This is the fifth article in my Visualisation Insights series. The purpose of this series is to provide readers with unique insights into the field of visualisation from the different perspectives of those in the roles of designer, practitioner, academic, blogger, journalist and all sorts of other visual thinkers. My aim is to bring together these interviews to create a greater understanding and appreciation of the challenges, approaches and solutions emerging from these people – the visualisation world’s cast and crew.
Alan Smith is Head of Data Visualisation at the Data Visualisation Centre for the UK Office for National Statistics (ONS). He is a visualisation expert with an extensive portfolio of innovative data visualisation solutions and is a well known champion of the importance of graphics in data analysis.
My reasons for approaching Alan about conducting a Visualisation Insights interview are quite straightforward. His position as head of this important unit gives him unique influence and opportunity to effect best practice visualisation techniques across the critical hub of data and statistical analysis provided by the ONS.
Just to be clear the views Alan expresses in this interview are his own and not necessarily those of the ONS.
Please can you give me some background about the Data Visualisation Centre? How long as it been established? What motivated its creation?
We started the Centre in summer 2007. ONS has a central Methodology Directorate – comprised mainly of small teams with expertise in the various elements of statistical production (designing surveys, analysis, estimation etc) and the one one thing that seemed to be missing was a dedicated team for getting information back in to people’s heads – what Hans Rosling calls ‘the final 6cm’. For some years prior to this, I had been running a GIS and mapping team in ONS’ Geography division, where I already had a cartographic visualisation remit. But it became pretty clear that the distinction between mapping and other forms of visualisation was becoming increasingly blurred, both in terms of tools and principles, so establishing the Centre was a great step forward.
What would you describe as being its principle remit?
On a broad level, I would say it’s to make sure that, as a producer of an enormous range of statistics, ONS takes the communication of its statistics seriously. What we’ve found over the last few years is that there are a lot of people producing statistics in ONS, and across the wider Government Statistics Service (GSS), who really care about their outputs, but lacked any formal support on how to implement things they were thinking about. Our unofficial motto is ‘making numbers meaningful’ and that is a principal goal – to allow users of our statistics to bridge the gap between data and meaningful information.
What is the make-up of the organisation?
We are a small team, currently just 4 of us. Our backgrounds vary, which is what I wanted from the outset – a multi-disciplinary skill set is much more flexible than 4 clones would be. So we have myself with a cartographic background, one colleague has a PhD in Visual Perception, one is a psychology graduate with a background in data collection and finally, we have a GIS/spatial analysis graduate.
Who are the key clients?
Key clients, inevitably, are internal to ONS, though we are keen to offer as much help as we can across the wider GSS. and beyond. ONS delivers a broad range of outputs right across the economic and social spectrums. With many of our web-based visualisation projects, we’ve been keen to share them across the public sector for reuse (with other data sets etc). There’s been some enthusiasm for that too, which has been tremendous. Looking forward, the forthcoming 2011 Census will deliver an avalanche of rich data which we are looking forward to working with.
How would you describe the profile of its work (advice, research, design work)?
It’s a mixture of everything really – which is what keeps it fresh, For a small team, we have a lot of hats to wear, so yes, there is an advisory/consultancy role, training, we develop and maintain standards on data presentation and also get involved with design and production work. This includes programming and more traditional DTP-style work.
What is your own background in Data Visualisation?
I can really trace this back to my time at the University of Colorado in the early 1990s, where I studied cartography – using ink and vellum. At the time, GIS was beginning to gather momentum and I remember being simultaneously excited and frightened by what was going to happen to the cartographic field going forward from there.
How long have you been aware of Data Visualisation as a subject field? What was your ‘eureka’ moment?
One of my major complaints about early attempts to produce any kind of maps on the Internet was how clunky, awkward and ugly they tended to be – essentially because software vendors were trying to replicate full blown GIS on the web. I began to miss the control over design you get with pen and paper. While I was fishing around for a dissertation topic for my MSc (in GIS), it became clear that the solution was not going to come from the traditional GIS community, because, at the time, they just didn’t ‘get it’. I began searching further afield and ended up finding many more kindred spirits in the visualisation community rather than in straight-laced GIS.
Who would you describe as being the most influential (to you) practitioners/authors around visualisation?
There have been plenty. For me, Andreas Neumann was the first to demonstrate that you could deliver elegant, interactive data visualisations, combining maps, graphs and other visual forms, using nothing more than a web browser and open standards. His early work – such as the social patterns and structure of Vienna map – was tremendously influential. In the cartographic field, Jason Dykes, Mark Harrower and Danny Dorling spring to mind as people whose work make me sit up and think. I also admire Hans Rosling who has been very keen to push visualisation to support decision-making and public policy. I am always inclined to support the people who think of visualisation in these terms rather than as a pure ‘beautification’ exercise.
There’s a new wave of statistically literate journalists who are producing some great work – for example, Mark Easton, Ben Goldacre, Michael Blastland and Simon Rogers. If we can help people like them get to the information in our data quicker, then that’s a key objective of ours met. In my own team, Dr Steven Rogers, who has a phenomenal understanding of perceptual systems, is a continuing source of entertaining discussions and new ways of looking at familiar things.
More recently, I’ve also spent more time looking backwards than forwards. So people like William Playfair, who invented modern data graphics as we know them, Charles Booth, Willard Brinton, Jacques Bertin. It’s tremendous to look at what they did and think what they would make of where we are now (probably, simultaneous horror and excitement).
What learning resources (eg. websites, blogs, journals, books etc.) do you most commonly refer to or immerse yourself in?
There are plenty. Edward Tufte’s books are probably the ones you would keep on your coffee table if you had normal folk coming round for a coffee. Stephen Few’s books are the safety-first manuals we keep around next to the First Aid kit. A little known magazine outside of the statistical world is ‘ Significance’, now published jointly between the Royal Statistical Society and the American Statistical Association, has some lovely articles which I would seriously encourage people to read. I’m always coming across interesting web content, which I try to bookmark on my delicious.com account. The New York Times website is great, there’s some tremendous stuff there. For people involved in communicating about statistics, there’s the BlogAboutStats. My favourite visualisation blogs are information aesthetics – and yours of course 😉
Are there any software developments that particularly excite you?
I am particularly excited by the second-coming of SVG (Scalable Vector Graphics), which is now embedded in the nascent HTML5 specification. It is becoming exactly what I first hoped it would be – HTML for pictures. With proper support across most browsers, we should see some really exciting things coming from that area now.
In terms of software, web-based tools like MapShaper and IndieMapper are very exciting. Also, the profound influence that something like Google Maps has made on the public subconscious. Although there are many concerns over the degradation of traditional geography skills, you can’t argue that things like GoogleMaps and Location-Based Services in general haven’t made geographic information meaningful to people.
Regardless of the fight between Adobe and Apple, I am still a big admirer of both companies. They make great hardware and software that makes our jobs (usually) easier and more enjoyable.
Are there any particular trends in the field that particularly excite you?
The move to open standards for web visualisation. Even Microsoft are now on board. A web browser has now got to be the most ubiquitous piece of software in the world and they now all have built-in support for sophisticated visualisation techniques like animation and interactivity. It’s a real pity when web content fails to take real advantage of this. Free libraries like Processing.js should encourage some real innovation in visualisation too.
More broadly, the fact that people now expect the web to be engaging means there are real opportunities for bringing information to people who previously wouldn’t have have gone near official statistics. Visualisation gives us a chance to let people see official data in a more personal, meaningful way. They can visualise their own inflation patterns based on their own spending habits, compare their neighbourhoods and cities with others, see how long they are expected to live, all over a cup of coffee, using just a web browser. That’s an exciting thought and I think we are only just scratching the surface. The real power of all this convergence is going to be what happens when we bring different data sources together – as long as its done skillfully, rather than just for the sake of it.
Are there any matters that particularly disappoint or frustrate you?
Yes – but a lot of the things that generate frustration also mean we are never short of work! I get frustrated by all of these things at various times, but have to recognise we have had minor and major victories in those areas too.
What is your perception of the era of open data and transparent access – do you feel this is a positive move or is it somewhat superficial given we are essentially making data available to non-experts (in both analysis and domain knowledge)?
Overall, I think it’s a very positive step, though it’s not without issues and challenges. If I ever needed to demonstrate that numbers alone, without context, do not equal meaningful information, then the enormous amount of content dumped onto things like data.gov.uk is almost a self-authored case study. Just making data available, by itself, solves nothing.
However, I find the issue of non-expertise interesting and potentially patronising. Traditional notions of statistical literacy, based on numeracy, are ripe for challenging, I think. For example, here in the UK, Camelot had to withdraw a lottery scratchcard a couple of years ago (‘cool cash’) because it required people to compare negative numbers and too many people found that difficult. Poor numeracy was acting as a barrier to problem-solving. However, it would have been perfectly feasible to present the same information in a different way (for example, to ask them to identify a warmer or colder temperature, rather than a higher or lower number) and many of the same people would have been able to solve the problem. We are sometimes lazy in that we are happy to let an issue like poor numeracy act as a barrier when there are ways around it. I’m not saying we should ignore numeracy issues – but improving numeracy is something with a very, very large turning circle. 15 million adults in the UK lack Level 1 numeracy skills – but these are people – intelligent people – who need to handle data now, so there’s a challenge up front for people working within the data/information visualisation.
What role can the Data Visualisation Centre play in safeguarding the positives of this shift in attitude towards data access?
On a simple level, we’d like to see ourselves as just one place where people can come to for ideas on how to visualise data, recognising that there are many others too. I am particularly keen that we play a role in what at ONS we call ‘Wider Public Reporting’, which involves how we engage with people beyond traditional statistics users. There is definitely a need to unlock the expertise and insight that is acquired during the process of making official data. One option for this is syndication of one sort or another. I recognise that not many people are going to browse ‘statistics.gov.uk‘ over a sandwich at lunch (2 of their least favourite words in a URL!) – but they might look at the same content if it was hosted on the BBC, The Guardian, or The 10 O’clock News for example. The Met Office is a good example, I think, of a recognised authoritative source of official data, which is served to many different channels in many different forms for a variety of audiences.
Could you provide a brief outline of how a typical visualisation project may come about and evolve?
Normally, a spark – and buy-in – from the parts of ONS that produce the data is the starting point. How things proceed from there depends on what that spark is. I remember presenting the first animated population pyramid as a fait accompli to our demography area, mainly because we were playing with their data and wanted something less complex than a map to test out our ideas. But most projects end up being a collaboration based on their knowledge of their data and how it is used together with out understanding of how we can exploit it.
Which visualisation tools do you mainly use in the DVC? Are there any that you don’t currently use that you would like to?
Our general approach has always been to figure out what we wanted to achieve and then work out a way of implementing it – so we never let the tool determine the way forward. Having said that, we very often end up with the usual suspects – Adobe Illustrator, Photoshop, Flash, various open web technologies (HTML/SVG/CSS) because they are so flexible and between them cover most of our requirements. Excel is never far away either, as most data at one stage or other seems to pass through it. We tend to avoid tools that have proven difficult to deploy across the web, sometimes restricted by our own infrastructure or the ability of users to access it.
How would you describe your progress so far in promoting and succeeding in the better practice of visualisation/design?
It is a never ending job. But more people across official statistics now take notice, so there’s a partial success.
Do you have a particular design style/standard that you use for reference in handling visualisation project work?
Yes – to try and make the designs unobtrusive and place the visual emphasis firmly on data and information. We are gradually evolving a house style that reflects that. As with all things, we don’t want the design style to be a straight jacket, but a little consistency is a good thing for the end user.
How do you handle the issue of aesthetic design with functional performance of a visualisation?
Carefully. It is not a crime to make a data graphic look attractive and engaging – but if it is over-egged, it can actually be counter-productive to your overall goal. So we tend not to go for gratuitous visual effects (drop shadows, 3-d treatments, shimmering reflections) as these can lead you down a slippery slope. We’re also very careful in our use of colour – both from a web accessibility perspective and from an aesthetic angle. Another thing we have to be careful of in the statistical arena is natural colour associations with certain topics. Often, the simplest symbology is the best. For example, it is easier for the eye to estimate quantity from length than area, – so lines, not blobs, are often more effective if you want people to interpret quantity correctly.
How do you articulate the true benefits of data visualisation?
When people use visualisation as a route into data that would otherwise pass them by – so, catching a wider audience, or revealing insight that would otherwise be hidden. These are things that data visualisation brings – to experts and non-experts alike. Playfair said that ‘…no study is less alluring or more dry and tedious than statistics…unless the mind and imagination are set to work‘ – that’s the role of data visualisation. The information in the data is beautiful, not the graphic itself.
Do you ever come under particular pressure from clients/customers claiming to know what they want in terms of presentation of information? How do you handle this?
Yes – and many of them are right! The “ageing map of the UK” we produced was a reflection of a very well developed sketch from Shayla Goldring in our demography team, who had a clear idea of what she wanted to achieve. Having said that, there have also been occasions where we have had to step in and advise our clients of best practice and, if necessary, enforce it. But it is very rare for things to turn into a stand-off.
Would you be able to identify an example of a particularly effective visualisation project you have worked on?
In terms of sheer market penetration, the animated population pyramids, which have proliferated into many different variants, have been a favourite – and a good example of starting out with something basic and rapidly iterating it based on customer feedback. We have also had plenty of people from around the world reuse these templates with their own data, which has been very gratifying. On a personal level, satisfying my cartographic DNA, the flow-mapping in our CommuterView product was a lovely project to work on, I want to revisit that one in the near future with a modern toolkit.
Have you any examples of particularly creative or innovative approaches beyond the standard graphical approaches (eg. bar/line graphs) that you have recommended?
Boxplots. We’re big fans of boxplots for multiple distributions. They were devised by John Tukey, the father of Exploratory Data Analysis, who championed the used of graphics for spotting things you’re not expecting to see. There have been alot of recent innovations – technology’s been great for encouraging it – but my view is that there’s alot of mileage in applying new techniques to the standard displays – the best of both worlds!
How do you assess the success of your visualisation services?
Repeat demand! We do monitor feedback, both good and bad, and hope it will help us make better products in the future.
Finally, how would you like to see the Data Visualisation Centre evolve/progress over the next 3-5 years?
In the current climate, survival is success! Beyond that, I’d like to feel that success is making the Centre recognised as an integral part of ONS’ workflow, delivering core outputs, not just an ‘added value’ thing. If it can help draw attention to the wealth of information in ONS data, then so much the better.
*************************
I’m extremely grateful to Alan for his wonderfully candid and detailed responses to the many questions I posed him! He has perfectly encapsulated the purpose of these insights articles providing some rich perspectives from his unique role in this field. I wish him and his colleagues at the Data Visualisation Centre all the very best for the future. You can keep a track on Alan’s bookmarked discoveries via his graphboy del.ici.ous account.