This is a quick post was inspired by Emily Kund and her Vikki Vonn makeovers here, about why using the right colour palette matters and how its easy to miss the real story in the data.
Take a look at this chart that was recently featured in the Guardian. Its looking at the change of average house prices in the UK, split by region.
It shows that prior to the credit crunch house prices across the UK were steadily increasing, then in 2009 all the prices dipped and are now recovering. However there is one major flaw in this chart. The use of a 6 colour palette when there are actually 13 dimensions. This has the effect that the colours are used twice, and in one case 3 times. So you have London, East Mids and N Ireland all using the red line. So its no impossible to pick out which is which. The chart software they have used doesn't allow you to select the region name in the legend and then highlight the relevant line. This becomes a problem when the biggest question i wanted to ask was, "Wow, what region had that massive spike and why is that region declined where every other one has increased." I cannot answer that question at all. So if we download the data and drop it into Tableau we can produce this, using the same colour palette
Already this is an improvement as we can select the region name from the legend and it highlights the name in the chart, and now we can see that its N Ireland that had that large spike and then the rapid decline. Thats the real story in the data, the exception. We can go one further by changing the palette to the Tableau 20 put some information in the tool tip and now we get a chart that lets us actually find the story in the data.
Another option in this case would be to colour ever line the same except for N Ireland as that is where the real interest lies. All the other lines follow the same trend, its the outlier that we really want to look at
Hopefully this shows why its important to be careful when you select your colours. You are potentially missing out on showing the most interesting part of the story, and after all , that's the whole point of the viz in the first place.
A few days ago someone going by the name of Mr Elite, who gets a least a couple of points for an ironic name) posted a comment to a blog post on the Tableau Website. The post was "advertising" a viz design competition along the lines of the Iron Viz Championship that has now become established as one of the highlights of the Tableau customer conference. There have been a few of these competitions that have been run outside of Tableau, in particular one at Facebook and one organised by Emily Kund at her workplace. Mr Elites objection appears to be that by Tableau promoting these events it means that only those of Jedi status or greater need apply and that unless you are able to hack your way through advanced table calculations and the like then you are of no interest.
Now when i read that comment I smelt the faint whiff of troll, not in small part to the fact it was an anonymous comment which always seems to have a certain fragrance, there seems to be a lot of hostility in the language. My initial reaction was they they were talking a load of rubbish and after reflecting on it for a few days I haven't changed my mind.
A year or so ago Stephen Few wrote a blog post about the fact that Tableau had lost its way by including viz types that fall outside the accepted best practices types, word clouds, bubble charts and the like. In it he argued that the thing that made Tableau so good was that it guided you down the accepted path. This started a big, and rather heated debate with people pointing out cases when these non-standard chart types can be valid and reasons why sometimes you need you viz to have a little bit more bling to catch the eye. Mr Elites comments, though no where near the quality or reasonably argued as Stephens did make me think back to how the product and the user base has changed.
Tableau is now at version 8.1, its had many years of refinement and development. There have been numerous new features added, some major, some minor, some game changing, others not so much. There are some glaring omissions (would it kill them to put frigging autosave in? Its kinda standard thing now and you just assume its there in the background, just in case you need it Vote for it here). The software has matured, and therefore so has the userbase. You have people that have been using it since the very first version, many have been using it for years and so naturally their skills have improved. They know all the little tricks to get every ounce of the program and make it do things that no-one knew it could do. There are people whos design skills create beautiful dashboards that look a million miles away from the standard bar chart. But each and everyone of those people at one point was a newbie. They all sat down with a blank page and wondered what these buttons did. What a pill was, and why was it blue and green. Why this function didnt work how they thought it would and swore at the screen.
A superb blog post by Emily Kund (One of the best tableau blog posts of the last 12 months) talked about her own Identity Crisis in the Tableau world as she had reached a cross roads and was unsure where to go in order to improve her Tableau skills. This generated a fascinating debate about peoples skills and specialist areas within Tableau. One of the things its really highlighted though, and here's what Mr Elite should note, theres not a single person working with Tableau that knows everything. Some people are fantastic at advanced Table Calculations, Some hack the xml, some concentrate on design, others on performance. Even those in the Zen Masters, those most revered in the community are not experts at everything, but what they do all share is the willingness to help others.
i think Tableau is the most intuitive and easy to use pieces of software that i have come across. Every time i click a button or move a slider or select an option you can sense the fact that someone thought about how that action should behave. A novice, wil very little instruction can go from zero to a fully interactive dashboard in a matter of hours, just by feeling thier way around the interface and experimenting. And if they need help, well its everywhere. Tableau themselves provide a superb manual and hours of instructional videos, all for free. You can download any public dashboard, pick it apart and see how it was made (this is a superb learning tool). There is the forum, Facebook, twitter where the community hangout, ready and willing to help anyone, at any level with any problem, no matter how big or how small. You never feel dumb for asking the most simple question, why? cos everyone had that same question once.
Far from being a tool for the Elite, Tableau feels like the tool for everyone. Elissa Fink said in a recent interview that they want to put data discovery into the hands or the normal user, and not the preserve of the data scientist or analysis. I've seen that first hand at our work where people are able to interrogate their own data and get real insight, without every knowing anything about data structures, SQL, warehousing. What they do know is what the dimensions mean and what the significance of a 0.5% increase in a measure this month is to their department.
Going back to the Viz contests that are cropping up and whos existence sparked the Mr Elite comments. Instead of seeing this as elitist and only rewarding those that know how to hack the software or are at Yoda levels of competence they should be seen as a learning opportunity. Give the same data set, what do difference people come up with. What techniques do they use, how do they get around those little problems that still hold Tableau back. They serve to show you whats possible to achieve and that anyone can also produce work to that standard.
See them as what you can do, not what you have to do.
So i am sitting in the airport waiting for my flight and i am thinking back over my time at TCC13. To be fair a lot of it is a blur as a combination of so much content, full on networking, fire alarms, hanging with Chuck Hooper (THE Chuckfather) seems to have muddled my brain somewhat. It was great to meet so many of the Tableau family that i have known from the interwebs for some time and finally put a voice to the avatar.
One thing that sticks in my mind came from the very first day, it was a topic that came up in 4 separate talks and i think its the next big hurdle that we face. How do you get your consumer of your dashboards and vizii to interact with them.
The first hurdle that Tableau faced, and any dataviz maker for that matter was getting past the "But i want it in Excel" mentality. Well i think that particular battle as been won, or at the very least victory is around the corner. The average person is now used to data visualisations, they are everywhere now, some good, a lot bad. Infographics are the new cool way of spreading information in an eye catching way. People are comfortable with trusting the pictures rather than seeing the numbers behind them.
We are in an isolated bubble. We that create the dashboards, attend TCC, know and love interactive dashboards. We spend hours crafting them, adding parameters, creating actions that allow data discovery that their own pace. We know that, but our consumers often don't realise that these dashboards can be played with. Often all that time we spend making our work dynamic and interactive never actually gets used. Its like buying a Ferrari and just driving to the shops and back. Sure it works as intended but there is so much more power that is just being under used. I suspect that i am not the only one that hovers over every chart they see on a webpage, expecting and not always getting a pop up or dynamic action.
From my own work i have experienced this. i have created dashboards with cool actions, clever parameters that let you change the entire look of the workbook, but when i have happened to go to see people they haven't realised they could do that. Maybe we have a greater level of curiosity than our consumers, or maybe we need to put a little more into our work to make it irresistible for them not to play. I wonder how much of my time has been wasted adding features that were not needed. Now maybe thats my fault and i should just produce what's required, but I want to make the very best dashboards i can, so if i spend the time i want to make sure people make the most of it.
So what's the solution? i don't know, but i suspect its not a single solution, but here's a few thoughts.
Time - Most peoples exposure to charts have been in the printed media, or on a projector in a meeting rather than on their browser. As time goes on and they see more of it on their desktops it will seem normal to be able to interact.
New Tech - Kelly Martin (@VizCandy) suggested giving everyone iPads, the tactile nature of touch screens just invites you to touch the screen. For some its now second nature, my 2 year old will stroke my laptop screen expecting things to move about.
Tell them - A simple solution is to put instructions somewhere on the dashboard, let them know where to click, how to interact.
Trick them - I've never used hover action filters but this seems like a great use of them. Having a select action in order to interactive requires active participation. Having it as hover you are more likely to accidentally discover that "It moves"
Depending on the dashboard, the consumer and the actions in the dashboard some of these might not be possible, or practical. Assuming that our audience knows how to interact with our work, might be wasting some of our valuable time. Maybe we need to spend a little more and teach them how to the a curious 2 year old, and remember how to play.
