image: thumbnail.jpg
Laura Gast @ USO | Data Science Hangout
Speaker bio:
Laura Gast is an epidemiologist specialized in simplifying complex data for high-level decision-making, constantly trying to 'solve' the idea in information communication to "Make things as simple as possible, but no simpler." Laura has a strong background in infectious and vector-borne diseases, gained while shaping malaria elimination strategies in Southern Africa and contributing to public health projects across at least ten countries. She started out in the AI/ML space in the early 2000s with her doctoral research which employed NASA satellite imagery and machine learning to analyze land-use changes in relation to mosquito-borne diseases in Peru.
Currently, she serves at a nonprofit focused on the well-being of enlisted U.S. military members and their families, tackling data challenges in development and impact measurement. Outside of her day job, Laura relishes pub trivia, live comedy or music performances, and long runs on the National Mall.
______________
āŗ Subscribe to Our Channel Here: https://bit.ly/2TzgcOu
Follow Us Here:
Website: https://www.posit.co
LinkedIn: https://www.linkedin.com/company/posit-software
Twitter: https://twitter.com/posit_pbc
To join future data science hangouts, add to your calendar here: pos.it/dsh (All are welcome! We'd love to see you!)
Thanks for hanging out with us! š
Transcript#
This transcript was generated automatically and may contain errors.
Hey everybody, welcome to the Data Science Hangout. Hope you all had a nice Thanksgiving holiday if you celebrated, or a nice week regardless, last week. If this is your first time joining us today, so nice to meet you. I'm Rachel, I lead Customer Marketing at Posit.
If this is your first Hangout, say hi in the chat because we'd love to welcome you in, especially anybody joining for the first time. The Hangout is our open space to chat about data science leadership, questions you're facing, and getting to hear what's going on in the world of data across different industries. And so we're here every Thursday at the same time, same place, unless it's a holiday. If you're watching this recording on YouTube later and you want to join us live, there'll be a link to add it to your calendar in the description below.
At the Data Science Hangout, we're all dedicated to making this a welcoming environment for everybody, so we love to hear from everyone no matter your years of experience, titles, industry, or languages that you work in. Totally okay to just listen in here, but there's also three ways you can jump in and ask questions or provide your own perspective too. So you could raise your hand on Zoom, and I'll keep an eye out here. You can put questions in the Zoom chat, and if it's something that you want me to read instead, you could put a little star next to it, and I'll know to read that for you. But we also have a Slido link too, so you can ask questions anonymously.
But with all of that, thank you so much for spending your Thursday with us. I'm so excited to be joined by my co-host today, Laura Gast, Data Science and Analytics Manager at USO. And Laura, I'd love to have you introduce yourself and maybe share a little bit about your role and something you like to do outside of work.
Yeah, so hi everyone. Thanks for having me here today. So my role right now, I work in the development operations side, so basically it's fundraising data. And like I'd say 95% of non-profit organizations, our data is kind of everywhere and kind of a little bit of a mess. And as happens with siloed information, or siloed organizations, maybe what we call one thing in one database is not the same in the other. So bringing together all of these different components, and we have multiple databases for incoming gifts and making them all talk together, to get at the end of the day, the question is how much money did we raise this month and from whom? And it seems like an easy question, but just like how many clients do we have in our organization? It's usually more difficult than you think it should be.
And so for those that don't know, USO is United Services Organization. If you've watched the movie White Christmas, Bob Hope sings at one of those. We provide services for us currently serving military members and their family. And so that's what I do for my job. My background actually is an epidemiologist, but in 2020 I stepped away from that for a bit. And the thing I like to do outside of this is, outside of my work, is probably running marathons or playing darts. So those are my two fun things.
That is awesome. Where's your favorite marathon location? I love Chicago Marathon. It's just, it's big, but not too big. It's not New York big. 50,000 people is too many people for me. 30,000 is plenty. And then it's always got great weather and amazing crowds, so. Nice. I love it.
Well, thank you so much, Laura, for joining us here. I get to start out with some of the questions while we wait for everybody to weigh in with their questions. But I know you gave a great talk at the Posit conference on design, and I'd love to hear a little bit from you about that. Like, what is one of your favorite lessons you like to share with people about data visualization and information design?
Lessons in data visualization and storytelling
Yeah, so my talk was exactly that. It's kind of my lessons learned and trying to think about how humans think when you're creating a report or a data visualization or even an Excel file, just a table. Because throughout my career, and again, I've worked many careers as an epidemiologist for several different countries, the common thing I came across was I thought I was saying something, but someone else thought I was saying something else. And so flexing those empathy muscles or flexing those stepping outside of what you're seeing muscles was one part of that.
The other part was, I would have to present both to, you know, a minister of health who had four and a half minutes for me. And at the end of that four and a half minutes, they had to sign off on a big spend, but also convince their policy wonks who wanted to see all of the data within it. And I had to do all of that within that four minute window. So it's kind of bringing those two things together and what I learned from that.
My favorite kind of instance of this, I guess, was, oh, let's pick this one. So we were working on a rapid response or rapid reporting system for malaria in South Africa. South Africa is almost to the point of eliminating, and so you get these like little pops, right? So instead of countries that have a lot, where it's just always a lot, and sometimes it's higher, and sometimes it's lower, but it's always going on, in countries that are eliminating, it's very, very, very, very low, and then it goes off the charts, and then you react and bring it back down, and then it's a little off the charts. And the important thing to do is when it starts to go up, if you cut it off, then it's going to come back down, right? So you need to know really quickly when you're starting to approach that threshold where everything's just going to kind of go off the rails for a bit.
And figuring out that whole system and how to demonstrate that this system that we'd spent a lot of time and money based on like sending nurses, sending SMSs, right? Because paper would take months to get there. Demonstrating that the system that we only put in a few Sentinel sites was actually going to catch that threshold. And while, you know, I'm making this up, the threshold would really be 100 cases in this large area, we were able to say at these two clinics, based on all of this past history, we see this thresholds first there. And so nine cases at that clinic really is the problem. And so how do you convince people who are used to thinking 100 is the problem, that nine is the problem?
And so demonstrating these graphics one over the other, this is what the path looks like. Usually, this is what the path looks like at this first, our Sentinel stations. So knowing that that's happening at our Sentinel station, and then tying that again to the economic impact of letting it get here versus cutting it off here, to that whole story. But putting those two visuals together, we actually came up with one single image, which was those two lines where you could see our Sentinel stations as being those highlights. So that was one of my, I was really excited that we were able to nail that one picture to tell the whole story. And then we could keep talking, but we wanted people to go, oh, and then come with us on the rest of it, rather than go like, ah, charts, whatever, just tell me when I need to do stuff. So that's my one good example I can make up.
I was really excited that we were able to nail that one picture to tell the whole story. And then we could keep talking, but we wanted people to go, oh, and then come with us on the rest of it, rather than go like, ah, charts, whatever, just tell me when I need to do stuff.
That's great. Thank you. I wanted to also just give a shout out to your new blog that you posted. Thank you. Because just when you were mentioning like the story in that, I was thinking of like, there's a line in that blog post I shared that says, those who tell the stories rule the world that you wrote. And you talked about how like we're wired to remember stories better than a list of facts. And so thinking about like our own presentations that we're giving, how do you actually like come up with the story that people are going to remember?
Yeah, absolutely. So just a note on the blog, I got COVID, which is why I didn't, there's not a new one from last month. So there are more coming. They're just half written and I slept for two weeks instead of doing other work and I'm digging myself out. So more are coming. I just, they're not there yet.
But talking about humans as storytellers, you don't remember the bullet points. You don't remember the numbers. Well, few people remember the numbers when you're given a presentation, right? You remember, oh yeah, last week, my coworker, Sam, he gave that presentation on that project. And I remember the red line went up, right? Oh, they must've been doing well. I think he mentioned something about them working with that data field and it was going well and they solved something with that other thing. You're not remembering it went up 14%. You're not remembering he joined it and on the 31st of 2018, that's when the data type changes. You might remember it, but you're remembering the theme, right? And you remember that Sam solved this problem. You remember that Sam solves this problem by talking to Paul. You're remembering the meat of it.
And while you're hearing that presentation from Sam, Sam is connecting all the dots for you so that you're brought along in that story. Each of those little story points is supported by the facts of we went up 14%. It was December of 2018, those little factoids. But really what you're doing is carrying someone through a story so that you can say, this is the problem we started with and this is where we ended up. And anybody in your audience should remember the gist of your story. And then they'll have it in their head when they need to come back and solve a similar problem or hire you to come in and solve the problem for them.
So that kind of idea. But humans by nature are storytellers. We don't necessarily remember the pieces. We remember an arc. And if you could put your facts into an arc, then you have a better memory of it. And if you think you don't do arcs, if you've ever written an academic paper, what is an introduction, background, methods, materials, results, and discussion? It's introduction. And then you have all the components of a story, the denouement, the conclusion. Everything humans build is kind of into that arc. Anything that lasts works in there. So that's kind of where my blog, my presentation are sort of in those worlds.
I love this. And I'm seeing a lot of reactions to this in the chat here. So I didn't want to jump off this topic if somebody had a specific follow-up question on this too. Or stories where it worked for you. I always love hearing stories from others. Absolutely. Okay. We'll give you time to weigh in. You can share them in the chat too.
Audience Q&A: neurodiverse individuals and data presentation
But I did see a Slido question. Or no, it wasn't Slido. Let's see. Sita, I see you asked a question. Do you want to jump in? Hi. Thanks. That's really nice because I enjoy storytelling. And I think science is all about storytelling and how effectively we do that. So I guess recently one of my relatives was diagnosed with a neurodiversity condition. And I was just curious, how have you incorporated or what kind of tools have you incorporated in your own version of storytelling to really accommodate for more neurodiverse individuals, individuals who may have faced stress or trauma and don't necessarily see the data in the same way as someone with a different neurological background?
Yeah, absolutely. So I'll take it kind of in two parts. So one is the, I have a family member who was recently diagnosed with a cancer and I'm this person's patient advocate. So I was reading the documentation so as to not stress this person. And so from that side of things, I am always careful to use warning symbols in a very careful way. So think about the color red.
One of my organization's main colors is red. I actively work against using that color anywhere. I don't need to go bad. I will only want red because we're in a Western culture. There are different colors for different cultures, but especially in the Western U.S.E., Canada, U.K. type situation, red is bad. Green is good. Blue is cool, is relaxed, is even, is baseline. So I am really careful to use those colors. In this report for my family member, they used red as the heading because red was one of their organizational colors. And when someone's going panicked, like, oh my gosh, I have this cancer diagnosis, this is scaring me, seeing that red color is going to connect to something deep within their brain where they're going to worry, why is that heading in red? Well, the heading in red is just telling you like what the tests mean and how they work. It's not bad. I mean, the whole thing's bad, but that section shouldn't be with some emotional tag like red, right?
So I'm really careful in using those. Also symbols, you know, an X symbol should mean no or bad or don't, right? A green check should mean you've already done this. An empty box should be this is a thing you have to do, right? And I've developed like consent forms for please participate in our study. Same kind of idea. If there's a big warning for privacy sections, I will put that in red if I have a lot of color because I want people to go, oh, attention, this could be bad, right? I want that attention. So leveraging that color when you need to invoke an emotion is really important. So that's kind of on the one side for the bad news type of diagnostic.
But for, let's say, maybe I had some neurodivergent people in my audience or maybe I've had a boss who was on autism spectrum disorder and would focus on things and I go, why are they focused? Oh, okay. That person is just looking at that quadrant, right? Like, why are they always looking at that quadrant? Well, one, I started to realize this person always looked at that quadrant of a report. So I started putting the main information in that quadrant of the report. So, you know, working with individuals is easier than working with a broad audience. But again, when you're working with that broad audience, basically, I assume I'm in a coding and programming space. The odds of someone being neurodivergent in my audience are reasonably high at any given time. So I'm always trying to make sure, again, I'm weighting information in any presentation in the same way. I don't want to catch that person, you know, this person's
attention in the wrong section first before I tell them the thing I want to tell them. And so in a presentation sense or in a presentation setting, what's called progressive disclosure. I'm sure you've all used it. That's the fancy term for it if you want to sound, you know, wonkish.
But you have a slide, right? And you want to say, first, we did this thing. Then we did thing two. Then we did thing three. Don't put things two and thing three on the slide. Only put, here's the order we did. First, we did thing one. And then tell them about thing one. And then make thing two appear. And then talk about thing two. And then make thing three appear. So even in, and I, you know, fairly neurotypical, I still have a hard time when someone puts that on the screen and it goes, one, two, three. I'm going to go read. They did one, they did two, they did three. And then I'm going to change my attention back to the speaker. And so I've now lost everything they said about thing one because I, my brain was focused on the reading part. It's really hard, even for the best multitaskers above us, to do two things at once.
So making sure you're drawing that attention both in an orderly way without giving later information before you've given it for first information and then the earlier information. And then the second side is using those emotional and attention-grabbing cues very sparingly. So did that answer a lot of that, I hope?
Thank you. That was really helpful. I see a lot of great thoughts shared here in the chat too. Does anybody want to jump in and share what they put in the chat too? I don't mean, I don't want to just call on someone if you're not ready to share. But Ted, I see you unmuted. So I'll have you jump in.
Yeah. So I mean, I discovered this when I had a dyslexic student and she was kind enough to explain to me what she was having problems with. And I think it's, you know, after kind of understanding, like, you know, like it's hard to kind of see things, to group things with dyslexia. Like, you know, that was one of the things that I thought was especially useful was using the Quarto visual editor because she would always kind of cut and paste code blocks and accidentally copy the backticks. And like she was wondering why it wasn't working. So it's just a little accessibility thing. So definitely. Thank you.
I'm going to jump. Oh, I was going to just say on the, like the dyslexia component, there are a lot of really cool websites out there that show you approximately what it's like to read if you have dyslexia. And those kind of helped me gain some empathy a long time ago when I was working with that, with that population in my health role. And those were really helpful. And also the font, the typeface design community spends way too much time, too much, it's useful to me, thinking about how to put space in letters and how you can scan what letters are more easily scannable. Some fonts are more easy to read for dyslexic and they dug into why, you know, serifs and a fat serif and how you weight the letters. And so if you're really working, you know, you're working with that group, there's so much information on that, especially in the font and the typeface situation. Highly recommend digging into, you'll see I use a lot of Roboto because somewhere someone said that that was an okay one for scannability and reasonably good for dyslexics. So that's one I use as a core component for my, for my, the meat of text when I'm doing that.
So I never thought about Roboto being that, which is brilliant. I love it. Thank you. Yeah. I asked in the chat too, if anybody had recommended resources for this topic too, and I see a few suggestions there. So thank you so much, Russ, Steven, and Damian. Oh, there's more, more and more coming in. So I'll make sure to capture those and pull them together as well.
Conflict between hypothesis and data
A Slido question that I saw come in was, Laura, have you ever experienced a conflict with the story you want to tell and what the data actually says? Not usually, because let me put it this way. Many times we start with a hypothesis. We think thing A went great, thing B went bad. But in the course of our analysis, we found it was the reverse, right? So during the analysis, I'm trying to test my story about what do I want to talk about this data? And for the analysts in the world, you know, there's a exploratory and explanatory data analysis, right? And even when you do explanatory data analysis, you're still doing a little bit of exploratory because you're trying to figure out why a thing happened and explain why that thing happened. In exploratory data analysis, you're just going, well, what happened?
So there is a story inherent in the analysis, and that comes from a hypothesis, right? And oftentimes, you know, I don't know, 60% of the time, my initial hypothesis is, or 90% of the time, my initial hypothesis is not perfectly solid, right? I was a little wrong. We found something different. We didn't have the data to prove A or B. And so I will talk about that in my story if I have enough time. If I only have five minutes to give a presentation or present something, or it's only a one-pager, I'm not going to talk about that whole thing. But there are many times where I've been wrong. My story that I thought at the beginning is not the story we found at the end.
And depending on your audience, you might want to build that in. Because if I assumed that it was thing A at the beginning, I'm going to guess most people in my organization also thought it was thing A at the beginning. So you may want to start, you know, clickbait headlines worked for a reason. We've all caught whys to it. But if you thought it was thing A, boy, do I have news for you. Come along on this journey. And, you know, that's obviously a little bit not the best lead for that. But working with that arc, especially if you have the right audience for it, that might actually cement your final, actually it turned out to be thing B, better.
Because everybody in the organization knows it's thing A. Of course it's thing A. We just asked you to go get the data on it because we're a data-driven organization. And you're like, hey, everybody, we thought it was A. It really isn't. And this is why. And I want to convince you it's thing B. And you have to take them on that journey so that they're with you. They weren't in that room when you did 40, 50, 100 hours of work on the data. They need to be brought along. You're giving them the shortened version of that to get to thing B. So it's sort of, you know, there have been stories I've wanted to tell. But from a political situation, I'm not allowed to tell that story. So we kind of have to dance around it a little bit. But usually the story I want to tell is the story that we found in the data. So that's kind of a washy answer. But thank you.
They weren't in that room when you did 40, 50, 100 hours of work on the data. They need to be brought along. You're giving them the shortened version of that to get to thing B.
I see Anjali, and apologies if I mispronounce your name, asked about if you had any internship openings. But I thought this might be also a good time to mention as well. If anybody is hiring for any roles, feel free to share them in the chat as well. I forgot to say that in the beginning. But Laura wanted to ask you specifically too.
Yeah. Let's see how to say this politically correctly. Watch our website starting early in the year. Some things are changing in good ways, and we may or may not have some positions coming up opening for data type folks. Is that, watch this space, January.
Go-to tools for storytelling with data
Sounds good. Another Slido question that came in. And just a reminder, if anybody joined a little bit after the start, there's a few ways you can ask questions here. So feel free to raise your hand on Zoom. You'll put it in this Zoom chat here. But we also have the Slido link that Robert shared where you can ask anonymously. But one over there was, what are a few of your go-to tools for storytelling or reporting with data?
Yeah. So for the me, the person writing it, I always start with, I put blank slides for the number of slides I need. And then I write as the header, what point do I want to make on that slide, right? We started, in the organization, we think thing A. Next slide. Thing A actually turned out to be false. The journey, we started with this. The data was here. And just kind of like the highlights. If you take a storytelling course, it's called the beats of the story, right? So you watch a movie, Elf, baby stolen from home in Santa's bag. Baby finds out human, baby goes to, you know, as adult goes to city, finds dad, you know, et cetera, whatever. So I kind of set my beats on my slides, let's say. Or in a document, I'll just write the sentences that I want to hit those beats and make them headers.
And then I go in and, okay, what do I need to tell that story point, right? Well, I need to tell that these three data points that we usually rely on, restricted doesn't actually mean restricted from that financial code. It actually means this kind of thing. So we did this thing. And then, you know, kind of build in what do I need to prove that. Prove, support, the underlying, the support for that story beat. And then I'll work my way through. Oftentimes, I'm finding myself writing like 40 or 50 words and I go, too many words. Either I need two beats in the story or I need an image. You know, and you can always, you know, line went up 15%. That's an image, obviously. And there's a lot of good information, blog post it, forthcoming about how to tell certain stories.
You can also, stories and data, you know, bar charts, great for context, like contextual, you know, categorical. There we go. Categorical, right? You want to say, we sold more tables than we did chairs this month, or what have you. You can put changes in it. There's a whole world of that. But in those beats, I'll say, I'm trying to prove month over month, there were more, right? So, you know, I'll go into the data viz guru world and think about, okay, I want to show there were more over time. How do people use that? I'll browse around and look, oh, on tidy Tuesday on the old Twitter was great. I'm hoping it ends up somewhere. But I could venture with what others were doing to show that same kind of thing. And sometimes I'd stumble across a graph and I'd go, that tells me that story when I just look at it. I don't have to dig in. And so, I'll steal that information and put it under my story beat.
So, in terms of making this, developing my whole talk, or my whole document, or my whole whatever I'm giving, in terms of my tools for that are work from a storyteller's perspective, work from a screenwriter's perspective, essentially, and work with my story beats. And then from a data viz perspective, there's a lot of information out there. But always my number one thing when it comes to a data visualization is, I'm going to hold it up in front of a coworker for two seconds and then put it down. And I want them to tell me what it was. Line goes up, line goes down, way more of thing A than the rest of the things.
And I want them to tell me what it was. Line goes up, line goes down, way more of thing A than the rest of the things.
I just want that one, hold it up on a piece of paper, move it down. What do you remember? And if they don't remember my salient point, line went up, line went down, more of thing A than the rest of the things. I got to go back to the drawing board on that graph, right?
Leverage maps whenever you can. You get way more looking at a map of the United States and health insurance rates by county than you do a big old table of the counties, right? So there's a whole world on that, and digging in is kind of that way. I hope that answers that question. It's a little fumbly as well, but I think I got there. Happy to answer follow-ups.
Thank you. Yeah, the point about maps is just making me think of last week during Thanksgiving, I saw a map of the U.S. that said like, which states prefer like pecan pie versus apple pie versus pumpkin pie. And I can, when you said that, I could just like very clearly see it in my head and remember it.
Knowing your audience and pivoting in the room
I see, let's see, Alan, you had a question in the chat. Want to jump in here? Yeah, sure. This is super fun. I'm just really, really enjoying it, Laura. Thanks.
I'm wondering, so there are some rooms that don't want the journey. They want like, they want the executive summary, like whatever that means. And I wonder if you've got an approach or a toolkit or a way to be informed about what does this room want and therefore kind of how to calibrate it so that you like do the right kind of preparation and you come in with the right information. And, you know, if not, how do you pivot when you run into that wall, when you, you know, when you start to give the talk, right, and immediately someone jumps in and says, you know, and they just they just rabbit hole into something. I'm wondering how you prepare for that and how you figure out how to respond.
Yep. So on the first half, the prepping for the room, Malcolm Hawker, CDO Matters podcast. He's a guy who works for Prophecy. He's worked for Gartner. I love listening to that podcast when I run because it's it helps me problem solve. And I find myself debugging a lot of code while listening to him. But he has this this this thing he talks about where it says if my C-suite right now in my organization, if I'm developing something for them, usually where the request came down a couple of levels to me and said, hey, they want to know this pivot table. I'm like, they don't want to know the pivot table. They don't want to. They have a business question they're asking and they don't know how to ask down the chain or someone down the chain doesn't know how to ask down the chain for the spirit of the information. So they can only ask for the letter of the information. Right. There's a business question. They don't want the data. They want to solve a business question.
And so Malcolm Hawker says, if you can get in that room, get in that room. You're not always going to be able to be able to in the room, but you can ask people who are in the room or you can ask people who work directly with someone who is in the room. You know, there's a VP and I talk all the time with one of her her assistant type people. And so I was like, hey, what did what did Sarah Jane want? Sarah Jane, what's what's she working on right now? I've got a request from her. Can you give me any more information on it? And so leverage that network back up. You're never working in a solo position ever. I hope if you're an independent consultant, you might be. But
hopefully you've given a little bit of leverage to at least ask some people what's going on and help guide you on the right track just to see, oh, they're looking at this huge project over here. I don't know much about it, but it's all about the West Coast, whatever. So you can think, OK, let me incorporate that into my synthesized analysis for them to make sure I'm hitting the spirit of the ask rather than just the letter. So that's one part.
The other part of the I think I've got my great story. I think I've got all my data and all my story beats ready. I go into this meeting and on slide two, they go, well, of course, it's thing A. What about thing C? And I'm like, I only thought about A and B. I didn't think about C. The upside is if you are the one that did the exploratory analysis, you probably saw something related to it. And I always hate giving the answer. I don't know, because sometimes you don't know, but you kind of have to say, I don't know. But you can say, oh, well, there were some inklings of thing C. We saw this, or we don't collect that data point. That's why we didn't talk about it. But we could. Here are some ideas that I've thought about while drowning in your data earlier. Or you can kind of pivot a little. And this is always why, if I ever give a presentation that I did not do more than 60% of the work on, I have that person in the room.
And I have been asked to present for my team on many. There's five analysts working on something, and I'm just the one bringing it all together. With every ounce of my fiber and power, I will have at least the main analyst in the meeting with me. And I'll have a little side conversation.
Allison, hey, can you step in here? Do you got anything? And Allison will be like, yeah, cool. And I'm like, I'm going to ask Allison, because Allison did the meat of the work on the West Coast. And I'll have Allison come in. It never works well when someone wants to spin something off the rails. That's also why, at the top of every presentation, I start with my hypotheses. I start with the beginning of the story, which was someone said, hey, I'm wondering how we're doing in California. So I start there, and I say, but it's only northern California. And then that's the path I took, because I don't want them to think I didn't think about their hypothesis to start with. I want to make sure that's all in there. And then the other thing is, in most rooms, well, we only looked at these things because that was our starting point.
There was some information there. The data wasn't great, which is why we pivoted away early. But let's talk about this concept, and then I'll take this conversation right now back to the drawing board, and we'll develop that and give you a follow-up deck or document or presentation.
It's led to disappointment in some meetings I've had back when I was doing more consulting-type work, but we never really had anybody mad.
Sometimes you're never going to get everything. So I hope that sort of answers. Yeah, absolutely. Yeah, thank you.
Thank you. It was actually just making me think, too, about even the presentations that aren't like data-related, but when you're showing somebody something that you're really excited to share them with them, and then they have great ideas because they got excited, sometimes I can get distracted by their great ideas and wanting to address that in the middle of the presentation as well. I see Jeff, you had asked a question a bit earlier. Do you want to jump in here?
Five-minute presentations
And Jeff, I think it was around like giving five-minute talks. Let me try. You can jump in if you want to add more context, but the question was, Laura, what would a five-minute presentation look like?
The five-minute, so I'm going to, we're going to go back to my little fake example here of we thought thing A and it ended up being thing B, right? So if we're going to talk in the meat of it and we're going to get down for the analysts, it's going to take longer because they're going to want to know every variable, right? But if I'm just presenting this to the C-suite or in my example of I have five minutes to get a minister of health to sign off on this big spend to fund my big research study or fund my program over here, the starting point of that, rather than going through the beats of my story, the starting point is what do I need them to know? What do they absolutely have to know when they walk away from this five minutes? And it's thing B was right, even though we all assumed it was thing A. It was I need half a million dollars because if I don't fund this, then this big thing is going to happen. I need the thing. I need you to know this thing.
And if you always go back to that touch point in your five minutes, then you'll be successful. And then with that touch point, make your little mini story, right? You start with, hey, minister, I need you to know thing B. I got five minutes to convince you. Here's my five minutes.
We all thought it was thing A. We did some analysis, and I'm very happy to present you a longer document we have later. But the main points were the data we had to support thing A was wrong. When we started to look into it, it really turned out to be thing B, even though the population was smaller. And then we also were looking at this other thing. And to solve thing B, we need this money. And so rather than I'm writing my story beats and filling in underneath, it's I'm starting with the absolutely critical thing that they need to know and then building from there, right? Building the mini story underneath it. Everything has to point back to that salient point. So with the line, if I'd had more time, I would have written a shorter letter. It's the same thing, is it takes way more work to do five minutes than it does to do 25 minutes. So play according.
it takes way more work to do five minutes than it does to do 25 minutes.
Keeping teams ahead of data-savvy leaders
Awesome. Thank you. Eugene, I see you had a question. Want to jump in here? Yes. So in my experience, I've sort of witnessed that in the modern, you know, last decade and a half of just data becoming the reason that people choose their business actions or policy actions, that everyone is just becoming a little more data savvy. And, you know, the leadership sort of really starts to cling to, you know, data and want to use that data to support their decision making. And so junior analysts or other people on the team will get blindsided rather fast sometimes by the leader that, you know, becomes more and more data savvy. So they start asking for more slices, more interactivity so they can find this stuff out literally in the meeting, you know, when you're presenting this stuff. So how do you help you? What's your way of, I think, helping teammates? Because I think through what you're talking about, it's pretty obvious you're already doing this yourself. You're thinking about this yourself. But I think teammates, you know, younger people who sort of are looking to that leadership are trying to figure out how to be ahead of those questions. And I don't think that's an easy thing for them to get a grip on until they've been sort of burned, you know, like 10 times. So I'm just curious if you have some strategies to keep people ahead of these, you know, Luddites that become more and more and more data savvy and start asking for more and more complex stuff that takes the data analysts much more time to peel apart.
Yeah, absolutely. So again, as I tend to do, I split it into two parts. So I'm going to start with the easier part, which is how to get ahead of the questions you're going to ask. So, you know, let's say I'm a middle person between a junior and a senior member, and that senior member is that VP, let's say, is the one that always asks way too detailed questions. I'm going to make sure I'm working with that junior member to know, hey, you know, VP Jane, VP Jane asks all the details, right? She's going to get in the weeds with you. Be prepared to get in the weeds with her because she's leading this meeting. And never just be able to never respond to Jane. Always respond to Jane. Okay, let's have a conversation, you know, off of the sidebar on this. And let me get into this details with you and we can walk together through this. It might not be worth building a whole dashboard for Jane because really Jane has an idea in the back of her head that she's trying to get out with the, well, I want to segment it like this.
And Jane doesn't need to segment it like that. She's just wondering, like, what happened to this group of people, right? So maybe you don't need to build a dashboard for it to use all that time. But I've had good experience with people who are really are data savvy. They're good with me live coding with them, right? They don't need a whole dashboard where they can interact. They're oftentimes okay with me just doing some live code, like, hey, cut it like this and show me what that crosstab table looks like. And they're great with it. So prepping that junior member saying Jane asked the questions, she's going to want you to go two steps farther than you did. So prepare to get ready to go two steps farther. Or especially if you have a good relationship, say, hey, Jane, let me, you and this junior member, the three of us are going to have a meeting and we'll cut this every way you want in this one hour shared meeting. I've had really good results with that, especially with truly data savvy people.
The problematic side of this is throughout my entire career, I have worked with many people who are data savvy in that they can use curated products for themselves, but they cannot use the chaotic mess, right? I am right now in a chaotic mess of an organization. It is very hard to rapidly turn around some results for people who are like, we're a big organization. Why can't I just know how many constituents we have? Anybody who's in data, think about it. How many constituents do you have in your organization? What do you mean? Finance has a different definition than does product sales, than does the fundraising side. We've got volunteers as a constituent. We've got service members as a constituent. We've got donors as a constituent. We've got congressional members as constituents. What is a constituent in our organization? That's a hard problem.
And it sounds like it should be easy, especially if you've always been served good data products, a great, solid enterprise data level organization. So that's where I've run into the problem, is where people go, well, give me all the data. But they don't know the context of the data. They're used to using a curated data product. They're used to more of a data scientist role than like an analytic engineer role, right? They came up through the data science, but they don't know how all this control flow processes work and why it doesn't talk to each other. That's where I've struggled. And if anybody has good ideas on that one, because you can't exactly tell your CEO, no, you don't understand what you're talking about. You have to answer that CEO's question as best you can without giving him, here's 22 million rows of data. Have at it, Mr. CEO. I don't know what to tell you, because you'll go a little crazy. And you'd also don't want to spend hundreds of analysts and dev hours building a dashboard for something when he only wanted that one question, or he's never going to use it. It's going to go to the graveyard of dashboards. So if anybody has any good ideas on those folks that are genuinely good-hearted and best intentions, it's just they're only used to working with beautiful data products. Anyone?
I'll watch for unmutes to see if anybody wants to jump in on this.
But I saw, Philip, you had- I was going to quickly say, you could scare them with the rough, you know, consolus font output of a table function that you run in R. And just snapshot, drop it to them, scare them a little. Yeah. A little bit of a red flag. Back off. I know what I'm doing. Tell me your business problem and I'll help you. But just give me data. It doesn't work. Yeah.
Philip, I see you had a follow-along question to Eugene's question and just wanted to make sure we covered that in the answer too. And I heard somebody was about to jump in. So I'll have you jump in first and then we'll jump over there then.
I was just saying that I've run into some folks that were more of, let's just- why don't you look at the data as a data scientist and you tell me what our problem should be that we're solving? One part lucky and one part me at my pay level? Do I need to play the strategy of the org? I mean, cool. But- You get the- well, you're good at looking at things and understanding how data, you know, getting answers from data. You should be able to also come up with the questions. That's something I've run into with certain executives where that's one that's- there's nuances to navigating that.
And now I'm curious, how do you navigate that? I go back to, I need to understand what we're trying to accomplish because that's the lens through which I will view the data. And without that, it's all just blurry to me. It also sounds like you've been cursed by being good at your job. I've also got 27 years in the same industry. So I have a fair amount of experience. So I have a fair amount of domain knowledge. So that's where- that's where that's coming in. Really? Yeah.
I think this may have been addressed a bit in the answer, but I saw Philip had said, cool question, Eugene. So added on to it and said, we get a lot of workshop-like meetings where you get a broad question thrown at you and doing the explorations and a lengthy meeting with management. How would you prepare or anticipate for that type of meeting?
Anything you want to add to that? Yeah, just the way I've done it is, you know, make sure you've got all your, you know, if you're small world, non-enterprise, make sure it's all on your desktop, in your environment, ready to go. Make sure you've got it pushed to something like a Parquet file that's really rapidly pivotable, especially if you've got big, big data sets. You know, if you're lucky to work in, you know, we work in Databricks, we've got a Unity catalog. So I can go connect to everything that I have permission for. I also have a great partnership with IT, so I know what data is available, if I have a reason to ask for it. And so if I'm in that meeting and they go, well, what about the volunteers? Oh, well, if you'd like that, just send me an email and we'll get the IT to approve my need for that. But, you know, because we have, you know, security permissions and siloed, siloed, siloed, I haven't had access to that. So really, those types of meetings, I've had several of them, like my example with, you know, Jane, the VP and a junior member, live coding, just like slash it this way. It's been like, make sure that your, you know,
digital desktop of all of your things, everything's within reach of you to be ready to go. And then you're both going to do your best to answer that VP, that VP Jane's questions. But also you're going to be seen as someone who can get stuff done really quickly. And so questions might come out of that that help you do the rest of your job.
So I, those are stressful, hard hours. They're really hard and you have to be quick and on your toes and know everything's right there. You can just grab it, you know, I can just grab, you know, whatever, my hair tie or whatever, I can grab it and then serve the information quickly. And if I can't totally serve the information quickly, we have that question, that hypothesis that we know Jane is curious about. And so those are so tough. And if you get to go into one of those, it's probably going to be really hard and really, really useful. So any chance you have for one of those, grab it. If you don't have Tim's 27 years in the business and know a lot of it already, if you're in that one, well, I think if you've got those 27 years of experience, I think you've got a little bit of a leg up on some of that.
Audience Q&A: storytelling for academic audiences
Thank you. I see, let's see. Sarah, I see you asked a question in the chat. I'll read it because I see the little star here. But the question was, I work in academic research as a grad student, and I often have trouble telling a story because it takes so long to get people the scientific background they need to actually care about the problem we are solving computationally. Any advice on keeping the background engaging or working it into the story along with the data?
Take the time, take your time and practice on your friends. My sister heard 400 versions of my doctoral dissertation presentation conservatively because I kept testing messages on her and she understood virtually nothing of what I actually did. By the end of it, she understood a lot of it, but not a lot. But that idea of that background is huge to get to we went here.
And that's part of why I have the story beats. What do I need them to know as we go along? On this slide, I need to take two minutes to make sure they know gene ABC4 is critical in these 11 diseases, right? Do I need them to know ABC4 upregulates this and downregulates this? But when it downregulates this, it then upregulates these two other things. Do I need them to know that it down and upregulates all of these certain things? Maybe not. Maybe I just need them to know that ABC is involved in some way in these nine conditions.
If you need to talk about each way of those conditions because it has an impact later, right? Maybe it's these five conditions, it's an upregulator and these three, it's a downregulator. That's that critical component. And later when we're talking about our analysis, we found it affected only downregulators, not upregulators. Well, then you need to spend a little more time on that touch point. But building from a story beat and everything within that story beat of service of that beat is a good way to tie it back. And then of course, practice on someone who doesn't know anything. Find your best friend who works in, she's a finance person and you're a geneticist. Tell your finance friend about it and see if she can follow the story even though she doesn't have the detail, right? And then get your lab partner and say, hey, lab partner, when I tell you this story, does it remain faithful to the extreme detail? Because there's always going to be an abstraction. There was a great, when all the posit talks are up, there was a great talk about leaky abstractions, one of the keynotes, I think.
Wonderful talk. Go pay attention to that talk. But I always, now I'm trying to pay attention to where my abstractions might be leaky and give a little note, maybe it's just a little footnote of just, hey, there's a thing you might want to pay attention. This is a generalization, but upregulator, downregulator, and this is where the change is. There's some downregulation in these. Do they need to know? Maybe not. So it's sort of go back to that, make sure you're hitting your touch points. And if it's the five minute, it's in service of that key message. If it's the longer presentation of your story beats.
But building from a story beat and everything within that story beat of service of that beat is a good way to tie it back.
Thank you. I don't always get to read the chat while the Hangouts are happening. I just wanted to remind people, you can save the chat because I see there's a lot of great resources being shared in there. There's that little three dots at the bottom of the chat where you can save it. But I was laughing at these office jargon phrases that you all are, office jargon that you're all sharing there, like boiling the ocean and all the low hanging fruit. There's quite a few in there.
Using appendix slides
Ethan, I see you asked the question a bit earlier in the chat. Do you want to jump in here? Yeah, sure. Laura, I was just curious if you're giving like a slide deck presentation, to what extent do you utilize an appendix where you have slides on details, whether that's details you feel like you have to present to validate some of the stuff you're doing or the points you're making or just questions you anticipate. To what extent do you utilize that? Or do you try to keep the presentation really, really clean and save all that stuff for just if there's a follow-up question on it?
I always have extra slides because, again, my original presentations are always like an hour and a half. So as I'm building, I want to put my gene example, you know, here's the gene ABC. It's in these five and these three. I'm going to want to tell you 400 things about all of these things. All eight of these, I'm going to want to tell you. It's mostly a down regular, but it does these nine other things. And this one actually is involved in this thing, but not these things. And sometimes only in men, but not usually. Whatever. I'm going to want to tell you a lot of stuff about these eight, right? And so when I build my original talk, I'm going to overbuild that slide. And then when I go back through on my cleaning round, again, if I had more time, I would have written a shorter letter. I'll go back again and say, okay, I have six slides for this. These saying ABC is related to these eight. What in here doesn't go back to that touch point of I need you to know ABC is involved in a lot that continues the rest of my story. And I'll have a slide on each of these eight things. All of those go into the appendix. All of those are for later slides because someone in that room, to someone's point earlier, might want to try and throw me off. Well, you didn't consider that thing. That second thing is it really, you said it's an up regulator, but it down regulates these 400 other things. Well, hey, Mr. Guy, here's the slide that shows why I'm putting it in the up regulator category, right?
So when Mr. Snooty Guy wants to throw you under the bus, you've got the slides where you can go, Mr. Snooty Guy, this is why I put it over there. Super happy to have more conversations with you about this, but really we're here to talk about the, whatever I said, those down regulating three, and that's why it got put up here. So I use them all the time. And usually because I started way too long and they got pulled back. I will make sure some of that might be really, really needed to tell the whole story. That thing about thing two being an up and down, whatever I made up, that might be really critical to the rest of the story of why it went up here and not over here, right? So I need to include that information because it's in service to that story point I need to tell. So it might stay in, and then I have to shorten another 10 minutes, somewhere in that speech, I'm going to have to drop some information. But if you're only given 10 minutes to talk, you absolutely cannot tell the full story of an entire gene. Not going to happen, right? No chance at all. They didn't give me enough time. So going back to the, what do they need to know? And what are the story points to get there? Service of that, and then you've got that extra stuff that you had to cut out, cut to size in your appendix. I may or may not cut those appendix slides out if I have to share, like, great, I want your slides. I might not include those. If it's a recorded talk and we talked about that second thing, I'll put that slide in at the end so that if they're watching a video, they can see that slide. Or if we want to talk more about that slide, I'll put it in the middle, and then that can spur more conversations. But I almost always have extra stuff that just made my talk too long.
A glimpse into the role at USO
I can't believe how fast this hour is going by right now. But I know I said in the beginning here that I love in the Hangouts where we get to kind of just get, like, a little snippet into what life is like as a data scientist in different industries. And I was just curious if you could share, like, a high-level project that you might work on in your role, not without going into, like, specifics of course, but just kind of giving us a glimpse into it.
Yeah, so because I'm on the development operations side, the, like, the number one thing is where did we get our money and from who? And are there, you know, we have managed donors, you know, Joe Billionaire is going to have a special fundraiser assigned to them. But if you're just giving us $200 at the end of the year, you'll get a nice thank you and maybe a Christmas ornament. But we're not, we don't have a person managing you, right, to make sure that we have a good relationship. This is the same through all, any non-profit that does fundraising, it's the same kind of idea. But there's also that middle group of people of, well, maybe they're writing us $3,000 checks every year. Should we put them in the, should we give them a fundraiser? Or maybe not. Or this guy only gives us $25 a month. You know, this lady gives us $25 a month every month and has for 10 years, but our wealth rating says she's got a lot of money in the bank and she seems to be 82 years old. Maybe we should talk to her about a bequest. So let's put a person on her, even though she's a small giver.
So broadly, my reports fall into, like, three categories. Like, how are we raising money? And that might be a big, you know, quarterly report or something to go into a grant application. And then how are our fundraisers doing with their portfolios? And then what about these middle people that maybe we should have a guy on? Can you go, like, dig into that middle slice, that middle demographic, and find us some people that maybe we'll want to check out and do some research on? So it kind of falls into those three. And I generally am only reporting in the fundraising, the fundraiser space. I don't do a whole lot on, how many people clicked on this Instagram ad versus the dog Instagram ad versus the Veterans Day Instagram ad. We have another team that does that kind of the direct response kind of stuff. I work in the fundraising world. So bringing lots of stuff, weird stuff together and saying, is this person, is this John Smith the same as that John Smith? We do lots of the John Smith problem, which I'm sure anybody who's taken a data set classes, done that, and also set your hair on fire for how many constituents do we have? I don't know. Are these 84 people the same John Smith? Your guess is as good as mine. So a lot of that world.
Tool tips and dashboard design
Great. Thank you so much, Laura. I see there's a few questions that we might not get to. I just saw a few from the chat, but I just wanted to give everybody the opportunity, like, we have a few minutes left if you want to raise your hand here, if there was a question that we didn't get to of yours. And I'll just keep an eye out while I ask this one. I see Joseph asked and put a little star there, so I'll read it. I have a team member who wants to add tool tips to everything, but I worry that some audiences are overwhelmed by those details or that adding those options makes some people dive deep too quickly without seeing the big picture in the visualization. Do you have a way of thinking about when tool tips or a table view are useful and when they're distracting?
Yeah, that's always the, a lot of the times I go, we should, but we shouldn't. And I fight myself into a circle. And then at the end of the day, the highest title in the room is the one that decides whether or not we have them. If the CEO wants all of them, they're going in. I spend some time maybe designing that tool tip so that it's less intrusive or it's easier to scan. And again, in product design world, there's lots of talk about how the nutrient label on food, how scannable is that? That's a tool tip on a physical product, right? So I spend some time designing how that tool tip might pop up. But if the CEO wants it, it's happening. If a low level analyst wants it, or not a low level, but down the chain analyst, and this is being delivered to everybody, maybe I turn it off for everybody, but offer a packaged version to that analyst who really wants to do some exploration. If I can get the space on all of our servers, I'll have two versions of that dashboard, the executive version and the analyst version. And so sometimes the CEO wants, this is what you get shared at, unless you ask too many questions, then we give you this one, which hopefully spurs more questions. So it almost ends up as the highest title in the room. And then if I have fellow curious data people that I try and serve them because they're trying to answer questions too, or find more questions. So kind of a wishy-washy answer on that one, but sometimes that's what it is.
If the CEO wants all of them, they're going in. But if the CEO wants it, it's happening. So it almost ends up as the highest title in the room.
Thank you so much, Laura. As we get to the end here, I'm just wondering if people want to connect with you, what's the best way? Is it LinkedIn or through your blog? Yeah, LinkedIn or I'm on threads, but my threads is kind of not professional, but I feel like this is not professional. It's just fun nerds making jokes about bad product design. So LinkedIn is the best, the easiest place in a professional capacity.
Awesome. I did just want to share, we do have a data science hangout community LinkedIn group, maybe sometimes not the best way to have back and forth conversation, but we try sometimes, but I shared it in the chat there. If anybody wants to join that group to make it easier to connect with each other as well. I love getting to hear about people who are friends now because they met in the data science hangout or maybe met each other in the chat, sharing resources.
But thank you so much, Laura, for joining us today and sharing your experience. This has been so fun. Absolutely. Thank you, everyone. And I'm very open to questions on LinkedIn, just send me questions. Awesome. Well, thank you all. Have a great rest of the day.