 Thank you very much. How do I change the slides? This one? Okay. Sorry, just thought I'd make sure. I know what I was doing. Kira, thank you very much for welcoming me. It's a bit of a different kind of talk than anything that I've presented before. So I'm going to be talking about data stories, which is what I've done for the past five years. But I'm also going to be talking about organizational change and how to not quite fail at it. Because as anyone would know, it's really hard. And I did not know before I started that it was going to be that hard. And I did have a dream job, which I left last month. After four years at the Herald and a couple of years at Fairfax. And I'm going to do this talk. I'm going to talk about data visualization as well. And I'm going to talk about how to fail at that as well. So this is essentially a talk about failure. Because I've found that that's not what people expect to do when they're trying to change stuff. But I think they should expect it. Because otherwise they will fail. So has anyone ever done something new in some place really, really old? When I first started at the Herald, they were celebrating their 150 years. My very first month there, I think in the first two weeks, they were really celebrating this thing of having spent 155 years doing that. And I was basically going in there and saying to the editorial team, you don't know what you're doing. I'm going to tell you. And it was a really interesting experience because newsrooms are very, very strange places. So newspapers are kind of like memory institutions, but not really because they forget what they wrote two days ago and they start all over again. So for the statisticians, they're like Markov chains. They only remember the last day and not really going forward. And I was working on projects which were a month old or sometimes would take two months or three months. And no one could really comprehend that length of planning. So I know that people in museums and other places plan for years. Newsrooms at best plan for weeks. And that is only during the election year. And I've got this map. This is a map that I did this year. It's a crime map. It's the map of all of New Zealand's crime data for a period of 33 months. How many people have actually seen or not seen the Insights website? Very few. Okay. That makes it slightly easier. So Insights is a platform that I helped build and launch with my colleague Caleb Tutti about two years ago. And we basically built it because the Harold website wasn't great. It wasn't responsive. And you had to upload one file at a time. And if something went wrong, you had to upload all the files again. And now it runs on AWS and there's all kind of, there's a continuous deployment system, but I'm not going to talk about that. If anyone doesn't know, most of the common assaults happen in that area of Willis Street in Cambridge Terrace. And the worst time to be there is on Saturday and Sunday morning. That's a heat map. The darker dots are like 44 people getting assaulted in the space of an hour. It's a very, very different kind of storytelling that I did the Harold, because newsrooms tend to focus on the absolute worst. So if I give the data to an editor, just tell me what the worst treat is. We'll put it on the front page. And then the police will get angry and it will just snowball into a thing. But I wanted to build data visualizations. I didn't really care what editors thought because one of the editors I worked for had said a few years before my joining that we're journalists, we don't do maths. And I thought, I don't really care what you think then. Because what I cared about was people who cared about this stuff. And I never wanted statisticians to be angry at me. I was fine with editors being angry at me. And I was always fine with politicians being angry at me. So my crowning achievement of this year was when Winston Peters called me an Asian immigrant journalist and sent out a press release against my data interpretation. And I'm proud of being all three things. And so I'm going to talk about data visualization, but I'm going to talk about how I introduced it in the newsroom as well. And how it almost didn't work. I mean, it didn't quite fail because there are some really cool people who are going to run the website next and who are going to do awesome stuff and who are already doing awesome stuff. So it failed because I couldn't actually change culture in the newsroom. I wanted them to adopt the culture of looking at data carefully and to make them understand that data is a fantastic way of understanding complexity which exists in our society. And I failed miserably at that because we built visualizations and the latest built visualizations. I spent about three months on this. No one bothered me. It was a lot of fun to build and people used it. But at the same time, the way the journalists look at the data and the way I look at the data is quite different. I'm going to talk about that failure of trying to do that as well. Pierre used to talk about shiny stuff. So one of the reasons it actually worked was because it was shiny. And in a literal sense as well. Because things moved on your screen. And if you ever want to persuade people to do data visualization, just move things on the screen. It makes people really happy. Even if they don't understand the visualization, they'll be like, wow, that's amazing. So I got to build all kinds of visualizations. I took these screenshots last night because I wanted to show that how much fun data visualization can be and why it's a worthwhile thing to invest in and learn. The first one is about migration patterns in the corner. And then it's a volcanic map of Auckland. What will happen in Auckland if a volcano is erupted? I always tried finding the sweet spot between being completely sensationalist but having really good data to back it up as well. So this was done by scientific research. The last one in the first row is the most interesting one which I did during my first year, which was about education patterns and how there was a disparity between different deciles in what they achieved internally and externally. The middle one, has anyone seen this? It's the history of all the old black tests. It starts from 1903, got the data from Wikipedia, scraped it and it tells you when they won. So in 1999, there was a patch of six tests which they lost. My editors loved this visualization. But the interesting thing to me was that this worked for a very, very general audience as well. The next one is a unitary planned one, which I didn't really want to build, but Chris McDowell, who many of you know, was building one for a spin-off and he gave me a heads-up that he was going to do it so I had to do it as well. But it's one of the more interesting visualizations because people to this day e-mail me and they actually use it. Auckland Council cannot say that, that their visualizations are actually used because they are horrible and no one can actually understand that. Because you know, like caring about audience, I'm not really sure why data visualization people don't talk about that. Everyone talks about doing fancy stuff, but no one actually talks about caring for the audience if the audience actually got the data, if they actually walked away with understanding. And that's like the most valuable thing that I learned working at the Herald was caring about the audience. The last one is the last visualization that I did, which was we got a University of Auckland student to build an election prediction model and it tried to predict what was going to happen and it was fairly close in predicting. As you can see what happened when Jacinta Arden got elected as a leader, the two lines started fighting. So this is a talk about failure. And I learned this the hard way that you're going to fail at doing something new repeatedly. So whenever I tried building visualizations, they just didn't work for an audience. And inside the newsroom, there wasn't enough of knowledge about what I was doing to actually have a feedback loop where someone could tell me, you did this well, you did this wrong. And I think it's quite important to admit that we don't actually have the expertise when we don't. I found it really hard in the newsroom where you had to explain things over and over again. And someone asked me recently why I left and I said basically I had my fifth boss in four years and when they asked me what data journalism was again, I was like, oh, it's time to go now. I can't really keep on explaining this every year. So it's also what I wish someone had told me before I started doing data visualization and before I started building it for an audience and before I decided that it's going to be great introducing data in the newsroom and everyone's going to be on board and everyone's going to be really excited. And there were early signs that that wasn't going to be the case but I chose to ignore them. So I think it might have started this way. Five years ago, two editors had a conversation that we should start doing data journalism now and the other guy said, absolutely, so excited. And what is it? It's not really accurate because I don't think they actually said what is it. They just went along and they're like, hey, we're going to do this. So I did an interview with the Herald and, sorry, I did an interview with the Herald and they, I sent in my resume and I listed R and JavaScript as two languages. I know this is four years ago and I didn't actually know them like that well. Like I kind of knew how to get my way around debugging them and copying and pasting code and I was learning as I was going along because I trained as a journalist, I didn't really train as a programmer and I thought that was like okay because I could actually use them and build things with them. I wasn't an expert in them but they were really impressed that I knew HTML and CSS in the interview and I was like, oh god. You don't actually even know what the programming languages are and this is the really interesting thing about change. You're doing some change in an organization and you're doing something new. No one actually knows what you're doing. So one of my friends joked that just have a terminal open on your screen all the time running random code and then you can just do anything you want. But I actually wanted to do stuff and so what I did was I optimized for learning even though my industry and my organization actually didn't. Journalism schools don't teach data visualization. They don't teach data journalism and this was one of the key things like if you actually want to change things you've got to learn stuff and you've got to optimize for it because when I first started learning R how many people know about R? For people who don't know it's a statistical programming language and if you're a Kiwi you really should know because it was invented in University of Auckland so it's like a local programming language which is the fifth most used one in the world at the moment according to Stack Overflow. When I first started learning R it was a horrible programming language. It still kind of is but they've made improvements and what's interesting to me is that for organizations and institutions which are not optimizing for learning and learning continuously things change almost every day nowadays. It's true for a lot of tools in data visualization because you work inside the browser and your capabilities are decided by the browser so if you're not really optimizing for learning as an individual or as a group of people in an organization you're not really going to succeed and that was the hardest thing to implement in a newsroom because they don't really care what you know. You could make a chart in Excel and put it there and call yourself a data journalist it would just be fine because no one actually knows but I didn't do that. This is my example of when no one knows what you actually do. I built this visualization for stuff when I was working in 2013. It's the visualization of the New Zealand budget. You click and things zoom in and zoom out. People were really impressed but I actually knew how I built it. It was horrible. It's got four iframes so if there are any like dev people each of the tab has four iframes I told you it was a talk about failure. I'm sharing my best failures and it's got four different iframes because I didn't actually know how to create tabs back then and I didn't know how to do data but the best thing about this is that I actually hand coded the JSON which has like a nested data structure while I was inside the parliament working on the budget on a deadline I wouldn't recommend it. It's not good for your health but what was interesting to me was that it worked, failures work as well because it was one of the first times that in a news organization at Fairfax they had a visualization which was like this which had this level of complexity and this level of detail. It didn't work for users it didn't work for me it worked for the editors because they thought it was great and I got employee of the month, that month. This is the first visualization I built for the Herald. In 2013 the new electorate boundaries were released and I... while someone suggested on Twitter that I should plot the boundaries on top of each other and then I should put all the polling booths on top of that so that we could actually see what was going on. Has anyone ever worked with the polling booth data in New Zealand? Yeah, it's horrible. Yeah, let's just say I'm glad someone understands my pain about making this but it's a horrible data set because it's great that we have it but at the same time I spent about a week cleaning it and then I built this visualization where you could filter for different electorates and you could zoom in and you could see stuff. And one of the first responses I got on Twitter was that this is the worst infographic I've ever seen. If you're a journalist and you're working in an environment where no one actually knows what you're doing Twitter is one of your best feedback loops and I just... like for a couple of minutes it just paused and I was like it's probably not the worst infographic ever made because I'm pretty sure that there are worse ones and what happened was that I replied to the person, I said could you tell me what's wrong and did you actually zoom in? And five minutes later they replied to me, they're like oh I'm really sorry I actually didn't zoom in I just looked at it and I thought it was really bad and because Harold hadn't published a lot of interactive maps at that point and then they deleted that tweet and to this day I regret that I didn't screenshot it and the same interactive visualization which was one of the first I published and was also called a superb NZ Harold graphic by Daily Blog How many people know what Daily Blog is? It's like the leftist sort of version of political commentary and anytime they say that the Harold's done something good you'd think that it's actually good because you know we don't really get praise from political blogs so this is what the visualization actually looked like when you zoomed in and you could see what had happened in the previous election so you could kind of see where the political parties had a strength and where things were changing you could make a visual inference which was what my purpose was for building this visualization I didn't think it actually worked again I got nominated for a Canon media award for this and it was really silly because I thought it was terrible because it didn't work for the most of the audience it worked for political gigs but that's not really your audience it might be the audience for politics reporters but I really wanted to get through to people who wouldn't otherwise engage with the data visualization so the fact that I missed this guy was quite important to me and it took me ages to figure out how to optimize for people who are not engaged in visualization on daily basis and how to actually create stuff for them so one of the interesting things about visualizations and the reason why I say that it's made up of failures because most people think that someone sits on a computer types in for your lines of code and a visualization just appears and it's just like polished and looks nice you should look at people's drafts and the number of things that get discarded is huge if you want to learn about data visualization you've got to learn about the fact that most things are not going to work and don't do bar charts and pie charts that's just like generic advice for that so I built this visualization and then I had the opportunity to work with that data at least twice and sometimes I optimized for different things because I wanted to get different things out of it in the first instance the screenshot on that side is the election visualization we did in the previous election and on this side is the one that was published this year it's quite interesting to me that one is just optimized to get most number of clicks because it's got all the polling booths for the country when the results came in it's not nice the colors are not nice the charts look horrible there's too much text there and we deployed it at 6.55 and the results came in at 7 and we were really excited when it actually got the results because we weren't sure that it was actually going to work but it was a spectacular kind of failure which worked in another way that visualization was the only reason that the Herald actually continued doing data visualization and hired more people so sometimes if you're trying to change things you just have to prove to people that you know what you're doing and what was interesting about that visualization was that three people volunteered their time because they really wanted to see Herald do data visualization and a month after this I got to hire a new team member and it became like a proper thing once you have two people it's a team and everyone expects you to create great stuff the visualization on this side as the one we published this year there's a real contrast to what's there Chris McDonald had done a hexagon map after the last election which you can find on his blog and the thing was that everyone was doing visualizations where they would plot the actual geographic boundaries and New Zealand looks really blue because all the rural worlds go to national what was interesting to me this time was that we wanted to do this and my colleague Chris Knox built this map which has amazing level of complexity even though it looks really simple and the editors didn't like it because it didn't look like New Zealand and we had to tell them that this is a better visualization and but they were like why is it better it doesn't look anything like Auckland looks like because if you actually look at Auckland most of the electorates are there because most of the population is there and you actually get a much accurate picture of what's going on but when I had that conversation where they were like we don't actually know why you would want to do this I realized I had failed at communicating what visualization meant because this was not like the first election after that this happened and if you are going to introduce something new it's important to learn yourself but it's equally as important to make other people competent critics in your organization because then they would actually know what you're doing is why what you're doing is not bad and I'd recommend Alan Smith who's the FDA's data visualization editor has a talk on YouTube called a competent critic and I'd recommend watching that and this was one of the most important things I learned after I left so I hope it's useful to other people I built the bar charts on the side and I really love those bar charts they were much more elegant than the previous ones and they fitted the names of people just fine and it's an iteration that just took me about three years because I didn't have the opportunity to visualize it again at different stages as I said I was optimizing for learning at different stages throughout if you work in a small community of people an organization you end up optimizing for what they optimize for and newspapers optimize for analytics they optimize for clicks and it just becomes a thing that you end up caring about as well and it's not like deep analytics where you actually understand engagement and stuff it's just like raw numbers 200,000 people clicked on this it was great and they also optimize for awards I don't know if many people know news and media awards you actually send in your entry you tell them why you're great and I didn't really want to do that but I knew that in the organization it really mattered and I didn't go to the first ones then I went to the second one and I got two awards that's actually a picture of me holding two awards but I cropped it and it was really a mistake because I didn't want to go because I didn't think that was important but it was important to the organization and just to the industry to a person walking on street and to a person interacting with my data visualization or the work I'm doing it doesn't really matter but it really mattered there and the reason why there were two mistakes was because the next year I tried to optimize for the wrong thing because I got into this journalistic mindset of trying to do pieces that will win awards the year after that they disestablished those two awards because no one else in the country was doing it so we were kind of winning by default and like the really important thing is to not kind of optimize for things just your organizations optimize for because they don't really matter in the real world they might matter like for a day or two or a week but in the end what matters is what you learn and what you create so it was a finite game how many people have heard of finite and infinite games too so there's a book by James The Cars and it's really my framework for thinking about change it was my framework for thinking about change right at the start and it worked really well until I started optimizing for the wrong thing a finite game is one where you play to win whether it's an award or your colleagues approval or you want to get a better job the interesting thing about finite games is that you can actually fail because if you are trying to win awards and if you are trying to convince your bosses that this is worthwhile then there is a much larger chance that it wouldn't actually work so during my last year I went back to what I was doing earlier because they disestablished the awards I can't win any more awards now so I'm just going to do data visualization but finite games are played to win and then there are infinite games where you don't actually play to win you just play to play and you play because you want the game to continue you don't really want the game to end you don't want it to be like okay I'm done here now I won this many awards or this much approval and now I'm going to do something else I just wanted to do data visualization I wanted to do interesting data visualizations and I wanted to do persuasive ones so the distinction that I would like to make in this case is that in the industry everyone was playing to win everyone was optimizing for awards there was no collaboration which fascinated me really because New Zealand's media industry is so small just to give you an idea of scale the New York Times has a graphics department of over 40 people we had a department of two and a half people in the best of the times and the people from different news organizations weren't willing to communicate or to actually learn from each other but there was a wider community of people doing data visualization in New Zealand and elsewhere who were more than willing to help to the extent that you would ask them a question on GitHub or Stack Overflow and they'd actually spend hours and post solutions and what fascinated me with that was that they weren't actually getting anything out of it they were just playing an infinite game because they were learning from other people and they were passing on that learning and there was a community there's a community of small community in New Zealand which you can mostly find on Twitter of people learning from each other and I realized that the second game the infinite game is actually a much framework to use if you actually want to bring change because in the first instance your incentives can be gone you won't win any more rewards and you would just be optimizing for the wrong thing and my best example if you actually want to learn data visualization is how many people know about D3? So D3 was created by Mike Bostock who did a PhD and D3 was his output and this is Mike Bostock's page where he posts examples anyone in the world who wants to learn visualization can just go to Mike Bostock's page click on a an example and copy and paste the code because the reason why his library succeeded was because he built so many examples that almost every problem was solved and people could actually be more creative Mike Bostock was playing an infinite game it was optimized just to create better tools for data visualization and I think that's a much better framework for changing things because I was trying to change things in a small newsroom whereas he ended up changing things everywhere in every newsroom because he was optimizing for a very different thing and lastly the best piece of advice that I can give you is that you never need permission to learn things never, you don't need to ask anyone and that's the best part of digital tools programming data visualization you can just go home go to Mike Bostock's page copy and paste code google some stuff and you can learn and that's me, thank you very much