Storytelling and Data Visualisation – interview with Amanda Cox

Whilst in New York City last week I had the pleasure of meeting Amanda Cox, Graphics Editor at the New York Times, to ask her some questions around storytelling and data visualisation. This was a busy week in news terms (the Boston bombing and Texas factory explosion occurring), and owing to both our circumstances we met late in the evening in the Wall Street area. After failing to find a coffee shop or bar open in the area we settled upon the steps of Federal Hall. I thought it was a fantastic and unusual location for an interview…




CT) There seems to be a lot of conversation around storytelling in visualising data. One of the things with storytelling that occurs to me is that its about order and coming to a point. Whereas in data visualisation, sometimes, there’s less order and it’s more disparate or incomplete. Do you think it’s easy to reconcile storytelling with visualising data?


AC) Yeah, I think that the definition of what story telling means in data is a little sketchy.  In that I think really what people mean a lot of the time is more the idea reducing your data until the point where it means something. Shaping it I think, or moulding the data, that can either be by choosing a form that reveals something particular about the structure, as opposed to a more generic form such as a bar chart. I think the idea of what it means to tell a story with data is a little bit amorphous, like its probably true for a lot of art, like in a painting, is there a story within a painting…is there a story in Jackson Pollock pieces or in Mona Lisa.

CT) Do you think it’s possible to ‘see’ a story?

AC) Yes I do think it’s possible. I think to define precisely what that story is is a little more difficult to pin down.  Even in simple line charts, so you think of like the canonical Al Gore example, even if you did the average and think of it as one line, I think there is a clear story in that. Who the characters are in the story may be missing from the line chart, but when there is an axis of time the stories are embedded in the time axis, so it can be constant, constant, constant – then something changes, that to me is very clearly tied to the idea of a story.

CT) I read a lot about narratology and story structure. David Herman, a narrative scholar, said that there are four key things that qualify a text to be as story… 1) a story has to have events that happen in time 2) there has to be a named individual or people who have to face decisions  3) there has to be a disruption of a state of equilibrium 4) there has to be a foregrounding of human experience. Do all four need to occur in data stories?

AC) I feel that the Al Gore example of time series fits that quite well. There is a dramatic change in events, it changes because of characters. So I might not disagree with that definition.

CT) Do you ever encounter data visualisation work that is not a cohesive story that, has a lot of uncertainty or indeterminacy in it so you have to work quite hard to figure out with the story is?

AC) I think things with a lot of data – the image I have is of Aaron Koblin’s Flight Paths, which I think is great and brilliant and will hold up 20 years from now which is rare in data visualisation, but to argue that there is a story in something like that, well there is in that planes take off, but its a weak story, it’s not that compelling that plane A took off, plane B took off. But at a deeper level the pattern is compelling in itself, but it’s not really a story.

CT) I see a lot of your work has a strong sense of cause and effect.

AC) I think part of our mission or mandate, working for a newspaper, is that it’s difficult and off course to make things that just look pretty. There is a deep question of ‘why are we doing this?’ and ‘what is it?’ – so that being our mandate shapes the work and makes it more simplistic. Questions like ‘why are you showing me this?’ or ‘what do I think a reader should get out of this?’ – it’s our job as journalists to do some of that work.

CT) How much work is it reasonable to expect a reader to do to ‘get’ the visualisation – should the barriers to entry quite high and if so what does the reader get out of it?

AC) It largely depends on what the data is and how easy it connects to people’s experiences. If you’re making a local census map, where the reader is asking ‘tell me about my neighbourhood’ you can shove all of the burden onto the reader, things that people are deeply interested in we would expect people should have enough background knowledge and context to be able to figure it out. On the other side there’s business stories or things that don’t fit into people’s experience, things that people can’t be expected to interpret on their own.

CT) Do you think that storytelling is a way into adding that context of ‘why this matters’?

AC) There is a sense of handholding in data visualisation.  We ask ‘who is going to use this?’ at sketching stages. I feel like a tour guide. Instead of dropping you into an area and telling you to find a bar, I’m giving you a list of the best 5 cool places!

CT) Does data visualisation have a rhetorical angle (intentionally or not).  I’m thinking you may be able to curate or visually preference data. Is that a good or a bad thing?

AC) All visualisations, even those without a story are an interpretation of something. The choices that you make, like Koblin showing a full day of flights all at once, people will think its crazy, but if you reduced the lines then you’d get a totally difference impression. Editorial decisions, about form, how much data you’re showing definitely suggest some interpretation.

CT) Your readership presumably comes to you for an editorial perspective?

AC) I think what we value at the NY Times is the analysis and finding the right experts to tell you what things mean. Not like the front page of the Newspaper which is very edited.  It’s more hands off, they just run the baseball scores.

CT) How do you think storytelling devices from non-fiction, like for example closure, narrator voice, having suspense, or characterisation could play with data visualisation?

AC) I think the chief device seems to be surprise. It seems to be a common thread in a lot of the good visualisation that we see. The most successful visualisations have an element of surprise, something happens that you didn’t expect.

CT) That would certainly make a piece more memorable! Do you think, like people have favourite films, books etc, that its possible to have a favourite data visualisation?

AC) I think it is. That’s an interesting question. I suppose my favourite films or books made me feel something which I think is the same with data visualisation if something resonated with me in some way.

CT) Do you think that data visualisation can tell a story on its own, or is it a catalyst for something else, or does it need other text?

AC) I think it could by itself, but not maybe in its entirety. I think we are pretty good in data visualisation at handling the what, where, when questions, but we’re bad at the why questions, which are often much more interesting. I think its probably the same in something like film too – is it really possible to get in a characters head? I think I would argue that on it’s own data visualisation is always telling a story, but it might not be a sophisticated story.

CT) That makes me think of Hans Rosling’s Gapminder, where the application itself gives you the what, where and when – but when he narrated it at the Ted talks it brought so much more of the why.

AC) I think that Hans Rosling is a brilliant presenter and very enthusiastic. I think that Gapminder on its own is fine but there’s nothing special in a scatter graph.

CT) It seems having an enthusiastic narrator can makes the data more interesting and understandable! So, moving away from story and onto you! How did you get into all of this?!!

AC) A happy accident! I was in statistics grad school not having a very good time. I started applying for random things. I did a summer internship at the Times. Then an opening came up when I finished grad school.

CT) Has the field changed a lot in your time?

AC) When I look at some of the work that the guys at Fortune were doing in the 1950s, it was incredible and is better than the work we are doing today. At its core I think it’s still the same, but when I started at the Times we did nothing on the web at all.  The opportunities are so different because of the changes in technology.

 CT) Where would you like your work to go in the future.

AC) I feel like there’s a lot of room for more actionable work. The highest compliment you get is that your work is ‘interesting’, but I’d like to make more of an impact on the world. I feel that the work can go further in its relevance. I want to be able to look back on my work today in 5 years and think its terrible!



A big thank-you to Amanda for taking time out of her schedule to meet me, and for being a wonderful sport.


Storytelling here and there

I came across 2 articles about storytelling on the same day this week, one by Jonathan Corum (1) related to the recent Tapestry conference (on Storytelling and Data Visualisation) and the other from Emma Coates (2) at Pixar, relating to 22 ‘rules’ of storytelling.

Though I can’t at all claim that comparing both articles is a fair thing to do, reading them adds to my hunch that story is thought of so differently in data visualisation than it is in most other places. And this is partly no surprise at all, because in data vis generally the idea of story with a clear beginning, middle and end, with characters and narrated voice and such is not the norm.

But what really got me thinking is that in data visualisation a lot more attention is put on visual methods for displaying data and their effects on users trying to ‘decode’ the visuals — than it is for say the ‘craft’ of telling a story and entertaining. Can I even expect to have favourite books, films AND visualisations?!  This is where it seems to me that data vis can learn from storytelling in other mediums. Sure most dataviz is not really intended for entertainment more so than it is for orientation and understanding complexity, but nevertheless people surely want to be entertained and engaged with a compelling story

I am exploring what relationships exist between data visualisation and storytelling in some current research. What I am finding is a curious kind of storytelling mode that happens through data visualisation: curious because such data-led, objective and fact based stories (not to mention the way narrative can emerge in them through interaction), are hard to reconcile with a traditional storytelling approach. Basic story elements like plot, character, narrator, suspense, conflict and closure can all and often be missing

Yet this doesn’t seem to make the work any less interesting and when I see a compelling visualisation it makes me wonder about cause and effect, and perhaps how people may have been involved or affected in the phenomena that are presented. A natural thing to do, to try and make sense of it by making it a complete story.

One impression is that a lot of Data Visualisations give the same effect of a vignette, more so than a full story.

Link to 2 storytelling articles:

Mirko Lorenz: from the practitioner perspective…

As some of you who’ve followed my blog may already know, I’m a part time PhD student at the London College of Communication. That’s what I do for some of my time, and the rest is taken up running the Graphic Design course and teaching here at the University of Lincoln. It’s been too quiet on this blog I’m aware, and long solitary periods sometimes need to happen when you’re wrestling with research, learning and keeping everything just pinned down. Anyway, to save extending such an apologetic opening, I want to share with you now some insight’s gained from talking to practitioners recently about their involvement in storytelling and Data Visualisation.

As my research goes further down the road of data visualisation design and what I call ‘invitational’ aspects of datavis, in the context of data journalism and storytelling, it’s been very enlightening to talk with several people about specific aspects of their work in this area.

This time I want to say a ginormous thanks to Mirko Lorenz for his insight and response to the following research interview questions. What he has to say here raises interesting points around telling stories through data, both from the theoretical and practical points of view.

I turn now to the unabridged text of this interview. Thanks again to Mirko and comments are welcomed.


Chris: Please describe your job position and role.

Mirko: I am a journalist, information architect and trainer. Started in print in 1985 as a free lancer, then moved to online in 1995. I write, develop concepts, including wireframes, conceive new software and manage the process to get them working.

For the last 17 years I am working on new ideas for journalism and media companies. For the last six years I am a member of the Innovation Team of Deutsche Welle. We participate in EU funded ICT research (e.g. semantics, cloud computing) and aim to extract new concepts that could help in the newsroom.

Since 2010 I am active in the space of data-driven journalism. I organized a conference in Amsterdam together with Liliana Bounegru from EJC, speak often at events about it, tweet and network.  With Nicolas Kayser-Bril and Gheoff McGee I wrote an article about a potential new position of media organizations in the future, which was published at and re-published by Nieman Lab.

Right now I am developing a curriculum, did trainings with journalists at organizations like Der Spiegel, Deutsche Welle, ABZV and recently for Mediacentar Sarajevo. One bigger current development is an open source data visualization tool called (with Nicolas Kayser-Bril and Gregor Aisch).

Chris: Please describe any challenges and opportunities data present to you in your daily work.

Mirko: Data, if understood, can add a layer of information that is deeper and more correct than other forms of journalism in certain cases and contexts. As a future perspective I would like to see techniques currently described as “Business Intelligence” into something new I would call “Public intelligence”: Media, institutions and individuals should be enabled to see and understand how an issue might affect them.
My favourite example here is the Rent or buy calculator, courtesy of the New York Times.

Challenges are primarily training – the technologies that can be used are there, much of it is open source and free. Another challenge is to avoid making mistakes that could turn the generally positive development of open data, open source into something harmful.

Chris: Is it important that your audience can interact with the data in your work? (this could be for example via inviting user input and manipulation, social commentary facilities, sharing source code and data). If yes, can you explain how you feel this is beneficial?

Mirko: This depends. The main goal of the data transformation is to move away from quantity to quality. The aim is to tell stories. This can in many cases be combined with data interaction, data download or crowdsourcing. But, there are other data stories where the audience or the user simply expects to “see” the picture. So, interaction and manipulation are additional options, but should be applied when appropriate and when they advance the story. A good example is what the Guardian did around the MP expense scandal in 2009.

Chris: In the context of communicating with data presentations, do you aim to capture and make use of possible audience/user contributions and interpretations? If yes, can you explain how and why? If not, is this because it is less applicable/appropriate to your communication intentions?

Mirko: Again, depends. Take Hackathons and other forms of one day creativity creation. If planned well, these can build new communities, new ideas, etc. If there is no concept, they just waste a lot of time.

So, user contributions are great, but they should advance the story. (I know, I am repeating myself).

Chris: Is part of your communication goal to help your audience to analyse data themselves (exploring), to provide a representation of the data for your audience (explaining), or both? If both, does this depend on any external factors such as source material and communication context?

Mirko: Both. There are three main categories of news from my perspective:

(1) Assumptions about the world: This is breaking news – people check to see if anything happened close to them. If not, they quickly forget, even with important issues. So here, explaining why a remote issue is relevant, should be a big goal.

(2) Opportunities. New jobs, new things, anything that can be viewed as a move forward. People around the world are searching through information to find something that benefits them. Here, exploration of data can add substantial new options for reporting and information offerings.

Of course, both forms can be mixed, depending on the subject

(3) The fate of the others: Although quality journalists, etc. don’t like this too much, there is a third category, consisting of people and home stories as well as gossip. This is essentially “yellow press”. We are interested in anecdotal reports about celebrities, etc. though, it’s quite human. The reason why (from my point of view) is that we are enabled to compare whether someone richer/poorer might lead a better/worse life.  Driven by social media this is actually gaining more attention. Put positively this can work as “social glue”, more negatively it is just diversion.

Chris: Is storytelling an important part of what you do? If yes, can you briefly explain how you see the role of data in storytelling?

Mirko: Story is key. To transform dry statistical material into something that provides new knowledge and might even change how people think, is the goal. Prime example is how Florence Nightingale reported about “Diagram of the causes of mortality of the army in the East” in 1854. This is a big story, on one sheet of paper. Transformed healthcare.

Chris: Do you anticipate that your audience will have little or lots of topical subject specialist knowledge and in what way might the form and scope of your data presentation be contingent on this?

Mirko: This is why story as structure and convention is so important. You can introduce almost anything to any audience, but to succeed something that is hidden must be transformed into scenes, that follow a certain structure (Beginning, middle, end). The simplification that is done by developing a story helps to make the content understandable.

Good recent examples are the reports about Olympic disciplines like swimming and 100 Meter Sprint by the New York Times.

Even further advanced are pieces about top athletes, again from the New York Times. Here, data is used to create scene after scene – after 1:30 you know more about baseball or tennis than ever before.

The stories below are built along classical narrative structure: A protagonist is introduced, then explored step by step. Both stories have a culmination point (e.g. when they freeze 1.200 throws of the baseball player in mid-air). Data and visualization techniques are use to provide new perspectives and insights. Watching both there is a certain pattern (e.g. when they compare the number of spins of the balls for both sports).

NYT: Mariano Rivera

NYT: Speed and Spin – – Nadals Lethal Forehand

Chris: Where you are communicating through data, how do you/your team typically make decisions about what and how much source data is necessary or appropriate to include within any single article/communication?

Mirko: Depends. Nightingale transformed healthcare with a small datasheet. Other instances, e.g. Arab Spring might have millions of data points. Again, not quantity, but quality and the potential story coming out of it are interesting.

Development is done by storyboarding, based on traditional questions like who, what, when and how….

Chris: In cases where data visualisations feature in your work, can you describe whether they are normally positioned along with accompanying text/images/visualisations or just used on their own? If either one or both, can you explain any decisions behind such placement?

Mirko: A picture as well as a graphic should always have one byline and tell a brief introduction story to enable the connection to the reader. These texts, if good, can be very, very short.

Chris: How do you evaluate or judge the success of the data presentation format that you create or use in your work?

Mirko: I would focus on getting quality feedback – like: I never thought about this like that. Quantity builds over time if you achieve this point of information and understanding.

Chris: Can you describe how practical limitations such as available time, skills and resources impact on your work?

Mirko: There are just too many offerings currently, plus too many areas of knowledge from code to software to techniques. So, one way is to enable both “quick data visualization” for some stories as well as really working a week or longer on bigger interactives.



Thanks once again to Mirko Lorenz for his insight and generous contribution to this research.