 Welcome to MIT's supply chain frontiers from the MIT Center for Transportation and Logistics. Each episode features center researchers and staff who welcome experts from the field for in-depth conversations about business, education, and beyond. Today, David Carell reflects on what we heard in February speaking with truck drivers on our Voices of the Open Road episode. He compares these driver stories with findings from three of the MIT supply chain management master's thesis projects relating to the complexity of America's freight trucking networks. Take it away, Dave. Early in the year, as part of the MIT Freight Lab Driver Initiative, we talked with three working American truck drivers about their experiences in trucking. We asked them what shippers and receivers could do to help or hurt their utilization and ultimately their take-home pay. Because when we consider the driver shortage problem in the U.S., people often think that it's something that drivers are doing or not doing that could fix the shortages within over-the-road transportation or perhaps help rein in the currently high costs. But our conversations and our research is indicating that while it is within the sphere of concern of drivers to increase their utilization and decrease their dwell time, it is perhaps only within the sphere of influence of the shippers and the carriers they service to actually make the changes that could make these improvements. When we spoke with over-the-road truck driver Mark Kavanaugh back in March, he shared a simple version of how shippers can help improve drivers' lives. Yeah, when you show up on time for your appointment or even early and they get you right to a door, within two hours they've given you a green light and you've been given your paperwork and you're back out again. That's perfect. That's my ideal customer service. Instead of looking at what drivers themselves control inside the cab, we begin to look at how the institution's drivers work with, control their experiences of dwell and utilization from outside of the cab. A good experience for me is when you arrive and they're ready for you and they have a door for you and they have a bathroom for you to use and a clean driver lounge where you can have a seat or microwave something and get some snacks out of the machine if you need to. Pre-e with respect, a lot of them don't. Desiree Wood, known to many people as Trucker Desiree. And if you're out of hours having a small area that you could park the truck and do your 10-hour break, you go to some place like that, you're like, oh, I'll come here again. We're also joined by Paul Merhofer, known to many of his adored fans as Long Hall Paul. I'll tell you about the greatest food warehouse I've ever been to. It was 14 years ago. It was Cisco Foods in Alachua, Florida. And whoever designed and implemented that warehouse was to food, warehouses and trucking. What Temple Grandin is to cows and slaughterhouses. Got there, you checked in with the guard. He said, at three o'clock in the morning, I'm going to call you and we'll assign you with the door. Boom, three o'clock in the morning, here comes your call. They put you in a door and by the time you could even use the restroom, you're empty. The person who designed that should be canonized. There should be votive candles of that individual for sale and every truck stop. So truckers like me could light a votive candle and quietly meditate by it. One thing that a researcher can do is to take a look from an outsider's perspective. We are in no way really experts at the industries that we study because we don't do the actual work. We don't really play the game. Instead, we are more like commentators, people who watch the game from a distance, looking for fresh or interesting perspectives and insights. And in my world, that typically means data, lots and lots of data. This is a story in two parts. One, what we've learned about what trucking companies can do to improve driver's utilization and retention. And two, what shippers and receivers can do to reduce waiting time at their facilities. I'd like to introduce you to the members of the graduate student teams who worked with me this past year to follow up on what the drivers told us this spring. First, please meet my friends, Liora and Michelle, who studied unplanned downtime and recently graduated from the MIT SCM masters in supply chain management. So hi, my name is Liora Sotter. I graduated from West Point in 2015 and spent five years in the active duty army as a logistics officer. Finished out my military service last summer and joined my class here at MIT. Very thankful to be graduating here shortly. Hey, everybody, I'm Michelle Roy. I graduated with my undergrad in supply chain management from Texas A&M in 2016 and then spent several years doing procurement in the oil and gas industry before I joined the SCM class of 2021. How would you explain what you studied for your capstone project? So I guess to give a bit of background, a daily shift for a driver is 11 hours, driving hours within a 14-hour total workday, but our sponsor company found that their drivers were averaging about 6.5 hours driving time per day, which is pretty consistent across the industry. And that's due to various sources of driver dwell. And we define dwell in this context as unplanned, unproductive downtime while drivers are on the clock. So we were looking to really understand where unplanned downtime comes from in a driver's schedule and day-to-day activities. We looked at four main hypotheses to understand where this could be coming from. So whether that had to do with the geography of the node they were visiting, what time of day they arrived to their appointment, whether it was drop and hook or live load. So kind of that element of how many hands are in the pie, how many hands are in the pie of when they make a drop-off or pickup. And then we also looked at driver demographics. So does age play a role? Does months of experience play a role? And does gender play a role? And then finally, we considered do repeated visits to the same node by a driver. Does that have an impact on how much dwell they experience? Now that we understand what motivated your study, how did you research those questions? How did you conduct your study? So I'll start kind of broadly talking about why we focused on what we did. And then maybe Michelle can talk about kind of our quantitative and qualitative methods. But before that, just talking about measuring utilization. So utilization is a ratio of actual use to maximum capacity. So for truck drivers, this is hours driven over 11. And that's why we really focused on hours of service as our metric here. Talking to different players in the trucking space, we learned that some companies use things like dollars per week or miles per week. But these address productivity more than they do utilization. We were fortunate to be given a really big data set from our sponsor company that gave us a lot of opportunity, both TMS information, so transportation management system information, and ELD data. So that comes from the electronic logging devices that drivers use that give us information about where they are, what their status is, and other details like that. We had to define dwell within those data sets. And so for the TMS data, we defined it based on the duration of their visit to a node. So what we did there is we took, is this a drop in hook or live load? If it's drop in hook, the expectation is that it takes 45 minutes. If it's live load or unload, the expectation is that it takes two hours. And so that's what we compared to to calculate our dwell within the data set. So we had the arrival time of the driver and then the time that they left the facility. And so was that beyond the expectation depending on whether it was live load or drop in hook. So that's how we calculated dwell for that data set. And then for the ELD data set where we focused more on the drivers themselves and the demographics, we used their time in on duty status, which basically means that they're using their hours of service, but they're not actually driving. So it's not really productive time for them or for their company. With that in mind, we did some regression testing against our, with the context of our hypotheses to understand, are these factors significant contributors to the dwell that we're seeing? So could we say, for example, that as somebody's number of visits to the same node increases, that their dwell time proportionally decreases. And that's the kind of relationships that we were looking for when we did this hypothesis testing. The other side of the testing that we did or the research that we did here was more qualitative in nature. And we had the opportunity at our sponsor company to talk to really a broad array of people that helped paint a picture for us of what does it look like to be a truck driver today and what really makes up that landscape. So we were able to talk to some people more in senior leadership. And then also we were able to talk to a few drivers and hear from them, what's it like to actually be on the road? So the really cool thing about the timing is that we had some of those higher level conversations first leading into our quantitative research and testing. And then after we had some initial conclusions from the quantitative testing and the regression analysis, we went back to the drivers and we're really able, through asking some broad questions about their experiences, we're really excited to see that there was a match between what we were finding in the data and what their experiences were just really anecdotally describing their day to day and what happens if you're late. Excellent. What did you find? I would say we have three big takeaways. The first one is about driver demographics that we found that there really is not a perfect driver demographic when it comes to dwell that they experience. And so the dwell that a driver experiences is not a reflection of their age or their months of experience or their gender. And that leads me to parts two and three of what I think are the most interesting things we found where that dwell really is more in the control of the people that support drivers in their jobs. So dispatchers and people working at the nodes, as far as dispatchers go, what they can really do is send drivers two nodes during peak hours of activity. And this may seem counterintuitive, but what the numbers show is that when you are a driver and you show up during those peak hours, which we defined as between 7am and 5pm, you're getting in and out of a node a lot faster than if you show up during off peak times. And then the final thing that we found is that drivers who make repeated visits to nodes do see a decrease in their overall time spent at the node. In one of our hypotheses, we did find that repeated visits to the same node decreases dwell at that node. We use the TMS data and we counted the distinct number of orders between a driver and a node and used that count as a proxy for the number of visits. What was really interesting is that in over 90% of the data, at least from our company, those driver location pairs were either one or two. So in simpler terms, that means that the vast majority of the time or the 90% of the time, at least in the six month period that we were exploring, these drivers at our company are only visiting the same location once or twice. So we really think that that's a big area where the company could focus and that there's a big opportunity there too. Because I mean, it makes sense for probably any of us to say, if you move to a new town and then go to a new grocery store, it's probably going to take you longer in that first trip to a new grocery store to understand what the layout is. And then for subsequent visits, your time in and out is going to get a lot faster because you know where things are. And so really that same principle can be applied to our truck drivers and repeated visits to nodes that just in a very human way, they understand, where do I check in? Who do I talk to? And that really increases or decreases the dwell that they experience at those nodes and the total time that they're at those nodes. So for me, those are our biggest, most interesting findings. So so many people right now are thinking about driver dwell. What do you think people might not yet understand about the causes or the solutions to the driver dwell problem? When Michelle and I were first assigned this project, I fully expected that driver dwell would come from things that drivers were or weren't doing or would come from some characteristics of the drivers themselves. But through the course of our project, as Michelle mentioned, we really learned that it's really the people around the drivers, both their own dispatchers and then, you know, perhaps staff at the shipper and receiver facilities that have the most impact on drivers dwell. And I think for trucking companies, you know, maybe they don't have a lot of control over those shipper and receiver facility staff, but they certainly do have control over their own dispatchers. So sharing our findings about sending drivers to nodes during peak node activity times and sending drivers to the same nodes over and over again, those are totally things that are in the company's control and things that they can share with their dispatchers. I agree. The name driver dwell is almost a misnomer because I think it leads to the idea that it's something that drivers are doing or not doing that's causing this unplanned downtime or underutilization. But that really is the core of what we found is that drivers themselves don't have the power to change their own utilization. You know, we can look at drivers collectively really based on our study as a group of people that want to be out on the road driving and making money and being productive. And so it really is incumbent upon the companies that they work for and the nodes that they visit to help them in that endeavor, because it's not really something that they are equipped to do by themselves just because of the structure within which they're working through no fault of their own. This really goes back to that same idea of driver dwell is not the fault of a driver and it has nothing to do with their age or gender or months of experience. And I love that idea because I think we've talked about that there are some people that may feel interested in being a truck driver and that companies should take that interest seriously across the board and really pursue whoever would like to be a truck driver, especially obviously through the context of dwell that just because somebody is maybe older or has more experience or any of those factors doesn't mean that they're going to perform better in this space. And so I think to me at least to a degree this level is the playing field a little bit. So really kind of equal opportunity here and equal opportunity for the companies and the nodes to better serve the drivers that visit them. So a takeaway here for me was what Michelle said about how much is outside of the driver's control when we look to reduce dwell time. Okay then who else has influence? What if we look even one step closer into a trucking company's organizational chart? What if we talk about the dispatchers? Please meet my friends Danny and Paolo who happened to study just that in their paper Goldilocks and the three dispatchers. Hello my name is Danielle and score proctor. Hello I'm Paolo. In the simplest terms what did you study in your capstone project? What we looked at is freight truck drivers in the U.S. which we all know and love from highways and from getting all of our things and we looked at what has the greatest impact on their performance measured in a couple of different ways and really where we focused in was the what we felt was the understudied impact of carrier dispatchers. So the people that aren't on the road with the truckers but are actually back home with the carriers assigning loads to the truckers telling them where to go how to get there and what's next. We were looking for somehow to improve truck driver utilization in the United States. We wanted to provide some insightful recommendations from for management. Now maybe with a little bit more detail how did you do it? We know what you wanted to study how did you approach this question? What we had access to was six months of data about truckers. We partnered with a mid-sized Midwestern freight carrier and they gave us information from the electronic logging data. So this is information captured by the computers on board of the trucks themselves so it captures what the what the truckers were doing at every hour of the day while they were working. Their transportation management data so that captured sort of what the freight loads were and who the dispatchers were that assigned them and then some employment and tenure data on the truck drivers. And we used that to do some data analytics and some machine learning to find what were the features that most impacted truck driver performance and what of those came from the carrier organization or from the dispatchers themselves. Excellent thank you. Paulo same question and maybe if it's more comfortable to focus on some of the analytic tools you applied. Sure we have investigated both traditional method and a more sophisticated method we use linear regression to investigate some features that impact utilization and efficiency but we also have used some unsupervised machine learning techniques to cluster dispatchers. What we really wanted is to find if there are some common patterns between dispatchers that can deliver more expected results for the company and we have clustered them to identify this common features among those. Excellent thank you and what did you find? I would say that we have three key takeaways. First is that drivers working fewer Mondays they show higher efficiency according to this data set teams with less evenly distributed plans show higher productivity and this is related to what the dispatcher can do and third contrary to our first assumptions drivers on larger teams they tend to be able to achieve higher efficiency and our utilization. So I think the first idea of weekdays mattering so we found that when drivers drove fewer Mondays they actually had a higher mile per day efficiency and you know it's sort of long been anecdotally understood in the industry that Mondays are challenging days for drivers because they're leaving home challenging days for dispatchers to assign loads to those drivers but that Mondays are crucial to carrier productivity and what we actually found is the opposite and so this is not to say that carriers should avoid assigning loads on Mondays but I think you don't have to we don't need to think that a truckers week is the same as an office workers week they're different and that perhaps greater flexibility in those schedules or greater understanding in the differences in those priorities of drivers can help carriers better assign their loads when we talk about the the the unequality of distribution of freight plans which is quite the mouthful but what we what I think that really means is that it's not necessary to treat your drivers all exactly the same in order to have a cohesive and efficient team you can assign loads and miles and trips based on driver interest based on driver availability based on some of these different driver profiles and still see really successful driving teams overall and perhaps one of the most interesting pieces is this idea that drivers actually do better on larger teams this has two two interesting pieces one is that we found that this performance of these larger teams are highly correlated to greater disparity in this distribution of freight plans so dispatchers with larger teams actually have more pieces to move around the board to do some of that that work that I was just saying to assign plans based on driver needs but also it means the carriers don't need to focus on sort of lowering that driver to dispatcher ratio they can instead focus on maybe hiring fewer dispatchers capable of making more sophisticated decisions and making sure that those dispatchers are really empowered to make the right decision for their driving teams and not worry so much about just getting more of them excellent thank you so many people right now are thinking about the underutilization of American truck drivers and I know that you all have thought about that too what did you learn about driver utilization or underutilization by looking at this data that they could inform the the national conversation about truck driver utilization definitely the most surprising but one of the biggest takeaways for us was that there was this trade-off in the metrics that we looked at and the the big trade-off was between what I'm sort of summarizing as productivity so our utilization and then miles driven efficiency and retention and so really we when we cluster the dispatchers we saw these three distinct classes and the ones that had the highest retention had the lowest productivity in their drivers and vice versa and I think that for a long time there's sort of been this focus on retention because everyone's a little bit freaking out about this idea of a driver shortage and I think it's possible that we might be thinking about the driver shortage in maybe not the wrong way but there are other ways to think about it so when we talk about the driver shortage we talk about 60 000 driver short today we talk about 160 000 driver short by 2028 that's terrifying that's alarming we're not going anywhere we're not going to need fewer drivers we're going to need more what do we do and I think what Dave what your work has done when you look at our utilization is that perhaps when we talk about the driver shortage we shouldn't be talking about physical drivers short we should be talking about available driver hours short and in that case maybe retention isn't the most important metric anymore or perhaps there's there's sort of a different way of thinking about it and when we looked at the trade-off between retention and productivity sort of when I when I think about it I think what does this mean for carriers and it could be that retention isn't sort of like the one key the one sort of silver bullet there's other metrics that we need to focus on and we can focus on things like our utilization and improving that accepting that that might bring our retention levels down a little bit but knowing that we're getting better utilization from our drivers while we have them and that that might be sort of an alternative solution to this you know impending driver shortage issue yeah I would just say that our study indeed verified that through our data set that they are underutilized they drive around 6.5 hours per day when they could drive 11 hours and well we we have found a trade-off between retention and our utilization so maybe dispatchers or companies may be willing to worry less about retention and to focus more about how to improve their truck drivers utilization as well. Excellent thank you both I came up with a hard question so if you don't have an answer say we don't have an answer but if you have an answer I'd be curious so say we get some people who are managing truck infirms listening and they hear you and they push back and they say I don't think so I think the people that work for me want to make money and I don't think there's a trade-off I think utilization and retention go together drivers that drive more stay with me because they make more money to me that seems counter to what you all found and how would you convince that listener that the trade-off you identified is accurate. I'm going to convince them using the most convincing thing that I can say which is that you're totally right you're right you're absolutely right yes so we found these different classes of dispatchers and we found this trade-off but we also found that there was a dispatcher class that found the balance there was high retention and relatively high productivity it wasn't the highest retention and it wasn't the highest productivity but it did sort of find that maybe perfect balancing point between the two I would say truckers are like anyone else they're willing to work harder to make more money but at some point you hit a limit at some point the trade-off isn't there anymore and especially when you're talking about a group of people who are as a rule older and have families there is going to be a point where it's not worth that one more load it's not worth that one more trip even though it's worth more money they want to go home they want to take a break they want to see their families and so yes if you want to keep truckers happy you have to pay the money everybody wants to make money but you can also you can overshoot it yeah I agree with Daniel there should be a limit for everyone right you want money you want to work but yeah everyone has life right in other objectives maybe other things to solve and to live for as well I'll also say that maybe we don't need to think a driver as a standard driver we have different drivers different ages so maybe some drivers that are more younger they are more willing to spend more time out of their homes but those who already have established family or thing or another context maybe they are not willing to spend so much time out of out of their homes so we should realize that there are different drivers as well imagine now that it's the end of a work day and you've been to work and now you have to go to the grocery store and now you're pushing the cart that one cart with the one infuriating wobbly wheel back to your car in the parking lot you just want to get home as paul put it you have other things to solve for and other things to live for truckers are just like anyone else we can push them too hard take them too far away from the other things that they have to solve for and to live for I think we do that far too often because we don't manage our dispatching to meet their needs if we've looked at what trucking companies can do how can warehouse staff get better maybe we can't all achieve that sainted status the long haul paul brought up at the top of the hour but maybe we can all change our practices even just a little bit in order to do incrementally better please meet my friends zoe and roger hi my name is zoe hi my name is roger marino in the simplest terms what did you study in your master's capstone project we started automation of warehouse decision making we tried to eliminate the human decision factor and the decisions for the warehouse so our study focused on the inbound side of the warehouse uh it's from the data we received it's where we saw the biggest opportunity for improvement so here it's where the trucks arrived to the warehouse and then they're either assigned to a doctor or they have to wait uh at the drop lot which is which is a parking space and they wait there until there's like labor and space available and from what we saw uh the trucks were waiting there for a long period of time and our goal was to reduce this waiting time until they were unloaded in our model any the company's situation like when the trucks entering the set uh a checking time will be locked in the inbound coordinators need to prioritize the based on whether it's my new idea or whether this shipment content uh short if there is any dock available they will send these prioritized shipment first to the available dock uh otherwise the arrived trucks will form a waiting queue at the drop loss and once the trailer moves to the dock for unloading the warehouse coordinator will assign assign special counterbalance forklift and forklift the driver to unload the truck and then once the process finished the truck leaves the dock will be able to discredit even simulation model by arena um this model contains several decisions and we use several distributions to simulate the reality and generate these sarcastic variables we use this model to test different scenarios and the policy combinations to find out the best policy for for different scenarios so for our simulation model what we did we created a scenario that we checked with the company to confirm that our numbers were correct and once we had like our baseline model uh we created different scenarios and each scenario changed the truck arrival distribution so in our first scenario the trucks arrived normally on the second one we had a little spikes it was a lumpy the distribution and for the third scenario it was a quarterly push this means that at the end of every quarter there's increased demand so more trailers arrived to the facility and through these three scenarios we tested different unloading policies the first one was treating a certain type of trucks called manual ids we would unload them first at the end with the lowest priority the second policy would treat them with the highest priority meaning they would be unloaded first the third policy treated these trucks as any other truck and we proceeded to unload them with a 5-0 model which is first in first out and the final policy um designated a special dock and the special labor only for these certain types of trucks uh when we use the simulation model to simulate the different scenarios we use we call it a job load waiting time but it's actually the time intervals between the truck checking at the company's gate and the truck and loading start time like we use this time interval to index to marry our results so we test all this combination and then we just compare those time intervals to find out the best policy excellent thank you and to roger what did you learn what are the lessons from from the research for what we learned from this project we were able to determine that by keeping a constant unloading policy um the results were better than having the current approach which was having warehouse manager make this decisions and this is because their decisions were not consistent and also that by knowing ahead of time like if there's going to be a spike in demand you're able to adapt each policy to the one that fits better and once you're able to do this your results are going to be better instead of having just like a inconsistent but consistent approach from the person from the warehouse managers based on those takeaways what is your advice to a warehouse manager sure i think a warehouse manager should like knowing more collecting more data about the truck arriving behaviors then they can identify like what's the current situation or what about the next situation in the warehouse then they can choose the best policy to unloading the truck more efficiently is there anything when you were doing the project that surprised you for me i think i learned it's quite interesting that the warehouse is the whole things like it it operate they won't operate separately like say today i mean tomorrow on album or other days only warehouse operation so i think i learned when we do the simulation model we have to considering the warehouse decision as a whole we cannot just separate them and do only one part yeah so for me one of the most surprising things that i learned is that like the overall waiting time for trailers uh inside the warehouse which the average was like 21.5 hours so like in the u.s i know there's like a shortage of trailers and this not only happens in our sponsor company but throughout the industry and i believe like the truck industry is able to reduce this waiting period it will become a lot more efficient than what it currently is and they might not necessarily need more drivers they just need to become more efficient and just shorting these waiting times what zoe and rogers simulation suggests to us is that inbound scheduling should vary based on the demand that he's experienced and expected on any given day and at least in fast-moving retail some demand variations are predictable notably the quarterly push of inventory as suppliers try to get their product move off their accounting books when this is the case dedicating doctors to labor intensive freight unloads works to reduce time spent waiting however in other cases including business as usual and periodic storms that same policy is the worst we are committed to continuing to explore this fruitful area of research why try to get this right well because the truck driver resource is scarce and dwindling and because in all systems time is money maybe that's all you really need to know but you probably already thought of that let me offer one more let's start with mark what would you want a shipper receiver listening to this to know about the working conditions of the drivers carrying their loads they they all around need to change their attitude towards the drivers we're not we're not just you know hauling their freight we're also providing for the country you know we are essential to this country and they need to understand that and treat us with a lot more respect than what we're getting at a time when as a nation we are hopefully starting to come out of a pandemic that if you were listening you unlike many survived then perhaps we should now reflect on the many efforts that enabled our survival in addition to the never repayable efforts of frontline health care professionals over the last two years the rest of us like you and me the work from home podcast listeners the plain and ordinary citizens of the home body economy we all made it through in relative comfort in part because of what truck drivers did for us what they sacrificed to keep our supply chains running the links that they traveled to make the home body economy possible it behooves us as researchers I think to now bring our skills and our focused attention to the problems the truck drivers face maybe we can't solve everybody's problems probably not but efforts like this one are at the least I hope a respectable acknowledgement of a very large debt if you like this please stay in touch with me and the driver initiative to see what next year's crop of students and I come up with all right everyone thank you for listening I hope you've enjoyed this edition of MIT Supply Chain Frontiers my name is Arthur Grau communications officer for the center and I invite you to visit us anytime at ctl.mit.edu or search for MIT Supply Chain Frontiers on your favorite listening platform until next time