The Data Pyramid

Summary:

At Prodacity, LTC Melissa Sayers, with 23 years of experience in the Army as an Operations Research Systems Analyst (ORSA), shares her innovative approach to effectively communicate complex data science concepts to government leaders. Utilizing the Data Science Pyramid, Melissa demonstrates how to make technical projects accessible to non-technical senior leaders, focusing on practical steps to empower decision-making and resource allocation. This talk provides valuable insights into tailoring communication strategies to bridge the gap between data science experts and government leadership, ensuring successful project implementation and strategic alignment.

Transcript:

Melissa Sayers (0:18)

I am 23 years in the Army. I'm currently serving as an ORSA. I gotta love it. It's an acronym, right? Because we have a lot of acronyms within the government. It's an Operations Research Systems Analyst. I work a lot on the data side, like with data, doing data engineering, doing data science work for the Army. The biggest reason I'm here, and the thing I'd like to talk about today, is communicating with government leaders. So I have about seven years of kind of in the trenches communicating with government senior leaders, specifically like G.O., general officer level and above. Like, so how do you do that? 

(1:00)

I am gonna be using the data science pyramid as a method of doing that, with the idea that you can kind of tailor this approach to the needs of what you're, what you need to communicate. All right, so I'm gonna start with this idea of, I'm gonna refresh my audience. And this isn't necessarily you, this is my leader. I'm gonna introduce an idea and then I'm gonna apply it. All right, so a problem. A lot of times I felt like this, where I have people that are giving me things to do, and it's just piling on. Like one more thing, one more thing, one more thing. And I'm already trying to climb a hill. And I already have a bunch of stuff I need to do, and you're giving me one more thing, okay? So operational government leaders, they may not have a technical or data focused background. They don't necessarily understand how to make the thing it is that they want you to do, but they also may not understand how to influence and resource the project. We shouldn't expect them to be able to do this. That's our job. I have a deep respect for government leaders. They have a lot they need to do. They have a lot they need to figure out. There's a reason why we're here. So we need to help them. So my thought is, why don't we flip the script? 

(2:23)

Now here's the analyst. That would be like me in this picture, and I'm gonna guide the leader forward, with how we need to move forward, to help make that thing that they want to do a reality. Okay? So how do we empower these leaders? Well, use familiar, accessible ideas. I'm gonna give you examples. I'm not just gonna talk about this. I'm gonna show you real examples of things that I've done that have worked, and then clearly identify where leaders can influence and resource projects. So not just talking generalities again. I'm gonna specifically point out, for the thing that we're doing, how I need that leader to interact with it. 

(3:04)

Okay, so step one, refresh. All right, there's this thing called Maslow's Hierarchy of Needs. What's cool, is that in the government, this is something that most government leaders have heard of. So I can just say, "Sir or ma'am, have you ever heard of 'Maslow's Hierarchy of Needs?'" And I get, "Well, yeah," or I get, "What was that again?" And I'm like, "Sir, ma'am, have you heard of this idea that a child, if they're hungry and they're not sleeping at home, it's gonna be really hard when they get to school to learn. They're just not gonna be able to concentrate, because they're not having their basic needs met." They're like, "Yeah, like I studied that in my master's degree, or my PhD program. That was part of my thesis." Or, "We hit that in this project that I was at." Generally, GS 15, kind of across the board, 06 level and above. I really haven't met a leader that has not been familiar with this idea. So you're immediately getting like engagement, buy-in. Like you're not talking about some crazy topic that I have no idea what you're talking about yet. You're like, they can understand you. This is one that's posted on the internet about it. But generally, this idea of being able to learn, self-actualize, you need to have certain basic needs met. Okay? 

(4:35)

So step two. So they've agreed. They know what Maslow's hierarchy of needs is. Introduce. So the idea is I'm gonna introduce this data science pyramid because there's something that we're trying to get to, and I'm gonna show them how we get there. However, there's already a talk about how to use Maslow's hierarchy of needs for the data science pyramid. This is the one that's used. Here's the article down here in medium towards data science. However, this is still not quite accessible. Anomaly to detection, infrastructure pipelines, ETL, like these are not terms that government leaders understand. They're not ones that they're familiar with in general, if they're from a non-technical background. Okay? Well the internet has a lot of data science pyramids, so why not one more. Okay, so this is one that I have tailored for the Army, for the work that I do, okay? But my thought is depending on what you're doing, you would adjust this. Okay? All right. 

(5:45)

So I'm gonna use familiar accessible terms. I'm gonna go through this in detail. All right, so first. At the bottom I have this idea of a secure environment, a perimeter. This is your security layer. Like everything that you do when you're creating a software, you're creating some kind of a solution involving data. You need to make sure that it's completely secure. It's been talked about already that that just needs to be a part of working with the government. But this gets it in their head because if you have a problem with this secure environment, what's great is you can call it out. And I'll show you an example of me calling out a problem with the secure environment, which is why we need to address it. They're familiar with this idea already of a perimeter defense because it's like, it's military, right? Like I need to secure my environment. 

(6:41)

Okay, the next thing up, all right, what did I include here? So the very bottom is data collected in a structured way. What I have is my example down here. I don't usually include this data science pyramid in detail when I talk to a leader, but I have these examples in my head already ready to go so that I can pull them out to have the conversation. This is the type of conversation that I would have at the collect layer. So first we need to collect the data in a structured way. What does that mean? That means that we're not emailing around a file. Like maybe there's like a form or like a workflow or like something that somebody pulls up and it like connects to something else and like lets them do something. It's not a PDF in an email that's then signed and then emailed, and then there's version control issues and a bunch of other things. So is it collected in a structured way? And the example I have in here is that we had a transportation funding request. So this was one my leaders were familiar with. Hey, did you know that you currently have this problem within our organization? We have these forms with these funding requests that are being emailed around and I'm not able to do anything with that information because of that. So I already have a problem here and I can point to it. And if I know what the problem is, I can put something against it, I can put a contract in place to say we need to create some kind of a solution to do this. Or we need to, we need to do something to change that if that's where it's broken. 

(8:24)

Okay, the next layer up, are we limited to only a portion of the data? So I wanna be able to connect all the data I've collected, to the location I'm gonna use it. So I need to bring in all the data in, but data connections, this idea of data connections, I mean that's not important. Like the government leaders are not like familiar with why government, why data connections are so important. But they're essential to what we do. If we can't get to all the data. My favorite thing is we download it to Excel from one of our solutions, then we upload it to another location. I like to call those bandaid solutions. We don't want bandaid solutions. So are we limited to only a portion of the data? The example that I use with leaders, if this is the layer that's broken 'cause I wanna make sure that I'm providing them examples of the place that it's broken and something that they can relate to. What's a problem with a data connection? Well, in the Army we have something called central issue facility database, like it's CIF, it's like wear our uniforms and hats and helmets. Like everything that's given to us as a Soldier. Where all that's tracked in. Well, if I'm an analyst and I actually wanted to do something with that data, I can literally only download 200 rows at a time. So Sir or Ma'am, that would take me 10,000 people eight hours a day for a year to download the whole Army. Then I might be able to answer your question. Do you want me to do that? No, I need a data connection. And then I can answer your question like that, okay? That's why it's important. So it has to be relatable, all right? And that could also involve hiring positions, it could be flipping positions within an organization. 

(10:24)

Okay, the next one up, this data analyst step. So what I like to say is, can we count? Can we count? Can we count how many of something we have? And does it require the specific software to be tailored to do that? Or can my general staff, can my staff within my organization who's trying to solve a problem, can they do this? Can they do Excel like tasks? So if I'm working within a platform that has the data coming in and I'm enabling my, not my technical people, but my day-to-day worker bees, am I enabling them to be able to count? Because can I have someone in my financial section open up a table that has 2 billion rows in it and then do a filter? Because you can't do that in Microsoft Excel. You can't download that many, that big of a table to your desktop and then open it in Excel and then like manipulate it. Like are we empowering them to just do Excel like tasks? Not optimization, not machine learning. I just wanna do like a filter. I wanna join two tables together. I wanna do some logic, like if this, then this. Maybe I wanna make a little bar chart. And if you think about it, if we have in these big organizations, if we have say, I don't know, a hundred thousand people in the organization, and say only 10% of them wanna do Excel like tasks, okay, then we have 10,000 people that are trying to do Excel like tasks, but then each of them has at least 10 questions. So now I'm back up to a hundred thousand questions that wanna be answered. What's the chances of a single piece of software being able to answer a hundred thousand like unique questions. Maybe, maybe not, probably not. You're chasing your tail. So are we empowering the workers to be able to answer their own questions without having to download to Excel? Okay? And my thought is we have data literacy training. So what's cool is that then that leader can apply, if he realized that we're green here, we're green here, but this is where we're broken. My staff can't answer this question, then now we have the opportunity to do training like the Sagely team, how do we upskill? He can now apply that to the problem. 

(13:02)

All right, next one. COPs and dashboards. I love it in the government when we make COPs and dashboards where it's like a download from Excel and upload to Excel, and they don't automatically update and they only do one thing. So they're not drillable, you can't branch off from them. They show one chart and that's what you're allowed to see. Maybe you have some filter criteria that you can like apply on the side, but again, the idea that it can answer a hundred thousand different questions, we need to be able to branch off and do other things. So if this is where it's broken, is everything else green below it? So we have it collected, we have it connected, we can count, but our dashboard doesn't. It's not drillable enough. Well, then that's where the problem is, now we can apply resources against it. Okay? And then in general, I just put predict AI optimization, like the idea of just doing anything beyond just counting in general, the things that you can't do in Excel is at the top, but I usually don't even list all the things. I just say, yeah, guess what? Your problem lies up here and we're never gonna get there because we're broke down here. It allows laser focus on where the problem is. 

(14:20)

All right, now I apply it. So first, have you heard of Maslow's hierarchy? A child cannot learn if they're hungry. Yes. Then there's this idea of this pyramid where we need to collect, connect, we have to have literacy. We have to be able to create a COP and then we can do prediction. Okay? However, now I'm talking to a leader. So he comes to me with a question, he or she, comes to me with a question. The question is for transportation. I wanna know how much I spend and where the equipment is that I ship. This seems pretty straightforward, like Amazon, right? You just like, what did I spend, what did I spend it on? Where's it at? Like that seems pretty straightforward. I should be able to do this, okay? But now I can pull out my little pyramid and orient them. All right, so I am broke here because each DOD service uses different databases for shipping different types of shipments with different databases to track financial transactions. So, well that shipment was rail, so it went in that one, which means that's where it got shipped out of. And then that's where the financial transaction happened. But that one was air. So that one and this other one, and then the financials over here and then that one's ground. Oh, and that one's FedEx. Yeah. So we have a problem right here. And so how do we fix it? So instead of saying like, why can't you make me my COP? Why can't you tell me my answer? I can orient them on the problem and it can be relatable, because these leaders are really good. And if you can hack the bureaucracy and apply that power against the problem, that's what I'm proposing. 

(16:13)

So how do we empower, in this scenario, a government leader? First, demonstrating how crucial getting that data is. How do I do that? Maybe I give them an an example. Let me show you what I could do if I had this data. Okay, now get approval. I need to know each database that I need, maybe get example data, demonstrate what would be possible, and work to potentially create a strategic engagement. This is where general officer to general officer is actually having a conversation about a data connection. What would they need to know in order to actually be able to have that conversation? If it was me and I was setting up a general officer to have a strategic engagement about a data connection, I would wanna know everything I could about that. What type of database is it in? Who's the program manager? Who do they fall under? What's the hierarchy? What are their steps for approval within that organization? It's a lot of work, but that's the type of, this isn't technical. This is like figuring out how another organization works. But if you wanna apply that lever to really make stuff happen, then those leaders need to understand and that's part of figuring that out. 

(17:32)

Here's another example of a application of using this pyramid, okay? Across the board, an organization, this could be the Air Force as a whole, each of the different sections. This could be a specific program within an organization. This could just be a single section and maybe the divisions within a section, but each single one we can look at and say, hey, for human resources, how are we collecting all the data that we use on a regular basis? Is it all connected to the place that we can use it? Oh, you're red on literacy. Hey, I know exactly what we need to do to fix that. Let's get you set up for training. All right, let's try a different one. Operations. Ooh, you're yellow, yellow, yellow. So you do have people in training, you have some of your data, some of your data's connected, so that's good. You're actually probably starting to be able to answer some questions, but maybe we need to like apply some, we might need a contract down here really quickly. Government leaders love green, amber, red, right? So quickly hone in. Like is there a contract solution we can apply to this right here, this collect step. Ooh, look at this one. Maybe we need a education program or ooh, you guys have started, that's great, keep it up. You know, ooh, what's this yellow on the security, what's going on with that? Like we probably need to get after that quickly. What's going on with the security? Okay, so thoughts are collect, connect, prepare the bottom two, continue that yeoman's work to make those red and yellow, green. That's hard work. The bottom two are hard, but to be able to communicate that. The next one, visualize literacy. That's easy. That's scheduling classes for staff and predict we can't get there until we at least, like here, we got some stuff all the way up before we can start that and then secure this example right here is that, hey, if we're transporting stuff via email, that's not the most secure way of doing it. Maybe maybe we need to fix that. 

(19:50)

All right, another one. This was one that we used to just help with who we needed to hire. So the idea here was, hey, it doesn't matter where in the pyramid it is, but if it's quick turn, we can take it on. The next one, if it's stuff at the bottom or stuff at the top, that's hard. And we currently have capacity limits, those are hard people to hire. Bottom, the middle ones, hey, we've got plenty of people for that and we're knocking stuff out. So if they come with a project, where does it fall, and can we do it? Helps 'em quickly orient what type of project it is and if we have resources, right? And the conclusion. So familiar accessible ideas, identify exactly where they can help. 

(20:34)

And here's examples of how they can help at every single leg of the pyramid. Resource security software or hardware solutions. Approve security methods. These are things that general officers can do if they're guided in the right direction. Resource and prioritize collection. Hire people, strategic engagements, prioritize training, empower their staff to create the BI business intelligence solutions, instead of sending it to their more advanced people. COPs, not going for bandaid solutions. Making sure that when they ask for a COP, they realize they have to have a solid foundation, otherwise they're creating organizational overhead and at the top, hey, you want something at the top, but my foundation is poor. We can do everything at the top, but not until we fix the bottom. And that is it. So it's a a real tool you can use to help with conversations. I think the biggest thing is that empathy. When I go to talk to a leader, this is the last thing I'll end with, I think. I actually think about my husband 'cause he's like not technical. He's history, political science, doesn't do like anything with data or math really. But I respect the hell out of him. He's an amazing leader and he's been in the government for as long as I have, for 23 years. He is an incredible manager and can direct teams, but he doesn't understand this stuff. But I need him to understand. I need him to understand how he can apply his leverage to get after it. So how would I explain it to this person that I deeply respect and is incredible in so many ways, but doesn't understand what I do? That's it. Thank you.