take care.