Evan: Hello and welcome to the Mining Your Own Business podcast. I am your host, Evan Wimpey, and today I am very excited to introduce our guest who is Kathryn Walter. She’s a senior operations research analyst at Avista—Avista being an energy company. And we’ve had guests from a lot of different industries here on the show so far, but this is our first guest in energy at all. So we’re going to learn a lot about what Kathryn does there, how analytics get applied. Excited to have you on the show, Kathryn. Thanks so much for coming.
Kathryn: Thanks for having me.
Evan: All right. To kick things off, can you give us a little bit about your background and how you ended up at Avista?
Kathryn: Sure. So my background going way back is I have a bachelor’s in OR from the US Coast Guard Academy. Did not know what or was when I decided to do it, but it was the math department and I like math. And then realized, oh, I really like this. I then did a search rescue coordination. Then the Coast Guard sent me full time, paid for my masters at Columbia, which is a great deal.
And then I did human resources analytics at Coast Guard headquarters. At that point, I was like, I really like this OR (operations research) stuff. This is great. I want to keep doing this. And in the Coast Guard, it’s not a career path. So it was like, well, guess I’ve got to leave. So I left, I became an optimization consultant at Ortec in their office in Houston where I’d helped with optimization problems for few years. I also did some professional development, teaching of courses. And then I was looking for something else and someone reached out on LinkedIn. A recruiter was like, Hey, you look qualified for this operations research job for a utilities company and Spokane, Washington.
Would you be interested? And I said, Nope, I’m not moving to Spokane. That’s not an option for me. And he was like, oh, well thanks for replying. A few weeks later they said, Hey, would, would you be interested in doing this? Being fully remote, they realized that between the pandemic and the nature of this job, like it would be fine to be fully remote.
I was like, okay, tell me more. All I know at this point is utilities company in Spokane, Washington. That’s it. And they tell me more and I’m like, oh, this, this sounds like what I’m, I’ve been looking for Sure. Tell, tell, and then I, I got hired in August of 2021, so it’s been almost two years. And Avista is the utilities company in Spokane, Washington.
Also, it’s pronounced Spokane, not Spokane. They had to teach me that. And Spokane, if you don’t know where it is, is near the Idaho border on the eastern side of Washington state. Because I, they also had to teach me this. So I had to learn a lot cause I, I never had reason to know before. And at the time I lived in Houston, I now live outside Washington, DC still fully remote.
Still have no plans with Spokane. It’s just not, not an option for me for personal reasons. But so in Avista—Avista is the power company. If you live in Spokane, your power bill is Avista. And that, that’s it. That they’ve been around since 1889. We have about 7,000 employees. And I’m the first and currently only, and therefore the best-ever operations research analyst at the company.
So about 10 or so years ago, around 2009, 2010, my boss, Clint Kalich, knew what optimization was, and he always says he’s an economist. And so he knew what it was, but he’s not an OR person, but he is like, oh, this could, this could be helpful. He runs the planning team within power supply at Avista.
And he started finding like a use case for it. And he did a very small optimization model for two the plants. So two hydro plants that are connected, meaning one the water goes through, one goes down and gets to eventually reaches the other plant and found that doing better decision making and better modeling of these systems could be $8 to $10 million a year in additional revenue with almost no extra cost other than the cost of making that model.
So it’s very profitable to make better decisions this way. And he was able to get management on board with, yeah, go create this application. So he used and still use, we definitely still use some, some excellence operation research consultants. He did the heavy lifting of the building the math model and so forth.
We also have a team of software developers who make the interface, deal with the database, get everything moving in the right ways. And that team has changed over time of getting bigger, getting smaller as the needs are. And now since about 2017 or so, there’s been like a fully functioning app. And like I mentioned about 2021, they realize, hey, let’s have an OR person in house.
So I’m kind of in the middle of this app of the, between the users and the developers and the OR consultants to, to get everything going. I realize right now I’m the only person who’s has both the access and the skillset to see what the user’s seeing—what’s on the screen, and when they try to run a model, how’s that turned into the data that gets sent to the code that builds the math model.
I can then see the math model and if it’s infeasible, I can see the why it’s infeasible, or I can see why I need to figure out what the why’s infeasible, and I can see a step back all the way back to what the user sees because either people can do that. They don’t have the skillset to be able to see the math model aspect or the OR consultants don’t have the access to the user interface.
And so I’m the person that kind of, in the middle, I can see the, the full thing and I, I work on this, it’s called the Avista Decision Support System, ADSS, and that what I do. I log aspects to it of, Hey, something got infeasible, what’s going on with it? Something is off, or “Hey, we want to add a new feature, we have a new requirement.”
We need to make sure that’s modeled correctly. You know, users are saying, Hey, these numbers look wrong. Why are they wrong? Okay, let make a look at why they’re wrong. So a lot of d different applications there within the same big picture application.
Evan: Yeah, very cool. Yeah, and very much tuned into the operations research world.
It, it sounds like an OR problem, which. One of the, I was very surprised when you said there’s no career path in OR in the Coast Guard because I suspect there are use cases where the Coast Guard could use some operations research.
Kathryn: Oh, absolutely. And they have them. They have some civilians who have long-term careers and there’s a couple people I know who are active-duty officers who are now having a second tour.
But since it’s not really seen as a career path, that’s not least, when I, back when I got on 2018, a whole five years ago, it was not seen as a career path. It was, I could maybe get lucky and go again one day, but I was like, why would I, why would I do that when I can leave and have a career in it and do this forever?
And also at other point, at seven years after duty time, I was like, I don’t want to wait 13 years to retire and or so, so booming and so much new stuff’s going on. It’s like, I don’t want to wait 13 years to. To be able to start doing this like and plus what they have a lot of or people don’t get to do the OR skills.
Mostly that has to do with some senior leadership, not understanding or skills. Sure. Like we had to run a joke in my office that because someone who outranked, like all of us, came by one time and he, he said something about, oh, you know, he wishes he could learn, you know, what we do. And, and he just doesn’t have time though to learn how to do VLOOKUPs and pivot tables.
Cause to him that’s what we did. Which is a set, cause we did do a lot of pivot tables and v lookups. Cause that’s the tools that we had that that’s what we had to do to ask the questions that we got. So it was that, that was the attitude that that’s what you do. And so a lot of like, or stuff you don’t get to do as much or it could be really hard to get going.
So it’s this, I could do this or I could just get out and then do, go have a career path and then get to do or every day. Yeah. And so that’s what I’ve been doing. But I mean, the Coast Guard set me up. Well they, you know, they set me up, they pay for both my master bachelor’s and master’s degrees in OR, and gave me some work experience so, I don’t regret it.
It’s very much the, they’re, they’re missing out cuz I’m not, I’m definitely not the only person who’s left the Coast Guard for this career path cuz in other bigger services, it’s there, but I’m not, I’m not the first, I, I doubt I’m the last you two.
Evan: Yeah. Makes perfect sense. Yeah. You have a master’s in co.
Masters in or from Columbia. You don’t want to spend your days doing v l cups and pivot tables in Excel. Absolutely. And so, so you went to Avista. You learned how to pronounce the city’s name. You learned. A little bit about Eastern Washington, even though you never had to move there. And in Western Idaho, I don’t know.
I guess you’ve never moved there, so I don’t, is it close at all to Moscow, Idaho.
Kathryn: Okay. I feel like I should know this. I do not. I know it’s, it’s near a Coeur d’Alene, which I’m not sure. Yeah. Cour d’Alene. Okay. Yes, it’s, it’s close. I hear it’s about 45 minutes from us. So even though I’ve been with the Vista since August, 2021, I went to Spokane the first time ever the first week of May this year, so a few weeks ago. It was very exciting. It’s very, people were very nice there, way nicer than I expected. Like I didn’t think people would be mean, but I was like, people are friendly. Everyone’s talking to me and having conversations with everyone.
Evan: You’re in the DC area, everybody’s outside of there is relatively nice. My little shot at the National Capital region. Yeah. But I guess I, I guess to, to get us closer back on track, you, you, you learned a little bit about the company, but presumably you had to learn about a little bit about hydroelectricity, how dams work, how things work. So can you talk a little bit about that growth and experience?
Kathryn: Yeah. Well, well, just like, I didn’t know how to, how to pronounce Spokane or where it was on a map. I also was not aware that people still use hydroelectric power. I didn’t know I, I’m from Georgia and then I was in the military. Like this isn’t something that comes up. So I just didn’t know that it’s very much still a thing.
And we also do more than just hydroelectric. We also have thermo plants. And we have some solar and similar things. We also, we, we model all of it, but I, yeah, I had learn all of it. But luckily my manager is just like, yeah, I know you don’t know this stuff. That’s okay. We’re not hiring you for that knowledge.
We’re hiring you for our knowledge. We can learn everything else. And so I’ve learned a lot by talking to people, by seeing it, by just reading things. And as I’ve need, I, I still don’t know a lot. Some stuff I, it’s more of a, I’ll learn it as I need it. And, but yeah, I, I didn’t realize this and I should have realized this, but as one as I never thought about was the power.
You know, you have to have everything balanced all the time and you can’t just have a whole bunch of extra power hanging out. And you also need to always be ready to move up the system up and down within a few minutes or almost instantaneously. So you have to account for that in your generation.
Everything else. I didn’t realize this, so, but we modeled that. So there’s all these things that I had to learn that just I feel like is one of the beauties of operations research is that it’s so industry neutral. You can apply it to anywhere and you just have to learn what you need to learn about it in order to apply it because the concepts are the same.
You know, the how to make a non-linear function linear enough that you can plug into a mixer linear program that’s the same where wherever, whatever industry you’re in, in.
Evan: Yeah, I, I think that’s a great, a very generalizable point too. We had a previous guest, I actually, the most recent guest we had, is in the pharmaceutical industry.
When they were growing their analytics team there, they were deliberately looking to hire people outside of the pharmaceutical industry so that they came in fresh mindset, being able to look, it sounds like you were sort of targeted being outside of, you know, the energy sector, the energy industry because of the mathematical skill sets you had there.
And you come in and you help build and maintain and, and apply this, this optimization work. This is, you know, going on 10 years-ish that this has been a vision. But how old did you say? Avista’s on the order of a hundred years old. Mm-hmm. So they’ve been solving this problem without operations research, maybe specifically for a long time.
So, Are, are folks eager to do things better? It sounded like you had a champion, your, your boss was, but mm-hmm. Is that, is that sort of across Avista?
Kathryn: Yes. I, I’d say so. People are eager to do things better and it comes down to, you know, the, the bottom line of if this works to make more money, it’s a for-profit company, of course, why wouldn’t we do that?
But people aren’t necessarily, don’t, they might not have the time to go, Hey, let’s just brainstorm. How do we do more analytics? How do more stuff? Because it’s just outside of their realm of what they’re doing. But when it comes to, you know, if it’s working, then sure. So while what we’re doing, you know, we make these decisions, or I should say we’re supporting the decisions using ADSS, we’re supporting the decisions optimization.
But these are decisions people made, you know, for, you know, like you said, decades. A hundred years before ADSS started. So it becomes this: Is it helping? Is it doing the same thing? First of all, if they’re not doing the same, all they’re doing, they’re not going to use it. Sure. So the doing the same thing isn’t improving it.
So there’s this constant feedback between me and the other people who work on the model itself. And then the people who are using it to actually make those decisions of, is this still making sense? You know, is this supporting what you need? Support? Are you getting what you need out of it? You know, oh, you, you want to use interface to show something that we don’t really have to optimize?
Okay, well we’ll get this user interface for you because we want you to be in the one place. We don’t want you to have to have, you know, we’re not trying to make your life more complicated. We’re, we’re trying to make this, you know, easier. So if it’s easier and it’s also helping, people are going to embrace it.
Evan: I think that’s a good mantra. If, if it makes it easier for somebody, then they tend to embrace it. And I’m thinking back, you know, I’ve never worked in energy, but in the consulting space, touch a lot of different industries and oftentimes there’s, you know, maybe the optimal technical solution is not something that makes things easier people are willing to embrace.
I’m curious if you’ve, you’ve seen that either something that you’ve built or tuned or decision support, that you said this would be the right way to do things, but it’s hard to get the change management or the buy-in from the people who are flipping the switches or opening the day. Obviously I have no idea how hydroelectric energy works either.
But for the people who are actually trying to use the decision support system, do you have to sometimes step back from what may be technically optimal just to get something that, that people appreciate and people trust and people use?
Kathryn: Absolutely. Yeah. Because if it’s not, if it’s not following what they expected to, they’re not going to use it. And even if we, we find that what we’re saying to do is better. If they don’t believe it, we’re not, we can’t convince them. So we had that recently where we have some. I have to keep this very general. But, but we had had a newer feature and we had this going for a few months to help make certain decisions.
And it was taking a lot longer than it should have. Like, it should have been pretty instantaneous. It’s taking hours to run and it got to the point that the people using it, were going, why are we waiting for this? Like, we’re not even getting the answers we thought we’d get And we noticed that they were doing a lot of targets in the optimization.
So rather than just say, okay, go optimize, or say, okay, but, but set all these variables to be these numbers. And they kept messing with them to try to get the numbers they wanted, because in the end they had the schedule that they were looking to get and that they were trying and make the optimization do the same thing.
And there’s. Sometimes there’d be constraints that they weren’t aware of or, or we weren’t modeling the best or things that we were told is always true. That might not be as always true. So it wasn’t, didn’t need to be as hard of a constraints as it was maybe had came down to, okay, they’re not going to use it.
What will the use, what what’ll do, and it’s matter of, matter of, we, we created something that was pretty much just redoing the calculations they had on spreadsheet, but if that’s what they’re going to use, okay, we’ve got to keep working on the original thing to make it so that it is usable. But if it’s not, if it’s, there’s some issue with it, even if it’s just taking too long, makes not use it, then yeah, we’re not going to try to force it and then we’ll just create a disconnect.
They’re just going to make their thing anyway. So why would we, why would we have that happen? And it becomes that now no one’s leaving this now, it’s not going to get supported. Now we can’t do more work on it. And becomes that cycle of no one’s going to use it anymore.
Evan: Yeah. I think, I think that’s, that’s, that’s signs of a mature analytics or, or pieces.
We’re not going to force this, it’s not going to be helpful. And it sounds like you, you know, accepting of the feedback that, that the people are giving. Kathryn, your background is in, or you, your title is, is or you, you do, or is there, is there a broader scope for using data and analytics to be, I guess is, is there sort of an analytics effort there at, at Avista or is, is.
Is it confined mostly to the operations research space?
Kathryn: There is a bigger data science focus. That’s, that’s the word that we like. I realize there’s some controversy or was data science, what does it mean? Data science and as far as I’m aware, we’re still working on, I’m, I’m very like focused on the application I work on, but I do know a data science team exists outside of us.
From my understanding, they’re still trying to get up and going and, and it’s hard because you can’t really hire someone like off the street. Like you can’t just hire someone from the outside of who’s some, you know, very experienced, very educated data scientist because they’re going to come in and it’s nothing there. It’s going to be very hard for them to get going. So it’s that chicken or the egg thing of like, do you have to have the systems to attract the, the right talent or do you have the talent to get the right systems going? And so it ends up be a lot of in-house training and so forth. And so there is the thoughts there, the efforts there, but what people are really doing, I can’t tell you as many details.
I’m just not, I’m not up on, up to speed on because I’m very focused on the system that I, I was hired to work on.
Evan: Sure. And I, I suspect that, you know, maybe I paint too rosy of a picture. I tend to be an optimist. But your, your decision support system and the work that you’ve done, and it sounds like really incorporating the, the user experience to it, that’s going to help to generate some buy-in.
It’s a lot easier to have, you know, the next analytics thing when you are very happy with an, some analytics that, that are already underway and are already making your life easier.
Kathryn: Yeah. Oh, absolutely. Especially cause you know, we have the tools now, so we have, you know, we, we have all these like, Things we do in Python developers who can and use that.
We also have a groovy license that we, that we use mostly for this, but it’s, hey, if someone has like a very small model and they want to test it out, I can run it for them on, on my computer with a license on it and, and do it faster than, than using whatever other software they might using. So we have the tools there and it’s definitely a, oh, hey, what if we did this?
What if we do this? And sometimes it comes down just the time of, do I have enough time to work on something? Does someone else have enough time to explore something? But we, it’s very nice to have the tool set and have the thoughts there of, Hey, let’s do something more.
Evan: Awesome. And on, on the time. Time is scarce. You are the one and the only, and the best, of course, , operations, research focused person at Avista. Do, does the, does the team look to grow? Do you. What, what does it look like for somebody else to come in? Do? Are you sort of growing operations, research talent within? And I’m thinking about a lot of our audience as they’re sort of starting up, maybe it’s not an operations research, but it’s in some analytic endeavor and they’ve had some success and they want to expand and where’s, what are the things that you need to think about as you sort of expand the capabilities and really the size.
Kathryn: Yeah, so first it’s budget. We, as much as we’d like to hire a whole team or people to really focus on this and do more stuff, we might not have the budget right away. And so that’s what’s holding us back right now. There’s definitely a, hey, it’d be nice to another person if it’s not a, we’re not in the place of, oh my gosh, we need another person right now. But we’re not, we’re not there yet. But that, that is already a concern that if once we get there, you know, will we have the, the funding for it. But there is, there is a general plan, I so should say, thoughts. Not, there’s nothing in writing, but a plan that eventually we’ll be able to grow the team of OR people, and I’m always trying to educate everyone around me about the different things we’re using of, Hey, this is why we’re doing this, or this is like optimization. There’s a—I gave a very basic training recently of what mixed engineer programming is. I had a whole example, cause I’m from Georgia about like Waffle House and like optimizing the what you’re going to buy at Waffle House if you have to be there for 24 hours. I had a whole thing.
Evan: Oh, is that public? Is that shareable? Surely it’s just an all-star special all the way down, right?
Kathryn: Yeah. Well, so I’m also vegan and I was like, okay, if I went to Waffle House’s, because I’ve got to go as a portion of Georgia. That’s just where I go.
What’s vegan on the menu? There’s only two things. And so it becomes this, if I only have the two things I can eat and I’m gonna be there for 24 hours. Cause I wanna bet now I have to do this. I don’t understand the idea of losing the bed and having to do this, but if I want a bed and have to do this, what would I eat in order?
Also get the right amounts of like, you know, calories, protein, not too much salt so far, something like that. That was an example I used of just like, Hey, this is what’s going on here. This is why it’s working. This is what happens. Something’s infeasible. This is what this means, and that’s, so it’s just trying to get that base understanding of.
What is this optimization thing? It’s not just, oh, we’re just making really decisions. No, there’s optimizations of certain type of math and certain things happen behind the scenes of just trying to get more people on board those ideas, and eventually I try to have more people who are outreach research analysts specialize in this type of math.
We do student projects. Sometimes we just sponsor project at George Washington University. And have them look at one of our inputs of, Hey, can we do a better job modeling this? It’s a student project and we realize the student project might not have, you know, the answer for us, but it’s those sure. Those little things that we have that like, Hey, it’d be nice to look at this one day.
Sure. We, we can plan the students to do that for us. And then interesting for us to see what they come back with and hopefully interesting for them and it works out that way.
Evan: Yeah. Very cool. And then if, if one day in the future the budget’s there and the appetite is there, you’ve got a network of folks who are familiar and hopefully have enjoyed doing.
Fun and interesting work.
Evan: Doing, doing work other than shifting cells in an Excel spreadsheet or, or pivoting them. Yeah, very, very exciting. And, also from Georgia, love the Waffle House example. I think that’s a very accessible example. Really cool. So you’ve had, you’ve had some success there at Avista at some point that the team may grow.
I’m curious if, you know, you’ve been there for on the order go, going on two years now. You’ve learned a lot about the industry. You’ve gotten to do some very fun stuff. If, if, let’s just say you, the budget is now Kathryn’s to decide and the team is now hers to decide, and you can point your or efforts at any problem that you want, realistic or pie in the sky, whatever it is. Where would you go? What, what do you think would be the, the interesting the, the Kathryn problem to solve?
Kathryn: I think I would expand ADSS way bigger than we have it now. So, for example, we have a lot of predictions that we use, of course, because we’re making decisions for the future. We have these predictions of like, you know, weather and different, you know, when the snow’s going to melt to bring the water into the rivers—all these things. I would do a lot more and we, we basically just get these predictions given to us as inputs. I would see if we can do a lot more analytics to better predict those things. Predictive maintenance to then when should we do maintenance, but also how’s that going to affect the bigger system. And really, cause we, we do have that a bit in right now in the system of like, if we have all these outages we need to do, how do we best optimize the scheduled outages?
But I would make it all one big model. It might take longer. If we get enough machine and enough people working on it, we could probably make it. Much more reasonable and then bringing more other people so people can like, have it on their phone going, Hey, what does it look like right now? If I made this decision, what would that look like?
Because you know, at like a plant, someone might go, okay, we can do summer maintenance. We got to shut down the plant for however many hours a day it’s going to be to, to do something. And they might only be able to see something like the overtime budget of, if we do it right now, how much it does it cost in overtime versus if we wait a couple days.
Now that might be all they see because why wouldn’t that be the only thing they see. But show the bigger system. Okay, if you shut down for this time, how does that affect everything else around the overall system? And then bring up power authorization to people to see, oh hey, if we just waited a day or we did a day earlier, whatever it might be, we’d actually save the company a whole lot more money.
Or it might be a better system overall for these other reasons and just be able to get more users and have even more analytics, leading into the inputs of the prescriptive analytics.
Evan: Awesome. I think that’s a great vision, and I think generally in operations research or data science analytics, people using data writ large, having a vision like that is such a, is such a good thing when you’re, when you’re champion, when your business comes to you and looks for ideas for where to implement or what, what could we do next? It sounds you’ve got a backlog of ideas. You’ve got a lot of things where you could point efforts, things that you can do. And I suspect they’re a lot better fleshed out than, you know, than you have time to talk about in a five-minute podcast question.
But that is super interesting. I think that that’s a great generalizable. Advice for other people to, to do that as well. Not just think about your current problem, but what could be next. And Kathryn, I do want to put you on the spot for, for one last thing. I saw you give a talk at a informs business analytics conference earlier this year.
It was very good. I really appreciate it. Thank you for doing that. I hope you do more conferences and, and, and spread the word. And it’s, it was, this wasn’t the only thing that made it memorable. But you shared some math jokes and I’m a bit of a jokester myself, so do you have any, any jokes on front of mind that you could share for the audience here?
Kathryn: Sure. So I’ll give you two jokes. One that I think everyone should just sneak into their presentations, and when you say you got to just stop and just stare until someone realizes a joke, make sure they laugh. Tell them to laugh if they don’t, and then move before you move on, which is anytime you show a graph, you go, you know, I can optimize.
And I can analyze. Graphing is where I draw the line, and then you show a line graph. If you do it right, it kills. Otherwise we were just stare at you, which also could be fun. And then, and here’s a joke for, for people to just tell to their friends and their family. I, I highly encourage it and I expect to hear back from everyone how, how much this, this joke went well at the, the dinner table.
Why do teenagers only travel in groups of three, five, or seven? Because I can’t even.
Evan: Oh, Kathryn, thanks so much. That’s a great teenage—angsty teenage voice that you can’t even. Thanks so much, Kathryn, for coming on the show today. Super interesting topic. Really glad you’re able to share it with us. Kathryn Walter from, from Avista.