0:00:06.4 Announcer: Welcome to the Analytics Power Hour. Analytics topics covered conversationally and sometimes with explicit language.

0:00:14.7 Tim Wilson : Hi everyone. Welcome to the Analytics Power Hour. This is episode number 260 and I want you to sit back, get comfortable by the fire, snuggle under a cozy blanket so I can tell you a tale. It starts back in ancient Mesopotamia over 4,000 years ago with the princess and priestess Enheduanna, who is often considered the first known author. For the purposes of this introduFction, we’ll say she is one of the first known and named storytellers while Enheduanna pinned her stories in Kannada form, the modern analyst uses digital technology, slides with words and images and data visualizations to craft data stories. And that’s the topic of this episode. What the heck are data stories? What are they not? Why do they matter? And what are some of their dos and don’ts? I’m joined for this particular podcast Narrative by Julie Hoyer from Further. Julie, what’s one of your daughter’s favorite stories at the moment?

0:01:19.8 Julie Hoyer: Hi there. She is really into If Animals Kissed Good night.

0:01:24.0 MK: Aw, that sounds really sweet.

0:01:26.9 TW: Aw, isn’t that cute?

0:01:28.6 JH: It’s really cute.

0:01:31.0 TW: And I’m also joined by Moe Kiss from Canva, a company that as stories go, actually is a unicorn. I think that’s right. Right?

0:01:40.4 MK: Sure.

0:01:40.7 TW: It’s your unicorn company. Yeah. But it also provides a platform that can help analysts deliver impactful data stories. Moe what’s a popular story with the Kiss Kids these days?

0:01:51.4 MK: Actually, we are really into AIs for analytics at the moment. It feels very fitting.

0:01:56.4 TW: Oh.

0:01:57.6 JH: Aw.

0:01:58.4 TW: Shout out to Jason Thompson and Hela. And I’m Tim Wilson from Facts and Feelings. I’m also the co-author of Analytics: The Right Way. A business leader’s guide to putting data to productive use, which is a non-fiction narrative available for pre-order now from Amazon, Barnes and Noble, Target and more. Apparently not in Australia, though. My kids are all well over a decade past relying on me to read stories to them, but I do have a nephew who will be getting a copy of Mo Willems’ Don’t Let the Pigeon Drive the Sleigh for Christmas. So we’re excited about that. But for today’s episode, we wanted to get someone who’s put a lot of thought into this topic. Duncan Clark is currently the CEO and Co-founder of Flourish, and he’s also the head of Europe at Canva, the latter of which is a position he took on when Flourish was acquired by Canva in 2023. Earlier in his career, Duncan was literally a storyteller in that he is a published author and among other storytelling roles, spent time as a data journalist at The Guardian. And today he is our guest. So welcome to the show, Duncan.

0:03:03.4 Duncan Clark: Thank you for having me. Great to be here.

0:03:04.8 TW: All right. So I think maybe a good place to kick things off is to actually nail down a good definition of data storytelling. Maybe, that may be the entire episode and we will get into, we’ll come to blows on it. So we’ll start with Duncan, if someone asks you to like, explain what data storytelling actually is, like what do you tell them?

0:03:27.4 DC: Well, I guess fundamentally data storytelling is about using data to communicate something. And that’s quite different from using data to understand something. It’s the difference you might say between what sometimes gets called in the data viz world, explore versus explain. If you’re explaining something, you’re communicating something, you are articulating an idea and in some sense, therefore you are telling a story. But beyond that, I think it’s one of those phrases that people do use in very different ways. I mean, there are people like John Burn-Murdoch who talk about storytelling being very much about how you use text in a chart and making sure a self-contained chart can articulate what it’s trying to say without supporting words. But there are, the way that we at flourish and before that kiln I’ve been thinking about data storytelling is really a little bit more like a traditional concept of narrative. Like a traditional story has a start, a middle and an end. It goes through an arc and so it progresses through time. And I guess what I’ve been working on for quite a long time is visualization that can do that, that can transition through time to actually tell a story With a start, a middle and an end.

0:04:38.8 MK: How much of it do you think is the, as you mentioned, the visualization and how much is it the narrative that goes with it? Or is it just like, it’s a bit of a dance and it really depends on the particular data story that you’re telling?

0:04:51.6 DC: I think it’s fundamentally about the narrative and the visualization is a really crucial part of how you tell the story, how you tell the narrative. It’s the reason that you can articulate a lot of information in a very succinct way. It’s how you can make something visually interesting. It’s how you can make something that doesn’t need you to justify every point you’re making. ‘Cause it’s justified in the visualization. But I think ultimately, if you’re trying to tell a story but you don’t have a message, then however good your visualizations are, what you’ve really made is something almost a bit more like a dashboard. It’s a collection of charts.

0:05:31.1 TW: So the communicate versus understand. I love that. And the going to the narrative. Would you then say that like if data storytelling is about communication and at the core of doing that communication, you need the narrative that really you should always be figuring out the narrative first and then the data visualization is just one piece that gets dropped in along the narrative as opposed to…

0:05:56.5 DC: Well. I would put it, but in a way it’s the other way around in as far as the narrative has to come from the data and how do you understand the data? Well, you do that visually. So it’s always a bit circular and a bit iterative. And I think data visualization often starts with let’s visualize something just to see what this data is. Okay, let’s change the visualization to understand it. And then once you’ve understood it and you’ve sort of picked it apart in different ways, it’s at that point where you start thinking, okay, I’ve actually understood what’s going on here. I need to be able to articulate that otherwise, ’cause you can do all the data analysis in the world, but unless you can explain why it’s relevant and get something changed as a result, then it’s an academic exercise.

0:06:39.6 DC: So for the data storytelling bit is that bit that comes at the end of that circle. Maybe there is no such thing as the end of a circle, but something that comes, you’ve got that slightly circular process of visualizing for understanding the common visualizing for articulation explanation. So the narrative layer has to come out of that. But it’s almost like if you’ve got the story clear, you can actually tell the story without the visualizations. It’s possible. Whereas if you just throw the visualizations that people, they’re not gonna understand what you are trying to get across. So it’s really a unity of them both that’s required.

0:07:16.4 TW: If you throw a visualization at them, you’re expecting them to then figure out the story. I think like that feels like the big miss. If I am trying to understand and I put all the understanding in front of you, it’s who’s taking on the burden of figuring out what it actually means.

0:07:34.3 DC: Exactly.

0:07:35.2 TW: Yeah.

0:07:36.1 DC: And you could almost see it as a spectrum, right? I mean, in theory you could just dump the raw data in front of them and of course, no one would expect them to be able to understand what happens.

0:07:43.0 TW: In theory, no that happens in practice. And it’s a problem. So.

0:07:48.1 DC: No, you’re totally right. I mean you see those Excel files that are circulated and people have drawn a cover page on them, almost like it’s a presentation and you put a big white box and put the title, and then you go to page two and it’s just loads of numbers. And so that’s the kind of dump the data and expect them to do not just the interpretation, but the sort of analysis. Then there’s the version where you’ve pulled out a few charts that make the data easier to digest, but you’ve still not explained what’s interesting about it. Then there’s the version where you get your charts to be sufficiently good at articulating their own message. And this is what I mean, go back to say John Burn-Murdoch from the Financial Times. He’s a brilliant data journalist. For him it’s always very much around how do you make a chart a self-contained piece of, almost like an encapsulated piece of content where the title is an absolutely key surface area where it’s kind of, this is explaining what the chart is saying.

0:08:47.9 DC: The annotation is then like the glue that binds the user’s attention between the title and the supporting evidence in the chart. And so that’s like the next level up where you’ve made charts that almost you could drop in front of people and they’ll get them. And that’s why, apart from him being a brilliant analyst, it’s why John Burn-Murdoch’s charts often go really viral on social because they tell a story in an encapsulated way. But then there’s the version above that where you actually sequence charts together and you construct a narrative. And actually to continue with the John Burn-Murdoch example, what he does brilliantly on social is he actually strings a bunch of charts together with tweets and tells a story. And it’s kind of, each chart is a self encapsulated piece of information design. And it’s a kind of scene in a story, but actually it’s when you string them together and draw a conclusion and tell the narrative that it becomes really, really powerful.

0:09:43.9 Speaker 1: It’s time to step away from the show for a quick word about Piwik PRO. Tim, tell us about it.

0:09:50.9 TW: Well, Piwik PRO has really exploded in popularity and keeps adding new functionality.

0:09:55.2 S1: They sure have, they’ve got an easy to use interface, a full set of features with capabilities like custom reports, enhanced e-commerce tracking and a customer data platform.

0:10:07.0 TW: We love running Piwik PRO’s free plan on the podcast website, but they also have a paid plan that adds scale and some additional features.

0:10:14.2 S1: Yeah, head over to Piwik.Pro and check them out for yourself. You can get started with their free plan. That’s Piwik.Pro. And now let’s get back to the show.

0:10:26.0 MK: I feel like we’re gonna get into this explain versus Explore concept a lot and we’re definitely gonna focus on the explain side. I’ve never heard it framed that way and I feel like I’ve had like a light bulb go off in my head because I have honestly, like Tim and I have both thought about this topic quite a lot to be honest, because it’s something we are really passionate about. When you think of the explore category though, like is it just like dashboards that comes to mind or analysis or are there like other areas that maybe I’m not considering that also fall under that.

0:11:01.9 DC: Well, I think it’s a really good question. So I think, the archetypal example of just explore, I think is the dashboard. You’ve got some filters, you’ve got a bunch of visual representations of the data and you sort of explore that way. But I do think there’s one in the middle actually. So one of the things, I mean, just to tell a bit of prehistory about where Flourish came from. So I was a data journalist at The Guardian and that obviously is very much, it’s all about the story. Like what are you trying to explain. How do you get the user to care about it? And coming out of that, I co-founded a little company called Kiln, which in the early days was just doing, it was really bespoke visualizations to order. It wasn’t a tool at that point.

0:11:45.0 DC: We were kind of experimenting with how you tell a story with interactive content. That was really what it came down to. And so to answer your question, like what we found ourselves doing is we would often make a chart. Let’s say you’ve got a scatter plot and you’re exploring the correlation between two things, but that scatter plot might also have a time slider. So the things are moving, hands rustling style over time, but you might also have a filter. And so in a way that’s a dashboard. It’s a chart with a bunch of controls. It’s very interactive. You can use it to explore that data set. Every data point’s available. You can move things through time. You’ve probably got animation as you change the filters, which will help understand the relationship between the two views. But what you would generally find, let’s say you’ve got 50 slider positions for the 50 years you’re looking over and then you’ve got four different categories and you’ve got four different color schemes or whatever.

0:12:34.6 DC: That becomes quite a rich powerful, it’s like a machine with lots of knobs. And the analogy that we used to use when we were working on this stuff was a play a piano. And you’ve got a piano where you can play all the notes, but you can also feed in a piece of paper and it will play the notes for you so that it’s a playable instrument, but it can also play itself. And that’s where we got to with Kiln is we would make things that were richly explorable, not exactly dashboards, more like interactive visualizations. But then we’d say the best way to actually understand why that’s even been worth doing is to pull out the most interesting views, string them together with a story. And actually then what you’ve got is a machine that plays itself. It’s like a Pianola.

0:13:21.7 DC: And I mean maybe just to give an example to just make that a little bit more concrete. ‘Cause that sounds a bit abstract maybe. So the first project we ever did, it was called Carbon Map. It’s still live at carbonmap.org. And it was sort of an experimental visualization that squished countries using a Cartogram system, but then you could also shade them. And what we did is we made this, we were like, God, there’s so many interesting things in here. What should we do? And we had this idea, well maybe we should make a video that explains how to use it. And when you go to the page, you play the video and then once you finish watching the video, you then go and interact with it. And then Robin, my co-founder had this idea, well hold on, why would we make a video? The video’s just showing different settings of the chart, right? We’re basically just recording ourselves pressing the buttons, so why not just actually make the thing play itself.

0:14:15.0 DC: So if you go to Carbon Map, there’s a big play button, you hit play and it plays some audio of me sounding younger sort of explaining what we’re looking at. But the idea is it’s the same experience as if someone had pulled it up on a screen and they’re saying, look at this, this is why it’s cool. Look at this bit, look at this bit. This is how it works. Sure. Now you can go and explore it, but I’ve explained it. And that was really the origin of a lot of the work that we did as Kiln. And then really that was the infrastructure that we built Flourish around was being able to have visualizations that are designed at the tool level for transitioning between different modes and capturing them in order to tell a story, but still let the user go and press those filters and explore the data themselves.

0:15:02.3 TW: That does feel, I mean, I can’t help but think of like the infamous, the late Hans Rosling’s 2006 birth rates, where he was presenting and that was telling a story. And if you look at like putting cool and it sounds like carbonmap.org. It’s very easy to look at those things that seem like so impactful and so powerful and they have such a crafted narrative and then say, yeah, yeah, yeah, but I’m just talking about marketing channels. So here’s bar chart, bar chart, bar chart slide. And I feel like there is this thing to say, yeah, but that’s the goal. Like always be, maybe it is okay to be aspire to that, what does it, shoot for the stars You’ll reach the moon. Like that’s a…

0:15:54.5 MK: Aim for the moon. You might reach the stars Tim.

0:15:58.8 TW: No, really? I thought it was…

0:16:00.5 DC: On the stars further away.

0:16:00.5 TW: Aim for the stars. You might reach the moon, right?

0:16:02.3 MK: I thought it was aim for the moon. You might end up in the stars.

0:16:05.0 TW: Oh, if you miss the moon…

0:16:07.2 JH: If you miss and you go further, I feel like the other way makes more sense.

0:16:09.7 MK: Okay, alright. I’m definitely wrong because I’m getting outnumbered here.

0:16:13.2 TW: But, well my intention was maybe I might be botching it, but it’s like, no, you’re not necessarily gonna have something that’s gonna wind up on a TED talk, but if that’s what you’re aiming for, as opposed to aiming for I must deliver the facts, I must deliver the information accurately and clearly, then it is a lot like harder.

0:16:36.5 DC: Yeah, totally. And so I think, I totally agree and like, realistically, you can’t expect that every piece of data information you convey is gonna be the equivalent of a New York Times story. Like, it’s not, that’s kind of might be the platinum standard, but it’s not gonna be day to day. But what we try to do with flourish is build a tool that kind of, that means that you’re not having to always make a trade off. Like you can just easily make a bar chart, but then you can make two bar charts next to each other. And when you move between them, they’ll animate between them. Or for example, I mean let’s give a real world example. Like let’s say that you’ve commissioned a survey, right? So you’re a marketer, you’ve commissioned a hundred people to give some views on your product.

0:17:19.9 DC: Now the conventional thing to do would be to share a deck that contains every different subset. Like well here are the female millennials in the dataset. Let’s have a bar chart for them. And you see this with bigger surveys, it’s often like literally 200 page PowerPoint or PDF with 200 static charts in. And apart from anything else that must be incredibly time consuming to make, but more importantly, the person receiving that information just doesn’t even look at it because it’s so messy. Whereas if you deliver a single, even if it’s just a bar chart with some filters, you deliver a single bar chart that lets you filter between the views, then you’re only making one thing. It’s easier to update, it’s easier to look at. You can share it on a single link. But then if you’ve made that, why not just quickly grab three different views of that? Well here’s what we thought was interesting, like the millennials look like this, the women look like this. However you’re gonna cut it.

0:18:16.9 DC: And then the third view is like, go crazy. Go exploit for yourself. And if the only extra time that adds is hitting the view and doing save slide and then doing it again, then it’s not really extra work. Be extra work to then incorporate into a really fancy website and to write a long form article or whatever. But that basic principle of, I’ve captured the views that are interesting, I’ve told a story and I’ve avoided a 200 page PDF with repetitive charts in and I’ve delivered one nicely interactive chart that animates between states, really clear what’s going on.

0:18:52.8 JH: You have to do some pre-filtering almost for your audience to your point. Like yeah, you can just take and make the 200 bar charts from the survey. But like as the analyst and the person making the report, I feel like part of the delivery expectations is that you come to them with the most important points. And I think that gets lost. Like either when you aren’t clear with your stakeholders of what they’re looking for. So you feel like you have to bring everything ’cause you’re like, I don’t have the context and the knowledge to pair this down to what’s most important to you. Or as an analyst, you just lack some of those skills to make that narrative to your point, and really be able to make the call of like, what do I feel like is helpful for them to go take an action or what’s most pertinent to like change their mind or support what they’re trying to achieve. And I think that’s a big, like leveling up and it’s a scary leap to make as an analyst. And we just, I think fall short a lot of times for lots of different reasons.

0:19:52.0 DC: Yeah. It’s a really good point. I guess the fundamental point you’re making there is like… Comes back to that concept of storytelling. There’s no point in delivering the 200 charts unless you’re explaining what’s interesting. But I guess just from a sort of format perspective I think what’s interesting is it sometimes feels like there’s a trade off. You’re either selling a dashboard link or you’re selling a PowerPoint. And for me the Explain bit which is the PowerPoint deck doesn’t actually need to be a separate piece of content from the Explore which is the dashboard. And actually it’s much neater to have everything encapsulated in a single view. It tells your story it lets you explore. So that won’t work for every data set and every approach but that’s where data storytelling can blur the distinction a bit between Explore and Explain.

0:20:44.5 MK: Duncan I feel like you’ve made reference to this a little bit. I have some intuition about the direction that you’re gonna take on this particular theory. And as Julie just mentioned I feel like especially young analysts think that all of the work is in the Explore and not in the Explain. How do you I guess get younger people or I don’t know less experienced people to realize the Explorer might be the fun sexy bit but the Explain is actually the most important bit and often the most time consuming. Because you have to put so much thought into that narrative and ensuring that people wanna take the best actions from it. Any tips on that?

0:21:33.0 DC: Yeah. Well it’s interesting. I mean I guess there’s a sort of communication skillset which… Because of… There’s probably a slight pre-filtering thing that the sort of people who tend to go into data analysts are probably more on the sciencey side a little bit less on the communication side. Which is not to say they can’t do it brilliantly but maybe it’s culturally in the data analyst community there’s probably less of an emphasis on communication. And communication is hard. I mean a lot of people struggle with it in all different disciplines. And I think to an extent it’s a muscle that you need to build. And if we are gonna have really effective data orgs we probably need to treat that as something that we train people in along with analytical skills. And I think there are some pretty basic principles. I think we just pull out a few. One of them would be if you were to boil it down to a sentence what is the point you’re trying to make?

0:22:31.1 DC: Literally just that forcing that concision if you could only say one sentence what is it gonna be? And it’s incredibly useful that kind of lens as a way to force yourself to work out if you have actually crystallized in your head what the message is. Because even a sort of 30 slide deck with all sorts of detail in it if there is an arc if there’s a connection if it’s worth presenting you generally can boil it down. And so forcing ourselves to think what does that boiled down sentence look like? I think another one is forcing yourself to ask the question why is this interesting to the audience? The thing that really great communicators do instinctively maybe but everyone else has to… The rest of us have to force ourselves to do it is understand what we’re saying and how it’s gonna look through the audience’s perspective. Why are they interested? Is this actually interesting to the people you’re talking about? And it is amazing. I mean I’ve been in lots of conversations over the years where I’m giving someone advice on a deck or whatever and I’m saying do you think they’re interested in this? And often the answer is yeah actually maybe not.

0:23:46.6 DC: And what’s interesting there is the person knows what is and isn’t interesting but somehow they haven’t got into a habit of forcing themselves to think is this additive or is this actually subtractive to my story? So I think that’s a really key part of it as well.

0:24:05.6 TW: Is there a higher order? I wind up talking a lot about reminding people of the curse of knowledge. Everyone has experienced having someone talk to them on something they don’t understand. And I think with pointing out to analysts that when they’ve spent years getting ramping up and understanding the data and then they’ve spent hours or days digging into something that what is now clear to them because they’ve sliced this data 100 times it’s really hard as a human to recognize that no the person that you’re delivering it to hasn’t been just focused on this dataset with the kind of maniacal obsession that you have been working on it and they don’t need to meet you where you are. You need to get back to where they can actually understand it. And that doesn’t mean you need to take them through the entire rigor. That’s not the way to do it. They don’t need to follow the same path that you did which is let’s dig into all the minutiae. You have to actually come to that simplifying point. So it seems like that why is it interesting to them?

0:25:24.9 TW: And then what do they need to understand and what is the clearest way for them to understand the minimal amount of understanding of what it’s showing so that their brain can focus on why it matters and what they should do. That is helping an analyst realize that it’s a completely different path than what the analyst took to get to their conclusion is what they need to do.

0:25:52.4 DC: Totally. I think that’s exactly right. And I think also if you are coming into a topic fairly cold like you’re the person having the story told to you it can be quite disorienting because there’s… Kind of like often there’s a bunch of different contexts that you’re trying to understand along the way and some of them are interesting some of them are less interesting some of them are more critical and you are working out as you go like hold on what’s this gonna mean? And then you get to the conclusion but you are a bit overloaded by that point. And that comes back to that thing I was saying about trying to express your point in the sentence. And I think that’s actually a really good practical tip is that if you add a slide at the beginning it’s like this is what I’m gonna tell you. And it is like if you can’t boil it down to a sentence it probably actually is a problem with your thinking rather than your storytelling.

0:26:37.2 DC: Because if you haven’t worked out what the point is then the single point not the four bullet points but the single point then you could probably do more work at that. And then that’s so helpful for the audience. Like you’ve said here it is. He’s like a TLDR but the shorter the better. And then when you’re Explaining stuff you’re referring back to the conclusion that you’ve delivered in advance. And that might not fit with traditional fiction storytelling where you are leaving people with cliffhangers and all that stuff. But it really helps if you’re just trying to articulate something factual that when you’re doing the explanation you understand what it’s building up to rather than leaving that to the end.

0:27:21.4 JH: You know what’s funny is I actually run into that when I’m reviewing other analysts work. They’re like oh you’re an analyst. Let me run you through this deck for a client you’ve never worked on for a question you’ve never had to think about in their context. And they’re like running through slides and they’re like and then this and then this and then I made this bar chart. And I’m like okay I have never seen this before. I was like I’m gonna need to take a step back. I was like but sometimes it’s funny because as painful it is for me at times I think it’s a good gut check to your point of nobody in this presentation will have been as in depth in this as you so none of this is trivial. You’re gonna have to slow it down and be really concise about what this means. So when you said that I was having flashbacks to some of those reviews that I do.

0:28:11.2 TW: I’m having a flashback I’ve had that debate about… And this is maybe I’m pandering to the guest the… Do you tell them…

0:28:20.0 MK: Never.

0:28:20.6 TW: Upfront what it is? And people are like no ’cause you wanna like bring them along on your journey. It’s like no the set it up which is different from executive summary of I’m just gonna barf more information out. I think making that point.

0:28:41.9 MK: I don’t think that’s what an executive summary is just to be clear.

0:28:45.8 TW: No that happens. That’s how they…

0:28:48.5 MK: Oh it totally is what happens.

0:28:49.4 TW: I think the executive summaries get manifested that way. That’s the intention is the executive wants a summary. What happens is let me cram everything in. The other piece that I think I’d love to hear what you think. I’ve used the… I can’t remember who I learned it from the idea of horizontal logic of… Horizontal logic and vertical logic. So when you’re talking about the data journalism and kind of having the title that matters more of a McKinsey title and vertical logic says you have that title and then the visual just does nothing but reinforce. Those are linked. You make a statement why it matters. And then horizontal logic is in a deck. If you just threw out everything and just read the titles of your slides does that actually hang together as a narrative? And it can’t perfectly work but it’s another kind of grounding to say is this a way to go from my one sentence summary to my now just verbally give you something that flows in a logical sequence and then the deeper story kind of falls?

0:30:07.5 DC: It reminds me so I co-authored a book called The Burning Question which was like a data-driven take on climate change. And it was trying to do sort of systems analysis for climate change around things like you shut down an aluminum smelter in one country but unless you’ve got a global carbon market open somewhere else and then they reimport the goods and so trying to think of it really holistically. And my co-author and I actually went through exactly that process you just described where we basically as we established the chapter structure and everything we wrote each chapter in a sentence then in a paragraph and then we wrote it properly. And we almost imagined… We never got around to it but we had this idea of putting the whole book on a website and there’s a slider at the top like how long have you got? And it would sort of just reduce down to the if you’ve only got 10 seconds…

0:30:55.7 S?: Oh no that would be cool.

0:30:56.1 DC: You can read 10 sentences.

0:30:57.7 TW: Oh wow.

0:30:57.8 DC: And then they expand out. But yeah I don’t think we ever got round that partly because probably the publisher didn’t want us just to put it online. But I think the idea is still quite interesting that idea of you can… It’s sort of collapsible like the summary is built in somehow.

0:31:12.9 TW: We’re having you back in five years and you’ll say well there was Kiln and then there was Flourish. And then I don’t know what the story compressor…

0:31:20.9 DC: Collapse book.

0:31:21.8 TW: Something the collapse book. Oh you had a name. See you.

0:31:25.3 DC: Well I do now.

0:31:28.8 MK: Can I just hone in on something that’s rolling around in my head I’m thinking about your comment of what’s the audience interested in? And I suppose it’s something that I’ve probably always… Yeah you’re always… The way I would frame it is what do they care about? What are they motivated by? How are they being measured or whatever the thing is to try and understand why they would care about what you’re sharing. I suppose the challenge is from a data team perspective you are sometimes trying to get people to care about something that maybe they’re not that interested in. And you use those other things as a connection point. Do you think it changes the data story you tell when you have pretty good intuition that they’re probably not interested and you need to find another way to hook them?

0:32:21.4 DC: Yeah. Totally. I mean because the challenge there the problem statement is there’s this important thing and they haven’t realized it’s important. So actually the story becomes around here’s why this is more important than you realized. But I think the way to do that often is to be quite explicit. It’s like rather than saying I’m gonna talk to you about this and then try and find some lateral way to make it relevant. It’s like if you think of that one sentence version it’s almost like the one sentence might be you think X is boring and you’re completely wrong or kind of why the only thing you need to think about this week is Y. ‘Cause actually the key point to your story isn’t any of the detailed conclusions. It’s the fact that there are tensions in the wrong place. So I think it still maybe can fit the same model. It’s just that you tell a slightly different version.

0:33:09.7 MK: I feel like Duncan’s about to start getting a bunch of decks from me and he’s gonna be like oh good God Moe please no. Yeah.

0:33:18.9 DC: We could build a AI deck parser and score your storytelling.

0:33:27.0 JH: There you go. How do you feel about whether a story has to be positive or not? And that comes from we have a lot of discussions at work. There’s kind of this pull of you don’t want to upset the stakeholder or the client. A lot of people are very worried of like well we can’t tell them bad news. Whereas then on the other side it’s like no our job is to be completely honest and transparent. So if something didn’t perform right we should tell them. Now I’m not a believer of that was shit. Obviously you don’t wanna go tell them that way and insult them. But it’s a weird rub of I think a lot of times analysts feel responsible for putting a good spin on it as if they’ll get blamed for a bad outcome. But I’m not a big believer of that ’cause I’m constantly thinking we’re supposed to be the neutral third party that helps give them helpful information like a recommended next action or a recommended like Hey that wasn’t your best. Let’s help you ideate on what to change. But that it’s also not our fault because we weren’t the ones that made the decisions that drove the outcome. But what are your thoughts on like the tone of communication?

0:34:41.2 DC: Well it’s interesting. I mean I guess it’s a similar sort of question to how do you give feedback in general. If you’re the manager of someone how do you give constructive feedback in a way that’s taken well and likely to lead to positive change rather than frustration, resentment and disengagement. And so maybe there’s some sort of just general questions there about how to communicate bad news or constructive suggestions in any kind of context. But I think certainly from my perspective the worst thing would be to dilute the message. Because ultimately the only reason we look at data is to work out based on the data what to do. So if we sugarcoat stuff it would almost be better to do nothing and just leave it to guesswork because it becomes untruthful really. So I totally agree that you need to see yourself as neutral to the conclusions.

0:35:39.0 DC: And I think in general I think that there’s a way to win the trust of your audience by being a little bit sort of emotionally attached to what they’re attached to. If they’re gonna be disappointed or almost be disappointed too and express that as a sort of here’s why we’re all disappointed about this. We were hoping for this. So it doesn’t come across as you are hoping for that. And I’m here to tell you that it didn’t work Lisa. It’s much better. We were all bought into this thing. Sadly it hasn’t fallen as we expected. And there’s probably some sort of fun humor that you can bring into that as well. But yeah I think the key thing is telling it as it is.

0:36:26.0 TW: Today’s data story is a tragedy.

0:36:28.8 DC: Exactly.

0:36:30.2 TW: Yeah. But there’s something to be if you think about it as the loss aversion, the banding together, if it’s this story what I’m trying to do is bring everybody together to rally people about overcoming an obstacle like the obstacle hasn’t been overcome so let’s treat this as chapter one of our data story. And I want to come back in two months or three months and tell the second version. I mean that’s getting very very kind of abstract. But I think if you think of it as a narrative why they care about it is like this is not good. But leading them to… But what can we do about it? Which I think also falls to the analyst sometimes. We’ll tell them how to fix it. It’s like well I’m not the fucking marketer. That’s not my… It didn’t work and I’m not gonna tell you that you’re an idiot. And I knew it was never gonna work. I mean that I can swallow but I can be like okay we tried this. We were excited about it and it didn’t work. So the story we don’t wanna tell in six months is we did the same thing again. So what could we do differently? How could we write our own story that ends more positively?

0:37:55.6 TW: Could be can be as I say could be I’ve had those discussions where you still want to grab them. They may not feel great overall ’cause they’re not getting the rallying cry of this was great. But if they walk out of it saying oh but I’m motivated because we’re in this together and we have some challenge to face.

0:38:16.8 DC: 100%. Yeah totally. And it’s interesting what you said about it being chapter one of a story ’cause of course the classic fiction story if you read something like Seven Basic Plots or whatever they almost all have… There’s some point when they’ve descended to the emotional valley bottom and then they come back up. And so I guess it’s sort of capturing that somehow that this is… Yeah it’s a chapter rather than the end of the discussion, the end of the story.

0:38:44.6 TW: You can be the hero.

0:38:44.8 DC: Yeah, exactly.

0:38:47.6 MK: I just wanna add to that. The one thing that Tim was saying, oh my God, look Tim, I’m about to give you a compliment. Look at this shit. The one thing that Tim kept saying over and over, which probably will go unnoticed, but I noticed it in a partner that we worked with who had to give us bad news. She used the word we constantly. Whenever she was presenting back results to us, she even though she was an agency that we had outsourced research to, it was always, we, it was never about the company that I was working for. And there was something so different, me and one of the brand marketers picked it up that she would always use the word we. And it made you really feel that you were in the results together, that you co-own the results. And it’s something that has always just stayed in my mind and I probably don’t reflect on enough, but it’s like, now hearing Tim say it, it’s stirred it up again.

0:39:42.5 JH: Yeah. That’s a good one.

0:39:43.7 DC: Totally. And that fits exactly with what I was saying about the… With being emotionally invested in the same things, being sort of co-owners of the bad news.

0:39:54.7 MK: So Duncan, I wanna give you a little bit of a scenario and completely put you on the spot. You’re a data scientist. You’ve had this really interesting business question. I don’t know, maybe you run an experiment or you did this deep dive, you have spent a good chunk of time on it. Your stakeholders know that you’ve been working on it, you’ve been going back and forth. You’ve got like this big deadline looming and you know you’re gonna be presenting back. And the data basically shows nothing. Like, I don’t know, there is no story. It’s just like, oh, it was a natural decline. Or like everyone is looking for kind of the silver bullet or this reason for something and you’re kind of like, oh, I don’t know. I feel like Tim would have a really good interjection here about, like, there would just be… There’s no news. Like there was nothing to find. You’ve looked at everything, you’ve cut the data every which way, but you still have this presentation date in the diary. How do you handle the data storytelling in this scenario?

0:41:01.2 DC: So it’s interesting. I mean if you think of the, going back to those sort of principles of like, sum up the key point and like know what the one sentence is and think what’s interesting to them. Like, it probably depends a bit on the level of trust you’ve got with that audience. But if your key point is there is nothing really here, we need to think from scratch about something else. If that’s the one sentence version, it’s actually we need to move on and talk about a new topic or a new way of thinking about things. And then you think what’s interesting to them is probably not you demonstrating that beyond a reasonable doubt. Like you need to sort of go over it to show that you’ve done the work, but you’re probably not gonna use the whole meeting for that. And it’s almost like you wanna boil that down to the one sentence is we didn’t find anything useful, a brief segment on here’s some evidence of that and then actually repurpose that meeting for like, what should we do next? Like everyone’s gathered together, what should we look at? What should we try? What should we think about? But like to changing the purpose of the meeting almost rather than thinking I need to tell a story that doesn’t have much content where everyone’s just gonna leave at the end feeling a bit bored but also a bit disappointed.

0:42:12.1 TW: Does it make sense to prep to sort of pre… I mean, you could walk into the meeting for sale baby shoes never worn. Okay, so now we’re gonna do this other, that is the sort of thing they can say, Hey, upfront, just know look thoroughly didn’t find much expecting this meeting, come with your thinking hats on. ‘Cause it’ll be very tempting to say, let me burn the whole meeting by just walking you through all different paths I pursued so that then we all feel like we’ve had that. If instead you can communicate, we looked, I’m a good analyst. I looked in all the ways. The dicey part is like, well, did you look at this? Did you look at that? It’s like, well, no, like you… Yeah.

0:43:00.2 MK: I feel triggered. I feel triggered. Well if you just run this query and cut the data this way and you’re like.

0:43:05.6 TW: Maybe you’d find something, then.

0:43:09.7 JH: But that’s a good question. In what form or like when do you not need a data story then?

0:43:17.8 DC: Well, it’s an interesting question. I mean, it also relates a bit back to what we were talking about formats. Like, in a way, the ideal thing is that you’ve got your story in your deck but it’s also actually got interactive charts in it, so if someone says, oh, you need to cut it that way you can just move the filters to show them that actually that doesn’t make much of a difference. And of course that’s not always gonna be possible. It might be a whole different way of analyzing the data, but that thing of having everything there so that your presentation, although it’s a narrative is also a bit explorable can also help with that kind of reactive analysis stuff. But I think the key point there is, again, boiling it down to one sentence, the one sentence is we’ve looked like properly. There’s nothing there. Let’s discuss, maybe you need three semicolons in that, but it’s a sentence worth of words and it’s about reorientating why we’re here, what we’re gonna do. And it’s less about the data itself.

0:44:17.9 TW: I mean that’s an, I could see the follow up to that. The next sentence saying, did it vary by channel? No. Did it vary by date? No. Did it vary by like something else that’s like, we looked at all of this, did we look at everything possible? No, because that’s not feasible. But if you, ’cause what would make me nervous about making it exploratory in the moment is that it then could become just like a group fishing expedition without a guide. And then.

0:44:44.7 DC: Yeah, true. Yeah.

0:44:46.9 TW: And then somebody does find something that actually doesn’t matter, but they finally found something. They’ve caught a little minnow and now everybody talks about the little minnow and the time is yeah. Ooh.

0:45:01.8 DC: But it’s also, I mean, you think of… I haven’t had this thought for decades, but I remember as a kid reading those sort of choose your own adventure books, where you get at the end of a chapter and it’s like, what do you wanna do this or that. You can imagine that kind of, yeah, you have the one sentence as you say, we looked at this, we looked at this, we looked at this, we looked at this, and then it’s like, slide six is what do you wanna do next? Like we can either think big about a different path or we can obsess about and dig in infinitely into like endless more cuts of this and probably get to the same conclusion.

0:45:35.9 TW: And still wind up on a desert island dying. I mean, you get back to the same, you’re like, oh, not page 42, come on. That’s where the story ends. So.

0:45:45.8 DC: But that kind of, again, it’s about sort of framing things properly. ‘Cause if you wanna steer people away from that, almost taking ownership of that possibility and presenting, there are actually two paths we can start afresh with something else or we can dig into this forever and probably not get anywhere. And people will be much less inclined to go down that path if you’ve acknowledged that it’s a thing that could happen. Explain why it is probably the wrong thing to do in advance. And maybe, I mean, maybe that’s leading them too much, but it could work.

0:46:19.1 MK: I actually like am having several light bulbs today of things that I think will fundamentally change the way I work. Like I think that is so great. Like it’s so obvious you’re like, oh sure, I should call out the elephant in the room and by addressing the elephant we can move past it or whatever. That is definitely top of mind. But the one thing that has been plaguing me since the show started interactions on graphs. So the other day I was working on something and there was an option for interactions on the graph live. And someone in the team was like, no, it should be static. Like a senior leader’s never gonna click on something. I do think it can be very company dependent and personality dependent, but do you think it’s something that you almost need to train in a company of like, we’re going to make interactive data visualizations and like it’s a norm that people come to expect? Or like, do you think it’s a cultural thing? Or is it just some people are driven to want to explore and others are not?

0:47:30.1 DC: It’s a good question. I mean, I think ideally the interaction is like an optional extra layer. And there are a couple of reasons for that. One of them is just a practical thing that the way that we give presentations often that doesn’t actually give us access to the mouse and keyboard. Like if you’re there with a clicker and you’re like, oh, I can’t tell this story properly because I’ve got no way of clicking on that dropdown, then that’s not gonna work very well. So it’s almost like you need the story to be self-explanatory as much as possible. But then the data interaction brings in that element of explore that allows people to either pull out particular numbers or to go down different paths in a bit more of a, to follow the same theme, choose your own adventure kind of way.

0:48:16.4 DC: So to give an example, if you land on a slide and it says, among our retail stores, blah, blah, blah, blah, blah, and there’s a chart and it’s got a filter on it that’s preselected to retail stores. So you don’t really need to do anything else. Like if they don’t interact with that chart, it’s not a problem. But if they look at the chart and think, I wonder how that relates to online stores and it’s kind of self guiding that you can just click on that and move it to online and you’ll see the bars move. So I think in a way that’s the ideal. I think there’s also sometimes a question for chart design around this stuff. Like, do you wanna try and squash a data label on every data point of a line chart or every bar of a bar chart when actually if you just hover over the bar, it’ll give you the value.

0:48:57.9 DC: And I think that’s another layer where it’s kind of, maybe you wanna put on the one that you kind of need to know for the story. Like maybe you highlight the highest bar with an individual value. But if most of the people aren’t gonna wanna know what the score was in Q3 2023, you don’t need to write it explicitly on you. Just let them hover over it and discover it for themselves. And I think the more people work with tools that support that, the more that will become quite a kind of obvious thing to do.

0:49:25.6 TW: All right. Well, this has been a fascinating discussion and I feel like we covered about 10%. We covered chapter one of a…

0:49:37.7 MK: Dude, it was chapter 1…

0:49:38.9 TW: 27, 27 chapter textbook. So Duncan, if in one sentence you could get… No that was, tell us everything we need to know, boil it down.

0:49:52.0 DC: I think this is the… If you did boil it down, it’s funny that one quote that I remember writing at the end of our flourish pitch deck was a Daniel Carman, the famous social psychologist. He had a sentence, no one ever made a decision because of a number. They need a story. And that in a way, if you sum it up in a sentence length piece of text, maybe that’s it.

0:50:12.5 TW: Oh, pour one out for the late Danny Kahneman.

0:50:16.8 MK: And we know that I love Danny. So it just fits.

0:50:20.0 TW: I mean we’ve had some great, great guests on the show, but I still think it goes down that Moe got a personalised rejection from Danny Kahneman about becoming, and that he did not come on the show, but he did respond and politely decline, which is something.

0:50:39.8 MK: It’s a precious memento.

0:50:39.9 TW: Yeah. So before we completely wrap up, we like to do a last call, go around and have everyone share something thought provoking, interesting, not necessarily related to the topic of the show. Duncan, you’re our guest. Would you like to be the first to share a last call?

0:51:03.5 DC: Sure. So I mean, I think a good principle for sort of data analysis and anything working with data is always to be sort of thinking from scratch. Like there’s so many places where if you think from first principles you might question a conclusion or whatever. So I’m really into any kind of media that pushes you to think from First principles. And one that I’ve been thinking about recently is Paul Graham, the y Combinator founder. He has a series of essays on his website, one of which has gone very viral recently, Founder Mode, which they’re great at getting you to think from scratch. I’ve also been reading Matt Levine’s amazing newsletter called Money Stuff, which again gets you to understand the finance world from first principles.

0:51:48.6 TW: Nice, excellent. Julie, what’s your last call?

0:51:52.7 JH: My last call is actually an app that maybe some of you have heard about, but Blinkist. I’ve recently started using it and I actually am a really big fan because I get a lot of recommendations or I hear a lot about career books, self-help books. And like as much as I love reading and like I’m interested in those topics, I personally really struggle to make it through a whole one of those books. I usually find that to me they feel a little repetitive and it’s a trait of mine to not wanna not finish something, but like I cannot finish a lot of those books. So the Blinkist is great, especially because there’s a listening option. It’s like 25 minutes per book and if I’m really hooked on it then I’m like, okay, maybe it’s worth me reading the whole thing. So I’ve really been a fan of the audio part of it and the summarization.

0:52:41.5 TW: That sounds like that last part is you feel like there may be a point where you actually are inspired to read the whole book, but you haven’t found one yet?

0:52:49.5 JH: Yeah, not yet. I’ll let you know if I find one.

0:52:52.1 MK: Julie, this is insane because mine is on a similar thread. I actually messaged my sister and was like, do you have Blinkist? Should I get it? So my rule of thumb is that now I only read for pleasure on my Kindle. If I’m going to listen, if I’m gonna consume a work related book, I do, I listen to it. But, so we all know that I love Choiceology and Katy Milkman, I am a very big fan. I did go back and re-listen to the episode recently on Choiceology’s Guide to Nudges, the episodes from 2022. But I just love the concept of nudging and behavioral economics and like how we can use that. Duncan, I’ve actually been thinking about it in Canva’s, like step two plan of how we can encourage the company to do more good. And I’ve been thinking about nudging and how we could incorporate that concept.

0:53:46.6 MK: Anyway, I’m very obsessed with it, but Richard Thaler is a guest and he wrote the book Nudge and I was like, oh, I’ve been meaning to read this. It’s been on my list for like five years. I should go read it. And that’s what prompted the discussion with my sister about Blinkist. But instead I went to ChatGPT and I said, summarise the book Nudge for me. And then it was pretty shit, I’m not gonna lie. And I said, summarise all of the empirical research in the book Nudge. And it gave me a summary of every single study day reference and the findings and the outcome. And I was like, that was amazing. So I have decided not to get Blinkist for now. I’m gonna use ChatGPT to figure this out and see how I go. But I’ll keep you posted. The problem is no audio version.

0:54:35.8 JH: Nice.

0:54:38.4 MK: Over to you, Tim.

0:54:40.5 JH: Yet.

0:54:42.2 TW: Oh, well I feel bad ’cause I’m not gonna stay in the behavioral economics vein, but that’s a pretty good episode if we’ve got Thaler and Kahneman both cropping up. My last call was not, I didn’t… I had it on my list and now I’m realising it is a little bit of a data story, but there’s a guy who does a website called Stat significant, Daniel Parris, but he did a while back, he did a piece, he basically goes into pop culture and just kind of digs in with data, this sort of thing that I think analysts sort of fantasise about just having a data set and just going deep just for fun. But it was quantifying the Kevin Bacon game, a statistical exploration of Hollywood’s most connected actors and it’s got network diagrams. So I was hooked, but it basically took the six degrees of Kevin Bacon and says, let’s try to empirically define like who the most connected actors are.

0:55:40.1 TW: It used it, like used eigenvector centrality to sort of figure it out. And it kind of landed on Samuel L. Jackson and I think partly because of the, between Pulp Fiction and Ironman two kind of promoted him way up there. But it’s one of those where he just finds little asides, he looks then most, what’s the most connected movie versus sort of the most connected actors. It totally surfaces the abysmal representation of like women over 40 in lead roles and how they don’t bubble up and kinda why, so there’s like social commentary and supported by data. So it’s a fun read. I wouldn’t say it would count as a data story so much as an interesting dataset, a curious premise, and then various little nuggets on the side. But again, it has a network diagram, so therefore I’m in.

[laughter]

0:56:35.2 DC: I’m gonna have to go and download the data and stick it into flourish. We love a network diagram on that.

0:56:40.8 TW: [laughter], We have a network facts and feelings, like a network diagram is like core to how we actually engage with clients. So, yeah. Oh, so Flourish does network diagrams, that’s a visualisation type within it. Is that…

0:56:51.8 DC: Yeah, we do all the visualisations. Well, I mean I’m not quite, but not far off. In fact, yeah, I should call out Flourish studio at the end given that I will be slapped in the wrist by our marketing lead if I failed to do so.

[laughter]

0:57:05.6 TW: She’s already, she had her finger hovering over as she’s listening to the episode and then she’s like, oh, he got it in. Okay. Delete the email.

[laughter]

0:57:16.2 TW: Awesome. Well, this was a really fun discussion, Duncan, thanks so much for coming on. I really think we could have talked for another two hours and still scratch the surface, but I actually in the style of Michael Helbling, which is not my normal style, I have a page of notes that I’ve actually taken during the discussion, so…

0:57:37.0 JH: Me too.

0:57:38.8 TW: That’s awesome.

0:57:40.9 DC: Well, that’s great Tim thanks for having me on. It’s been really fun.

0:57:42.9 TW: Awesome. No show would be complete without also thanking our producer, Josh Crowhurst, who pulls all this together and gets the ums and uhs like I just demonstrated, dropped out. So maybe that one will stay in but gets all this cleaned up, makes the podcast come out in a reasonably polished format. You, our listeners, we would love to hear from you. If you have questions about data comedies, data tragedies, data novels, data mysteries that you’d like to share, reach out to us on LinkedIn on the measure Slack. I’m not sure we’re gonna continue to mention that other platform anymore. But regardless of what kind of data story you’re telling, whether it has a solid plot, a weak plot, denouement, a surprise, a hero, a hero’s journey, whatever you do, keep analyzing.

0:58:41.1 Announcer: Thanks for listening. Let’s keep the conversation going with your comments, suggestions, and questions on Twitter at @AnalyticsHour on the web at analyticshour.io, our LinkedIn group and the Measure chat Slack group. Music for the podcast by Josh Crowhurst.

0:59:00.5 Charles Barkley: So smart guys want to fit in, so they made up a term called analytics. Analytics don’t work.

0:59:06.7 S1: Do the analytics. Say go for it, no matter who’s going for it. So if you and I were on the field, the analytics say go for it. It’s the stupidest, laziest, lamest thing I’ve ever heard for reasoning in competition.

0:59:20.6 MK: Can I give you my answer?

0:59:22.1 JH: Sometimes a lot of bad news.

0:59:26.2 MK: Psych is it [0:59:27.3] ____.

0:59:27.4 JH: Moe you go ahead.

0:59:28.3 MK: Do I have a lag?

0:59:29.8 JH: No, you let me go last time.

0:59:31.0 MK: Do I have a lag though? ‘Cause I feel like I keep interrupting everybody and it’s driving me mental. Sorry folks. Alright.

0:59:37.8 TW: You have a lag and Julie’s just a little too polite so that the two seconds between the combination of the two of you means when Julie starts talking is when you’re gonna catch up. [laughter] Julie, you wanna take this?

0:59:52.7 JH: No, no, no. Moe let me go last time. Moe you go.

0:59:54.0 TW: Okay. Rock flag and choose your own data adventure.

Podcast: Download | Embed

Subscribe: RSS