Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Episode #93: Robert Kosara
171 Plays7 years ago
Welcome back to the PolicyViz Podcast! I’m back! After a few weeks off this summer to enjoy the nice weather and time with the family, I’m excited to be back with a whole new slate of guests. Over the next...
The post Episode #93: Robert Kosara appeared first on PolicyViz.
Recommended
Transcript
Introduction and Podcast Purpose
00:00:11
Speaker
Welcome back to the Policy Viz Podcast. I'm your host, John Schwabich. Thanks for tuning in. It's been a few weeks since I posted a podcast episode. I took some time off over the summer to rest and recharge. And now I'm all set with a new set of episodes with some great guests coming up over the next few weeks and months.
00:00:29
Speaker
And before I get into this week's episode, I want to take a moment just to reflect on the goal of the podcast and why I'm doing it. Over the last few weeks and months here in the US has been sort of a challenging time. There's been protests about race relations here in the US, terrorist attacks in Minnesota and Charlottesville and Kansas City and other places, and of course terrorist attacks around the world, perhaps most notably in Barcelona. But reflecting on why I do the podcast,
00:00:58
Speaker
What I'm trying to do here is to bring experts in the fields of data and open data and presentation skills and communicating data so that they can help you, the listener, improve the way you do your analysis, you do your communication of your data.
00:01:14
Speaker
Because if we can't do a better job of communicating our information and communicating our work, it really doesn't help anyone. And in this era that seems to be challenging facts and data, I think it's even more and more important to think about ways that we can
00:01:29
Speaker
present evidence to our audience, communicate that evidence, communicate those facts well so that we can make the best decisions to help our communities, to help our country, to help each other. So I hope you are able to listen to these episodes and get some expert advice, get some expert tips and tools and resources so that you and your teams and your organizations can improve the way you do your analysis, can improve the way you collect or use data,
00:01:57
Speaker
and hopefully improve the way you communicate data to make the case that the things that you care most about are worthy of people's attention and can be put into practice and can affect change and affect policy.
Guest Introduction and Storytelling in Data Visualization
00:02:10
Speaker
So I'm very excited to have on this week's episode, no stranger to podcasts, no stranger to talking about data visualization, and to talk about an issue that's close to my heart and something I've been thinking about a lot lately, which is the concept of story. So I'm excited to have Robert Croustard on the show from Tableau Software. Robert, welcome to the show. How was your summer?
00:02:30
Speaker
Hi, John. Doing well, thank you. And my summer was very interesting. We all, I guess, took a bit of a break from podcasting and from blogging. So I'm also ready to start things off again. Yeah. And now we can start the annual tradition of arguing about pie charts.
00:02:46
Speaker
And yelling at each other on Twitter. So good. Everybody can get back to work. I wanted to have you on the show because in the spring I think you put out a couple of papers you were working on. One on your own and one with Jessica Holman and Heidi Lam both on stories and how this concept of story applies to data.
00:03:07
Speaker
And I've obviously done a lot of thinking about this done a lot of writing on it but I wanted to get your take on how we as people working with data and visualizing data. Should tell stories whether that's a word we should even use when i ask you just to talk a little bit about the research and also your thinking on how we use this word stories with data.
00:03:29
Speaker
So this is something that's near and dear to my heart because I've been doing work in this area for a while now. And I've been looking at a lot of news pieces and what people like to call news stories and looking at their structure and like what kinds of data they use, what kinds of transitions they use, things like that. And over time, so I have these classifications of they're using this kind of transition type and then I use like
00:03:57
Speaker
Scott McCloud's classification for transitions and there's Neil Cohn who has done some work on also comics so McCloud wrote this book Understanding Comics and he talks about a lot of interesting things there but one of them is the kinds of transitions between frames.
00:04:13
Speaker
And then there's Neil Cohn, who talks more about overarching structure, like what kind of kind of semantic role the different frames play. And then as I'm doing this, I found myself just having another column in there that says, you know, this is not actually a story. A lot of these ended up being things that start out
00:04:34
Speaker
in a way that looks like a story but don't actually have an ending. So they end because they can't go on forever, of course. At some point, there's just no more step in the stepper or there's no more space to scroll to because there's just nothing there anymore.
Defining and Analyzing Story Structures
00:04:50
Speaker
But it's not because there's an actual ending there.
00:04:55
Speaker
brings it to some kind of closure or to end the story in any kind of structured way. And of course, in internalism, there's this idea of the inverted pyramid or the inverted triangle, where you give people the important information upfront, and then you give them more. And you know, as it goes on, it kind of becomes less important so that people don't have to read the whole thing. Or even if they don't read the whole thing, they will still get the gist of it and the main message. But if it's supposed to be a story, there's supposed to be at least a beginning, a middle and an end. But many of these things
00:05:25
Speaker
just entirely lack is an end. And so one of the two papers, the short paper that I wrote was essentially on that, that I said, well, I found a small number, like a half a dozen or so. Well, I should be a bit more than that. But I found a small number of stories that I thought were really good stories that in the sense that they had an actual ending, which is pretty rare, where the entry ties back to the beginning, and then a subset of those, and that's about a half dozen or so, had a very particular structure where they set up,
00:05:53
Speaker
a question or they ask a question or they make a statement and then they present some evidence and then end up with a conclusion that really ties back what was just presented as evidence back to the beginning. So you really have an arc and you have a sort of like an essay or like a more of an actual structure that brackets the whole thing. And so it turns into an actual story. And so that's what I'm thinking about is like, so what does it mean for the other things? Are those
00:06:21
Speaker
stories are those I mean they're still called news stories but
00:06:26
Speaker
What are those? And in many cases, I think they're not, they're just a different thing and not, they're not stories in the usual sense. So in some sense, there's classical definition of the story, but we all might intuitively think of when we hear the word story, you know, little, little red riding herd, right? Or like, right. And then there's news story, which in some ways is different because I like that idea that it doesn't tie back to the beginning. And then there's maybe data story.
00:06:56
Speaker
which is also a different thing and how does... Yeah, so the question is where does data story sit? Like, where space does the data story go? Right. And also, even with news, and that's why I've been looking at news stories because of new pieces. So I like the term... Yeah, new pieces. Okay, yeah. The idea is that...
00:07:19
Speaker
is to look at those and say, well, what do people do? What do different kinds of news pieces do? And some of those are just accounts. They just say, well, this happened, and then that happened, and then that happened. And so you get this sense of, okay, so this is what was going on at this particular point. And some of these have
00:07:39
Speaker
a sort of an implied causality. So they basically say, well, this happened, and then that happened. And because of those two things, the later thing also happened. So they, of course, they picked those because you can't just tell everything that happened. So you have to be specific and focus on just the things that you think are important. So there is obviously a choice there.
00:07:59
Speaker
But it's not always very clear what that choice is based on or why those things were chosen. But that's one example. And then there's this thing about what's called an explainer. And those often are also kind of historical. So it says, well, there are racial tensions here and there. And then it talks about what happened in the past. And that also very often doesn't really explicitly make
Storytelling Principles in Data
00:08:21
Speaker
the connection. But it does talk about things that probably or likely had
00:08:26
Speaker
had an impact or an influence on what happens today or what just happened and what was newsworthy last week, whatever. And so those are things that people build. And then there are more technical ones like explaining how a machine works or how a phone call works, a cell phone call, how is that connected and stuff like that. So there's more technical explanations. And then there's the
00:08:50
Speaker
The background like it just there's just more stuff around what's happening there, but those are so I'm trying to get to the point of saying well there is there are there are reasons for presenting information that.
00:09:06
Speaker
that we like to call story, but that often have a different intent. They're more about explaining something or laying out information or just kind of retelling what happened. And some of those look more like traditional stories, but I think many of them actually don't. And they're more like...
00:09:22
Speaker
So to your mind, what defines for you the traditional story? Is it that it resolves some conflict at the end and it's not sort of the traditional explainer? And then if I were to take that traditional story definition, how would I be able to apply data or data visualization to that model?
00:09:45
Speaker
I haven't entirely made up my mind on that. But I think what a traditional story, if we call it story, there needs to be some sort of arc. There needs to be a reason why it is a unit and not just
00:10:01
Speaker
a few steps through some pieces of evidence. And if you think, or actually what I should have added to this too is like presentation. So there's this idea of I'm presenting something to you, I found something in my data, so now we're getting closer to data and not just like, you know, general stuff. But if I present something to you, often I have a conclusion at the end. And that conclusion is a recommendation or
00:10:26
Speaker
some kind of problem statement that says so here I just showed you a few things so I think that so maybe I found that we're losing money on x y and c and we didn't we don't know why or why why is it different now than it was before whatever or why is our thing thing not working the way it used to in the past whatever and so I'm looking for
00:10:44
Speaker
not necessarily for recommendation, but at least I'm looking for an explanation. And or I've done some research, I've dug up some evidence. And so starting by saying, here's what I found. So I'm actually telling you what what the conclusion is, essentially. And then I'm showing you some evidence and I'm saying, and this is why I think what I found is actually true. So it's because it's not always obvious that your evidence entirely ties into all of that. Right. That's closer to story, in my opinion, because there is
00:11:13
Speaker
a macro structure that is an overarching arc that ties the whole thing together. But if it's just, here's a piece of information, and then here's some more information, and then I just end, I just don't feel that that makes for anything that I would call story.
00:11:27
Speaker
So does the arc need people or units or individuals for the reader or the user to identify with or connect with? Or does it just simply need the arc? And if it doesn't need people or units, how do you build that arc? So for the arc, I think the arc can be independent of people, but it doesn't have to include people.
00:11:53
Speaker
Let's see, right now, I think that I love forcing you to work out your theory on the story during the whole interview. This is like the whole research process right here. Well, it's not like we don't know. We haven't figured this out. And I change my thinking every
00:12:20
Speaker
couple months. And I'm glad I'm not in politics because people would call me a
Challenges and Techniques in Data Storytelling
00:12:26
Speaker
football worker. But this is because I learn more and I understand more about what's going on and that changes how things work. And so anyways, what I'm trying to say is that
00:12:37
Speaker
Given that I haven't seen any clear counterexamples so far, I think that we can we can say that that story doesn't have to involve people. It can work without people. But for the most part, you want to obviously have something there that's not just numbers. So if you're talking about data, obviously, we we need to have a connection to what what the numbers mean in some kind of real terms, like how does this impact
00:13:03
Speaker
a person, a business, I don't know, something that's tangible that I can connect with. And for the most part, those would be people because we care about people, obviously. But if my presentation is about marketing, then I might not
00:13:19
Speaker
care. I don't have to make up a person like there's this whole idea in in, I guess it's also true in marketing, since I'm using this as my example right now, but also like in user research that you make up these personas. Yeah. And the personas become these drawmen, where it they become way too rich and and over embellished for no reason at all other than just because it's fun to make up a new person. But
00:13:46
Speaker
that they're actually more of a distraction than anything in many cases. So being clear about what you want and being clear about what the story is really about, like what should the story, what are you trying to achieve with the story? What is your, what are you trying to communicate? And then building the whole thing around that rather than trying to make it an amazing story that has like all kinds of conflict and people and you know, protagonists. I think that's actually more of a distraction than anything. And that's also where this term storytelling has kind of
00:14:16
Speaker
taking on a bit of its own dynamic that isn't always that helpful because we get sucked into these rabbit holes. That's fun to explore because there's so much with traditional story, but it's not necessarily all that helpful or
00:14:29
Speaker
for presenting findings and data and analysis, but it can make for a fun story. Sure, but it can be a distraction as well. But at least my next question, which is, do these conversations really matter? When I say to someone, I'm going to tell you a story with these data,
00:14:50
Speaker
In some ways, we've all sort of accepted this word story. We haven't well defined it, but we've all sort of accepted it. So when someone says, here's a graph and it's a line chart, and they're like, oh, the story I'm telling is this, we're sort of like nod our heads and say, okay, when I think you and I would probably agree that that thing itself is not a story because it doesn't have, you know, it doesn't have an arc doesn't have a conflict doesn't really have a resolution. So
00:15:14
Speaker
Like, is this purely an academic discussion or does it have implications for people and how they think about communicating their data?
00:15:24
Speaker
Yeah, I think it's both because the certainty is if you want to be specific, I think there's this whole thing about line charts in particular and anything that depicts time lends itself to telling a story. So it doesn't actually do it itself, but it does give you the material, right? So you can look at the line chart and you can see, oh, it went up during this time and so and then it dropped and then it went and then it increased again.
00:15:49
Speaker
And you can say, well, oh, I know by drop because this happened, that happened, and it was increasing because we did this and that. You can add a lot of information to that because time is so inherent in story that it just kind of makes you essentially tell the story. It just forces it out of you almost. And that's why it's both a bit of a pitfall. And it also kind of doesn't matter all that much because in the end, because we're so good at telling stories, I guess we mistake the
00:16:19
Speaker
the material for the actual story, like we mistake the numbers for the story in this case, which to me is annoying, but it might not be that important in kind of a practical sense. But from an academic point of view, I think that distinction is quite important because we need to understand what is the story and what is just the material that makes the story. But this story, in my opinion, is the pieces that actually tie it together, like the connective tissue between the steps that actually make the narrative.
00:16:48
Speaker
I know I like that distinction, but I also think, like you said, it's a little bit of both. But I think in
Targeting the Audience and Storytelling Techniques
00:16:54
Speaker
some sense, being able to define what we mean by telling a data story then forces us to be careful with that word. And so if we're just making a graph.
00:17:06
Speaker
You have this list of examples in one of the papers. For me, the first one that comes to mind is Snowfall from the New York Times. That one is still, for me, a real story. I would contrast that with a graph where I'm not telling a story. If someone said, here's this story I'm telling you, and they show me a column chart, I'd say, no, that's not a story. This thing is an example of a story. If you want to tell a story with the data, you need to think in a different way.
00:17:34
Speaker
And so in some ways, making people think more carefully about the word story forces them to think about who they're communicating with and what they're trying to say. So the story also needs to be targeted and tailored, I guess, to the audience. So you want to make sure that you have information that they need, that they have, and that they understand who the audience is and what they know, what they don't know.
00:18:04
Speaker
But your example also is interesting because Snilfel is this very elaborate piece that came out in 2012 or so. 2012, 2013, I think. I think 2012, yeah. And at that time, it was this very long scrolly piece that had like, what, five, six pages? Five, six sections, yeah. And each page was a long scrolling thing that had animations and the serviced video. And so it was very elaborate. And at the time, this was a totally new thing.
00:18:33
Speaker
But and it's clearly a story that there are, there's a lot of narrative in there. There's a lot of like, narrative about people and this term too. But the it tells the story of what happened and of the people and so on. So this was about this this avalanche that that when that happened in and the people got caught in somewhere here in actually close to where I am in Seattle. But but the problem is that comparison, of course, is that snow snowfall is this enormous
00:19:01
Speaker
huge piece. And so there's lots of steps in between that and a simple chart. So there are ways you can turn a simple chart into a story by just using some highlighting and stepping through a few things or maybe building up the chart by saying, well, here's the competition, here's us. And then maybe over time or
00:19:23
Speaker
given different segments or whatever. You can build things up that way and you can actually use a story using a single chart by just hiding parts. I do this a lot actually in Keynote, just putting stuff on top of charts, then revealing them over time. You can make really interesting little stories that way. You just have to play that a bit. That's what I'm saying though, that the chart, almost any chart, if it's not too complicated,
00:19:50
Speaker
can become the raw material for a simple story. And then you can just kind of work with that a bit and figure out, well, how do I start by not overwhelming people, by not giving people too much stuff at the same time, but focusing them on one thing and saying, look at this first and then look at that. And that way you set their expectations, you set their baselines, you can compare and so on. And that then turns the whole thing into an actual structured presentation and perhaps into a story.
00:20:19
Speaker
So does this require either animation or interactivity or presentation where you are revealing the pieces of the graph? And does that mean that to tell a story with a chart with data, you can't provide a single static graph? Well, you can have a single static graph and then a bunch of text spread. And so that's the way you can always tell the story in the text. Yeah, yeah.
00:20:46
Speaker
But if you really want to use the chart to tell the story, I think you need... I'm not sure I want to call it animation. Builds, for me, don't actually have to have any kind of transitions. You could just actually do it as PDF pages and just kind of flip through it. So it doesn't have to be smooth animation, that's what I mean. So it doesn't have to be...
00:21:10
Speaker
Motion it can you can just stick the simple thing is just like not show something and then show another thing You can very easily build those things. You don't need fancy animation. In fact animation probably actually gets more in the way than anything but So yeah for sure you can do that You can also Highlight things you could just do like cold nose bomber talks about this a lot in her book She actually does these she takes a chart and basically turns it all it gray and then she just highlights by using a
00:21:39
Speaker
uh heavier like a stronger color like a dark blue highlights the things that are important you can do that with lots of things you can make like some of the lines in the lines are heavier and darker and that way you end up just
00:21:52
Speaker
focusing on those, even if you have other stuff there. So it doesn't mean you have to hide things entirely. In fact, you want to show context, you want to show comparisons. But by being a bit more selective about what you show and exercising some judgment there, I think you can actually walk people through the data and the presentation pretty easily and pretty effectively. And if you do it right, I think the tools are really super simple. I don't think you need anything fancy at all.
00:22:14
Speaker
No, I think that's right where you have the graph with the one line and then another graph with two lines and maybe some text on it. I wonder why people still don't do that. I feel like it's the same reluctance to do small multiples where more graphs is in some ways worse than having one graph with everything on there.
00:22:38
Speaker
And I don't know if that's a holdover from, oh, well, it's printed. Thinking in a print world where we want to get everything on a page, but that's not the case when you're online. So there still seems to be, at least in my experience, a reluctance to take a graph or a complex graph or even a simple graph and break it up into multiple pieces. Yeah. And that I think is because there's still this mindset of analysis.
00:23:06
Speaker
You want to have almost like a table. You want all your information there. So if you give it to somebody, they can see everything and they can then look at it and find the patterns themselves. But if your goal is to walk them through that, then you want to be very focused on just the things you want to show at any one time. And then, of course, at the end, you want to give them as much as you can, like all the data, all the charts, whatever it is.
00:23:31
Speaker
As you're presenting, you have to actually hold back a lot of information so that they focus on just the thing you're trying to say. And that way you can get your message across. That doesn't mean you're hiding the other stuff, but you're certainly focusing on the things that you care about that
Importance of Conclusions in Storytelling
00:23:46
Speaker
you want to show.
00:23:46
Speaker
So in that framework, let's just say you write a blog post and you have a chart with four lines and you break it up into four separate charts. So you're going to show the first one, the second one, the third one, fourth one. When you get to the end, to have it be closer to this traditional definition of story, do you then need to tie it back to the beginning or because you've layered or built this thing up, is that enough? No, you do have to have some sort of connection back.
00:24:15
Speaker
you just described this, I just thought of an example that I just saw recently. There's a piece, I think NPR did this, it looked at when are people at work, like in the office. And there's this survey that did that, and it looked at like different job groups. And so it has this overall, this like area chart that shows the kind of the combination of all the different people, all the different jobs. And it's like, these are large categories, I forget what they're called.
00:24:39
Speaker
And then it shows that there's a distribution over the day. And so this is over one day, so from midnight to midnight. And you see how people go to work in the morning. And then there's a bit of a dip at lunch. And then it goes back up. And then it drops down. And then they walk through four different groups. So they look at, well, two of the groups that are specifically interesting, of course, are people who work in restaurants. I forget the actual terminology, like services industry, whatever.
00:25:05
Speaker
or something, right? Right. And so, of course, they work later, but they work when everybody else is at lunch because they're serving them. And then they work much later at night because you go to dinner, and that's when your servers work. And then the other thing is they have security personnel, and those folks are up, of course, during the night much more than anybody else, and so you can see that pretty easily. And so they walk through four of those. So they show you the overall thing, and then they show you four snapshots.
00:25:34
Speaker
And then that's it. And so that ends at this point. And so it's really unsatisfying because you're like, okay, so why isn't there something else here that you learned from that? It ties it all the way back to the overall average. That's where I feel it's not a story because it has a beginning, it has kind of a middle, but the end is just because it just ends. That's the end of the web page in this case. Why not have something there that actually ties this back and has
00:25:59
Speaker
some sort of historical context, perhaps, or something else that gives me a sense of, so what does this mean now? You showed me a few examples, but by showing me the overview first, I actually feel like you almost owe me some sort of conclusion at the end. You just walk me through a few slices of this data. So that's where I just feel like there's a missed opportunity there to make this into a story. And again, for journalism purposes, this might be fine because
00:26:25
Speaker
It's a classical very pyramid. You showed the overview at the beginning and the main thing. This is the shape of this curve. And then you go into more detail. And I could just stop reading it at any point because I might not care about the next one that you present. But that means that the end of the story is just really not defined. There's nothing there structurally. And that's where I'm just not feeling like it's a story. It doesn't earn the right to call itself a story if it doesn't have that ending.
Conclusion and Future Episodes
00:26:54
Speaker
So really what you're saying is the folks who are listening to this who are from NPR, they owe you a t-shirt. Oh, for sure. Yeah, at least. I feel like we could have this conversation all day, and I'm sure we will. You have a few more things coming up over the fall. I'm sure you'll be doing some writing and more talking about stories. So I'm sure we'll all look forward to that. So Robert, it's great to talk to you. Thanks for coming on the show. Absolutely.
00:27:19
Speaker
And thanks to everyone for tuning into this week's episode. I hope you're back on board listening to the show. I've got a great slate of guests coming up over the next few months. So thanks for tuning in. Be sure to get in touch. Let me know what you think of the show, if you have suggestions for guests. So thanks again for listening. So until next time, this has been the Policy Dis Podcast.