Good morning, good afternoon, or good evening, wherever you may be viewing this presentation. My name is Jennifer Van Mullekom, and I'm the Director of the Virginia Tech Statistical Applications and Innovations Group. I'm going to talk to you today about Missing Marks Delay the Dig. This is a case study on how we use JMP Visualizations in a Root Cause Analysis.

You need to get some background on the problem before we can jump into how meaningful the visualizations are in solving this problem. I'm going to give you a little bit of background on underground utility location notification process. You may not think about this very often because you only notice it when it's not there. For example, if somebody is digging a hole in your neighborhood or near your house, and they hit a water line, you may notice that you're without water for a while and be frustrated, or you may also, God forbid, be without your internet for a little while.

Now, those are really innocuous consequences for hitting underground utility lines, but certainly we can have much greater consequences. If we hit an electric line, a gas line, we might be looking at property damage, injury, and in extreme cases, loss of life. This is a very serious topic, and it's very important. In fact, there's even legislation around this in the Commonwealth of Virginia.

Right now, we're going to zoom in and get some more background on the utility location notification process. What happens when we want to focus on locating underground utilities for excavation, such as a fence installation, digging a building foundation, or even installing landscaping in our yard, we call in, or we type in a web ticket for the utility location request. That goes to Virginia 811 in the Commonwealth of Virginia. Virginia 811 then notifies the utilities, the electric company, water, sewer, gas, telecom that might be running through that excavation area. They have maps of the cities, towns, municipalities, and counties throughout Virginia, and they know who to notify.

These utility companies are then responsible for locating their own lines, and so they're either going to respond back by saying, "We know that there aren't any lines in that excavation area," or they're going to send someone out to mark those lines with spray paint. Once all the notifications have been returned to VA811 and indication has occurred that the utilities have been marked, then excavation can begin. You can dig the foundation, you can install the fence.

But there's a problem with this process currently. Let's look at that problem. The problem really is that "No Shows" are increasing. Now, how do we define a no show? A no show is really one or more of those utilities per ticket that is entered in the VA811 system not showing up to mark their properties or not responding that they don't need to mark the property. You can see from the JMP visualization that's on the slide in front of you, that from 2019 to 2023, we've seen these tickets increase from 4% to above 10%, "No Shows". This creates delayed excavations, due to one or more "No Shows".

We want to know why have these "No Shows" increased? That's one of the main questions that we will want to answer. We need to do this for a mixed audience. When I say mixed audience, it's really mixed knowledge level. We have a steering team that's trying to solve this problem, and this steering team ranges from excavators that probably have very little experience with statistics all the way up to CEOs who may have taken several classes in statistics in their MBA, and folks in between that might be working in data analytics, QA, QC, or even be operators in the VA811 system, managers in the utility companies. Quite a variety of experience and understanding of statistics and data.

We wanted to do something simple that could help them get to consensus to solve this problem. So that we're no longer delaying excavations, which cost time and money. One of the first things that we looked at as a root cause was, is this really related to an increase in ticket volume overall?

You'll see that with our JMP plot, we have years broken out by different colors and different symbols. If we look at the tickets per week on our x-axis, we notice that we don't have much difference in the range of tickets per week. But what we do see, if we look at the "No Shows" per week, we certainly have quite a variety in the number of "No Shows" per week for the same number of tickets.

What does this tell us? Well, it tells us it's probably not an overall increase in ticket volume that's going to lead us to the solution to this problem. It's not making sure that we just staff up necessarily at this moment and that we're able to handle a large increase in ticket volume overall because we're not actually seeing that. Something else is going on in the process, and we want to know what that is.

Of course, given that there are folks from different industries, different stakeholders that are in this room on the steering committee, people have a lot of ideas and a lot of opinions. But we needed to come with data, and one of those opinions regarding what might be going on in the system is related to fiber installs.

You may remember during the pandemic, we really were able to expose broadband inequities within the US. There are a lot of people from rural communities that were having to drive their children to school parking lots in order to get access to Wi-Fi, in order to download their assignments, or be online with their teachers for help sessions, for courses. This created a large amount of money being freed up in terms of federal funds and state funds in order to install fiber. Virginia has been are taking several efforts to really bring fiber optic cable to every nook and cranny within Virginia, and there are definitely some rural locations.

In investigating fiber, what we see is that we have our years across our top panels, and we have our fiber install across our rows, and we definitely see an increase in those delay "No Shows" as we go across the years. However, what we want to notice is that as we go from not a fiber install, no fiber install to a fiber install, we see a pretty dramatic increase in these delayed tickets due to "No Shows" over the past five years. In fact, we're really at 2X the delays relative to a non-fiber install for fiber installs in 2023 when this project was completed.

That led us to another question because folks are realizing that fiber install trends vary over a year. We also know there's a seasonality to all types of construction that requires excavation within Virginia. Typically, we're working on more construction, more tickets, for underground utility locations in, say, March through October, maybe the first part of November. We have trends across the years as fiber installs increase, and we have seasonal trends. You can see that in this next JMP plot where we started investigating some additional sources of variability.

Also, some of the theories that the folks on this project had were focused on the geographic trends that they might see. As you might guess, when a fiber company gets a contract to install fiber, they're going to do that regionally for a while, then they're going to move to another region. In particular, what this plot shows is you can see an increase in the intensity of "No Shows" across the years in general. We see a big uptick, in particularly 2022 and 2023, in these "No Shows" as we see some of the darker colors on the map of the state of Virginia.

In addition, we see that those trends match up with some of their knowledge of the fiber installs that we're seeing some rural counties, for example, we can pick these out because of what we know about Virginia. Some of these counties are very rural along the Blue Ridge Parkway. You would also know that some of these South Central counties are also fairly rural, and even in Southwest Virginia, very rural as well, compared to maybe the Northern Virginia region up here where we're around DC, and we would expect a lot of tickets.

We see that intensity increasing, and folks could look at this map from 2023 and see the intensity moving around as some of the fiber construction moved around. It actually verified what they were thinking about. Now, of course, from a reasoning standpoint, we would say, well, maybe 683 "No Shows" isn't that many given that we have hundreds of thousands of tickets that are being issued each year.

It may not seem that many in that context, but when you consider the fact that maybe there's one utility locator for, say, a Tri-County area because they can usually handle that workload, 683 "No Shows" represents an impossible amount of locations to go and check out. This was very illuminating fascinating to the folks in the room with respect to what's actually happening within their fiber "No shows".

We also looked at a lot of other cases of root causes as we went through this PowerPoint presentation, it was probably about 50 PowerPoint slides of graphs, so there were a lot more things investigated, but these were some of the illuminating graphs that really helped us understand what was going on with no shows. Our conclusions from this analysis come in to play in the sense that overall ticket volume is not a contributing factor to "No Shows".

We do see an association between the fiber installs and the delayed response tickets due to one or more "No Shows", and "No Shows" vary over the geography and the time. The increases in specific counties track with known increases in the fiber installation due to the broadband. There's still a lot more to learn. We haven't really gotten to the root cause, and I would never go so far as to say that this is causal inference or anything like that, but it created a discussion, and it helped build consensus among a team that was very divided around what the solution should be.

There's still a lot more to learn. It prompted questions. Is there a locator shortage in certain areas? Do we need traveling locators that go around with the fiber installs? Are newer locators taking longer to actually mark to work the properties? Then we also got to a point where we realized that the electric company, the water company, sewer gas, and telecom aren't talking to each other. They don't have an overall prioritization process, so they might be working on separate sites leading to a number of incomplete sites being marked, and excavation is unable to begin on any of those sites, whereas if they coordinated efforts, they might be able to create at least a portion of those sites that are completely marked with their resources.

I think this really shows the value of simple plots. We could certainly do a lot of more complex analysis with this data, but the communication, the visualization, it created conversation, it created a group of individuals that were previously divided, and it created more focus with them in terms of problem-solving. I'd like to say thanks to Virginia 811's CEO, Scott Crawford, and the Manager of Data Analytics and QC, Kenny Spade, for their continued support of the Virginia Tech Statistical Applications and Innovations Group. We're proud to partner with them in helping them problem-solve with data-based reasoning.