Storytelling with Data

Subhasree Chatterjee at UXDX USA. Video: https://youtu.be/SxAKGVERkJs

Readable transcript: edited from the recording's captions for readability (fillers and false starts removed, punctuation and section headings added). Wording is the speaker's own. Timestamps are positions in the video. Names marked [?] could not be verified against the audio.

The shape of a data story

[00:00:05] Hello everybody, how's everybody doing today? Everybody well awake after Ken's great talk, and have enough coffee in all of your system? All right, let's get started.

[00:00:21] So we had some great talks yesterday and today as well, and can you tell me what all of those talks have in common? Yes, data and great storytelling. So my talk is going to be on storytelling with data, and I'm going to talk about some tips and tricks, how we all can become better at storytelling with data. And I'm going to try something out, and I'm going to attempt storytelling with data using storytelling with data, so that at the end of it we all become better at storytelling together. And also I am not just telling you what you all should do, but I'm also showing you what can be done and what good looks like.

[00:01:08] So what do all great storytelling have in common? It always starts with the beginning, it has a middle and it has an end. And always the plot starts with incidents, and then it recognizes the different characters, and then the story builds slowly, and after some incidents we reach a climax, and then we have some falling action, and then we have a resolution.

[00:01:36] But when you are talking about storytelling with data it looks a little bit different, so I had to create a graph of my own. So we will follow the similar narrative arc of beginning, middle and end, but in the beginning I am going to introduce the plot. I am going to talk about why it's important, why all of you should pay attention to me. And in the middle I'm going to talk about some examples and some do's and don'ts, where you can take practical tips to become better at storytelling yourself. And then hopefully, when everything comes together, we all have that aha moment when you realize how exactly it works. And then we will end with some key takeaways and a call to action.

Why storytelling with data matters

[00:02:20] So now we are here at the very beginning of storytelling. So first, why is it important? So data says that 63 percent of people remembered a story but only five percent remembered a statistic. That makes sense, right? We are all human beings and storytelling has been there forever. We have been using data and visualization to tell our stories, and some of these even remain now. Storytelling is engaging, it's educational, it increases your curiosity. It's also universal, because it can be translated into multiple different languages, and in terms of data and visualization you don't even need language in some situations. And also it's memorable.

[00:03:07] So that hopefully gives you enough idea of why it is important. But why does it have to be me talking about this particular topic? If you search in Google for storytelling with data you will get thousands and thousands of articles of what it means to be great at storytelling with data. There is a book called Storytelling with Data, and it is a great book by the way, you all should read it. So why am I here talking to all of you about storytelling with data?

[00:03:34] So I am a lead data analyst with LexisNexis, working for more than five years now. And as I have grown in my career, from an IC to now leading a team of data analysts, what I've realized is, even if you are great at doing the analysis itself and you have some great results with huge statistical significance, if you are not able to tell a good story of why that data matters to your stakeholders, that data will not be able to create the impact that it should create. So we all need to be better at storytelling with data, to be able to communicate with our stakeholders and to be able to create the impact that we can with data.

[00:04:24] So when I started looking at this particular topic, initially I was thinking that it's a me problem, that I kind of zone out in most presentations five minutes in and start scrolling through my phone. But then I realized it's not a me problem. 79% of people say that presentations are boring, and 46 percent have admitted, and admitted is the key word here, that they get distracted.

[00:04:52] So if we all are emotional human beings, whenever we communicate with each other within our group we use storytelling, then what is going wrong in this corporate setting, and why aren't we able to communicate in a similar manner when presenting our data or our work to our stakeholders?

[00:05:13] So I think there are three reasons. First, we don't always know who our audience is, or we don't pay enough attention to be able to speak their language. Second, we over-emphasize data. We want data to tell the story for us, but then data is only going to tell what the data says. It is your job to be able to share the insights, not only the data, with your stakeholder. And the last is not making an emotional connection. Again, we all are emotional human beings, so if you are not able to create that emotional connection with the people that you are speaking with, you are not able to create that impact that you can with including that emotion.

Knowing this audience

[00:05:55] So to be able to not repeat these mistakes when I am talking with all of you, I needed to make sure that I know who the audience is I am going to be speaking with. Now I know, of course, it's a UXDX USA conference, so there are going to be UX people here, but that doesn't give me enough information to know exactly who I am going to talk to. So that's why I reached out to Rory and Catherine, and I wanted to see what exactly the audience breakdown looks like.

[00:06:23] So as you can see in the left graph here, we do have a fair distribution of UX, product and development. But if you look at the right side of the chart, the majority of the people here are experienced practitioners and senior management. So that gives me a little bit better idea of who I am going to speak with. But then I also needed to understand your motivation behind why you are in this conference and why you are attending talks like these. So the majority of the people said that learning is a big part of it. You want to know about practical situations, practical case studies, which you can then go back and replicate in your own role. So that gives me a much better idea about what exactly you are expecting.

[00:07:11] So now we are here in the end of the beginning, where I have explained why it is important and why it is me here and why you should pay attention. Now let's get into the middle of it, and I am going to spend the majority of my time here explaining different elements of storytelling and how we can get better at it.

Element one: be the data detective

[00:07:33] So there are three elements of storytelling: data, narrative and visualizations. You start with data, then you build a narrative around it, and then you use proper visualizations that help you tell that particular story. So let's look at it individually. We will start with data.

[00:07:53] And the first thing that I want all of you to remember is, be the data detective. So whenever you have a problem in hand, first look for the evidence. What data support can you have to prove the problem you have and how big the problem is?

[00:08:13] So two pieces of advice here. First, look for data that rejects your hypothesis. How many times did you have a stakeholder come to you and ask you for data that exactly supports the hypothesis that they have? So this is a Dilbert meme kind of suggesting exactly that, and it was too funny for me to not share it here, because we all, I feel, can relate to this, and this is very real even though it's kind of sad in a way.

[00:08:45] So first, look for the data that rejects your hypothesis. And also make sure that you are doing the analysis right. So this particular graph talks about US spending on science and technology, and it highly correlates, as you can see the correlation is 99.79, with suicides by hanging. Now the correlation is real, it's a fact, but then that doesn't mean a causation, or that doesn't mean that one is impacting the other. And there is a great quote from John Tukey, talking about, if you torture data long enough it's going to give you exactly what you want. So we definitely do not want to do that, and we want to make sure that we are doing the analysis right.

Choosing the most impactful insights

[00:09:33] But then, once you have the data that you are looking for and you have done your analysis right, the next and the most important thing here is to be able to share the most impactful insights. And that's where I feel many of us go wrong, because when you are spending hours and hours solving a problem, talking with your stakeholder, trying to find a solution, you want to share every single thing that you have discovered through your analysis. And that's a very normal situation to have. But we will have to keep in mind that even if you have spent hours and hours working on that project, you are getting only 25 or 30 minutes to talk with your stakeholders and let them grasp every single thing that you have done.

[00:10:23] So keep in mind four of these factors to be able to get to the most impactful insights. And here is where actually our villain shows up as well, because we end up emphasizing more on the data, not thinking enough about our audience and not creating the emotional connection with them.

[00:10:43] So keep in mind four things. First, actionability. So the data that you are sharing, the insights you are sharing, what actions can the audience take based on everything that you are sharing right now? It has to be relevant to the audience as well. So if you are talking with your direct stakeholder versus you are talking with your CEO, the data that you are sharing is going to look very, very different for each of those stakeholders. So keep in mind who you are going to share your insights with.

[00:11:14] Next is the impact on users. So we always have too many problems to solve, right? There will always be limitations on resources, different challenges of technical feasibility. So you have to share the insights that are creating the most impact on your users.

[00:11:33] And last, arguably one of the most important ones, is alignment with the business goal. Again, when we are talking with our users, we are unrevealing some problems using our data, we tend to be very involved in there, and we want our product managers, our stakeholders, to be able to solve every single problem the user is having. But if it does not align with the overall business goal that the company has, there is a huge chance that that's not going to end up being implemented. So we need to keep in mind all of those things as well.

Element two: the narrative

[00:12:07] So now we have hopefully enough idea of what to do with the data part of it. Now let's look at the narrative. And here it actually goes really meta, because initially I have talked about following this narrative arc in my own talk, and you can keep in mind for most of your storytelling that you can follow this arc as well.

[00:12:28] But if you are thinking about sharing some insight with your stakeholder for them to take some actions on, you need to behave as the narrator, or like the advisor, and you need to also show them that you are the expert here. And in most situations the stakeholders do expect you to behave as the expert as well and share your recommendation. So start with that, start with your main recommendations that you have for that particular problem, and then back that up with the most impactful insights that you have found from your data analysis, and that's where all of those different factors come in in terms of selecting the most impactful insights. And then also tell them what could be. So if you implement a solution like this based on my recommendation, what is the potential impact that you are seeing if those things get implemented?

[00:13:29] So this is one of my slides where there is a lot of text going on, so bear here with me, but I wanted to share this particular one because I feel like we always go through these different stakeholders in our daily life where we are sharing our insights multiple times. So make sure that you are making that narrative personal, because as I have mentioned in the beginning as well, if you are sharing something with the CEO which is too technical, too detailed, then you are going to lose that interest very quickly.

[00:14:02] So for the first one, with senior management, keep it high level. Create an elevator pitch which you can communicate within minutes of your interaction with the senior management. And also keep in mind, three is the magic number. Three recommendations, three impacts, three data points generally create the necessary impact that you are trying to create. And always emphasize the business impact. They are always dealing with business metrics and business situations, so it helps if you can keep the overall business impact in mind and share it with them, how your recommendation aligns with those business impacts. And by format I mean here, basically I feel like we end up using presentations a lot more, and in every situation you might want to think about different formats of storytelling instead of just using presentations all throughout. But for senior management presentation mostly works well.

[00:15:04] Next you have your direct stakeholders. For direct stakeholders you have to start with the big picture and you have to start with the big recommendation, but there you have the opportunity to go into detail a little bit more, because they are probably working with you closely enough to be able to understand those details. But don't go too much into technical stuff. And for these particular stakeholders, interactive dashboards, Miro boards, tend to work better than just sharing presentations and sharing just the highlight of it.

[00:15:39] Peer group is the perfect place for you to be able to share every single detail that you want to share with them, all the technical stuff, all your challenges, all your learnings from that particular analysis. Here's your opportunity. And you can think about sharing something like a six-pager. It's an Amazon concept, I don't know how many of you are aware of it, but it's basically six pages of a Word document, or any other sort of document format, where you are sharing things in much more detail than just sharing highlights and sharing some data backing that up.

[00:16:17] For a mixed audience it's where it gets a little bit more difficult, because you need to keep in mind people from different backgrounds, from different roles, coming together and listening to you. So do not make any assumptions of prior knowledge, and also try to cater to different learning methods. And also keep in mind that all of them are there for a common reason. All of them are either trying to solve the problem, all of them are trying to understand the reason behind the problem. So find that and then try to work with them according to that.

[00:16:57] So now, as we have gone through these very text heavy slides, here's my way of rewarding all of you with this cute cat cave[?].

[00:17:09] So this is something that is very important while storytelling with data: keep in mind what they are going to do about it. Because if you are sharing some insights and you are not really calling to action, what they can do after they go out of the room, then again you are not able to create that impact that you possibly can. So always keep in mind what they can do about it after they leave the room.

Element three: visuals

[00:17:37] So that's all from the narrative aspect of it. Let's look at some of the visual examples. The first recommendation I have here is, use visuals to reinforce, not to distract people. And I am not even going to go into the details of these visuals, because that's the exact point I'm trying to make, that don't create a visualization which will just take 30 minutes of you explaining through this, and even then I don't think I can do a good job of explaining what's going on here other than it looks like a tornado. So let's not do that.

[00:18:12] And the last one is, let's not use pie charts. The only effective way of using a pie chart is this, where you are talking about how much pie you have eaten versus how much you have not. But even then, I can do a better job with a bar chart, because here I don't even know, have I eaten 25, is it 27? So I can possibly do a better job with a bar chart. So let's make sure that we are using effective visuals.

[00:18:42] So now we have made the connection of going from data to narrative to visuals. So here's just kind of a recap of all of those things, because I have gone through multiple different points, I have shared many different charts and visualizations, so just to make sure that you get the understanding of what exactly is going on, and for all of you maybe to take a screenshot right here.

Example one: How the Virus Won

[00:19:06] So now we are at this point where we have gone through the beginning and middle, but there is something missing, right? The aha moment. We now know the theoretical aspect of it, how you can use these different elements to be able to tell a better story, but we still don't know how it actually works. So here is my time to be able to share some of the great storytelling that I have found on the internet.

[00:19:33] The first one is How the Virus Won. So this is a New York Times article published in June 2020, so this is just around three or four months into COVID, and it talks about how the virus won at that standpoint. So we are going to look through this video, but just wanted to give you a context a little bit more. This will go very quickly, so don't try to read through every single thing that is there. The idea is to look at the visual and try to understand if we are going to the end of the story or not.

[00:20:09] So let's look quickly through it. It starts at the very beginning, when we identified two different cases in the US. Then it starts expanding, we start getting more and more cases throughout the US. And then there are hidden infections which are not detected, but they start forming these clusters across the US.

[00:20:37] And then we start seeing some travel bans, but then it was only from China, so different countries are still flying in, and you are seeing all of these different places flying in across the US. And then you start seeing cases spreading a little bit more, and you started noticing different outbreaks. But then we are still traveling, right, we haven't stopped yet. So you are seeing all of these different clusters of cases and how they are moving across the country and how the virus is spreading through them.

[00:21:36] Now it starts talking about different variants that we saw at that point of time and how much impact they had, so the Seattle variant compared to the New York variant. And then we start seeing some local hotspots, and now it starts making that emotional connection. Now it starts talking about the individual cases and how they ended up being the super spreader, and how, because they moved around within the country, they ended up spreading COVID in their community.

[00:22:12] Now it uses colors like green, yellow and red to showcase how things are changing and how things are slowly shutting down. So initially it was all green, now you start seeing yellow and red, and very quickly it all becomes red. And now it talks about, if we had done things sooner, how that would have avoided some of the infections and some of the deaths right there. And this is the situation as of June 2020.

[00:22:56] So I feel this particular story does a great job of starting with the data points that they had at that point in time, and then using visualization and using some of that crisp narrative when it was moving through the article. And it gives you a great idea of exactly what happened, how it happened, what we could have done differently to change it a little bit. And even if any of you are not aware of COVID for some reason, by the end of this article you have exactly an idea of what happened.

Example two: the climate clock

[00:23:30] So the next example I wanted to share is about this particular clock. It is actually right here in New York, and that's the address if you want to go after this conference and look at it yourself. But this is basically the climate clock, and it is showing the time frame that we have left until it is too late, and that's the quote it generally shows right after it shows the time, and the time is slowly ticking down.

[00:23:54] So this is a great example of using data to storytell and to inspire action, because when you see the clock right in front of you slowly ticking down, it is hopefully invoking all the emotions in you and asking you to do something about it. What can you do to slow this clock down? What can you do to hopefully stop this clock altogether? So this is a great example of inspiring action through data and through storytelling.

[00:24:27] So now we are here at the end, and I'm going to leave you with some key takeaways and a call to action. So these are the villains of our storytelling I initially mentioned. Now hopefully, after going through all of the different tips and tricks that I have mentioned, you can turn those villains into heroes. Now you know how to use your audience information. Now you know that you should not over-emphasize data and you should use narrative and visuals together to be able to tell great stories. And now you also know how to make great emotional connections. So this is my last thought and moral of the story right here: tell a story that inspires action. That's the end of my story, thank you so much.

Q&A

[00:25:19] Subhasree: I'll take any questions. Any questions? Okay.

[00:25:31] Subhasree: I'm turning to you. Okay, thanks for the assist.

[00:25:39] Audience: Hi, my question was related to the part where you kind of combine data as part of the story. So my question is, how do you ensure that when you're communicating something you remain clear but you're able to preserve that data accuracy and not fall into over-simplifying something just so that it's understandable?

[00:26:05] Subhasree: Yeah, so that's where you need to keep in mind the four factors that I talked about. It's not like you need to share only one data point, it's a combination of data points. But also think about the insights that you are getting from it, not just, this is what the data says and this is what we have seen the data is talking about. So keep in mind those four factors, hopefully that will give you the direction to be able to choose the right amount of data points. And then also take help of all of these different visual techniques that you have, so that you are able to tell a story that matters, and whether it follows a time frame or if you are trying to showcase something else, depending on how many data points that you have. But hopefully those techniques will help you select the right amount.

[00:26:53] Audience: Thank you.

[00:27:01] Audience: Hey everyone, I'm Luke, I work at a data visualization company. Thanks so much for the great talk. I was wondering, your point on starting with our recommendations as the expert: how can we kind of balance that without inciting bias in the audience?

[00:27:29] Subhasree: That's why initially I talked about, try to reject your hypothesis, because if you are trying to find the data that will support your case you are probably going to find that data somewhere, right? So we need to be very intentional about it and making sure that we are covering all of our bases, or we are looking at all the necessary places, so that we are not falling prey to some of those biases that we might have.

[00:27:56] Subhasree: Thank you. And we have a question right here.

[00:28:01] Subhasree: You got it. Nice, that was a good throw.

[00:28:09] Audience: I have a similar question, and it's about presenting data to senior leadership when the data you're presenting to them is not what they want to hear, it is not supporting their hypothesis. And in my experience, starting right off with your recommendation, that is, you know, what you wanted to hear is not what you're going to hear from me. I'm just wondering if you have any tips of how to maybe approach that differently when you know that your audience was expecting to hear A and you're coming back with B.

[00:28:36] Subhasree: Right. I mean, start with that, right? Like, this is not what you want to hear, but this is what the data says. And you have done your work to be able to be like, I tried your way as well, so I tried to see if I can find the data necessary to be able to tell your story, but this is not what I have found. And don't shoot the messenger, I cannot do anything about what the data is telling me and I am just here to share that message with you.

[00:29:10] Subhasree: But also, in situations where they are trying to find exactly what they're looking for, there are reasons behind it, right? So that's where I meant about making that emotional connection. So you just don't have that transactional relationship with your stakeholders where they are just expecting you to share the data, but build or develop that emotional connection, kind of try to understand where they are coming from, because they also probably have some goals that they need to meet, right? Like they need to develop something or release something because they are on a tight schedule. So try to understand and try to kind of work together to be like, okay, let's see, this is not what you wanted to see, but let's see what we can do about it, instead of just being like, well, it's not the one that you wanted.

[00:30:02] Subhasree: We can squeeze in one more quick question I think. I think you were first, so sorry.

[00:30:15] Audience: Thank you, I do like this thing, it's pretty cool. Thank you so much for this wonderful talk. I have a quick question. I love what you said about look for data that rejects your analysis. What I wanted to ask you about is, what about ambivalent data? So let's say you came with something and you don't have the really clear one aha moment, you have a bunch. How would you storytell that, and how would you inspire it for decisions?

[00:30:42] Subhasree: So always, like as I was talking about, aligning with business goals and with their roadmaps or their priorities. So start with the one that's closely aligned with that. And also, for situations where you are talking about ambivalent data, it always helps to look at the whole picture. So I always talk about, data doesn't necessarily mean just a set of quantitative data. You can talk to your users, of course, all of you know that. But you can look at customer support data, you can look at NPS data, you can look at exactly what they are doing in the product. So it's kind of combining all of these different data points, I feel, helps reduce that confusion a little bit and helps you tell that overall story a little bit better. So hopefully that helps.

[00:31:37] Audience: Awesome. Amazing talk, thank you so much.

[00:31:41] Subhasree: Thank you everybody.