Archive for the ‘Visualisation’ category

OMG, the world just went crazy

March 26th, 2010

I’m still not quite sure this is really happening, maybe its the flu jab I just had…  I was tinkering with settings for this site, specifically mucking around with widgets and thought I’d try the widget that lets me add a Google Gadget. I thought, I’ll see if there is something out there to do with BI, so I searched for BI and came up with…(giggles)…

A “7D Pie Chart for Google Spreadsheet”

7d-pie-chart

This really has to be seen to be believed, so I have embedded it (admittedly right down, waaaay out of sight over there on the right, scroll down a lot). For one thing, I have no idea what 7D really means but just to give you some idea, the creators have combined the usual 3D stuff with more dimensions, size, colour, and (this is the best bit) sound! I mean it really is quite incredible, you can spin the thing around and click and it makes noises and shows stuff and you can stretch it and really go to town. But what on earth is it for?

What I find amazing is that the code necessary to achieve this must be really complicated, someone really, really smart must have sat down and worked really hard to develop this but I just don’t understand why.  It doesn’t help in terms of conveying the information and at best it is just distracting and ‘clever’.  It reminds me of my early programming days (here he goes again) when people did ‘smart’ stuff which the next programmer could never work out when a simple statement would have worked just fine.  At 3am when you get called in to fix a batch finance job you really don’t want to be unpicking ‘smart’ code.  So just keep it simple, keep the layout, the format, the colours and the content simple, just enough detail to communicate the story.

Stephen is just going to love this one ;-)

Dashboard Makeover – Pt III – Don’t Make the Reader Work

March 14th, 2010

If you go to all the trouble of presenting beautiful information on a dashboard but leave all of the interpretation up to the reader then you’ve missed an opportunity.  The information should leap out from the screen for the reader.  This doesn’t just mean highlighting the good or bad news, it means performing calculations and providing different views of the same data if that is what is needed.

Take the example of the annual time-series; it gives the whole view for this year and last (remember green is always current year and black is last year.  The peaks are shown on both lines and the overall picture can quickly be realised.

annual-overview-chart

However, as this dashboard is dynamic, it is updated daily, we could provide another view of the recent information that would save the reader having to hover their mouse over the datapoints to read off actual values:

7-day-zoom-chart

Note that the y-axis of this chart doesn’t start at zero.  That’s ok though, because we’re not comparing the magnitude of the values as we would in a bar chart, we’re comparing the actual different between the lines over the last week.  This tells a completely different story from the above chart – the measure has plateaued in the last few days.  However, we’re still expecting the user to compare the different values and then work out the difference between them, and in our case, work out the percentage difference between them.  If the lines were crossing over and changing quite dramatically relative to the prior year, this might be significant.  So we could again provide a different view of the data to meet this need.  This time we’re using a bar chart where the bar colour changes if the value is positive or negative.

percentage-variance-chart

In combination, all these charts tell a different story; we can see the overall picture, the local situation and trends in both instances.  We can view actual values or percentage variances without the need for any mental gymnastics.

Finally, a simple overview bar chart, provides an overall summary of where we are relative to the prior period.

zoomed-in-improved-bar-chart

Hopefully you’ll note in all of the above examples that the colours are muted and we’re toning down the non-data ink.  We’re not trying to dazzle the reader, we’re trying to communicate with them.  When we put the whole thing together it tells a more detailed story and hopefully the information leaps out.

improved-bar-chart-on-dashboard

Next time we’ll add a few more details to the layout and then put it all together as a finished product.

Eyes Wide Shut?

March 8th, 2010

Where have I been for the last twelve months? Somehow I’ve totally missed an incredible resource for creating data visualisations.  I’ve actually discovered two sites but more of the second one another day.  What I’m talking about today is Many Eyes which is a site that contains a bunch of tools (including good old Wordle) that let you visualise your own, or anyone elses data in a variety of ways.

many-eyes-logo

What I’m most interested in are the tools that enable visualisation of free form or unstructured data as we’ve seen before with Wordle.  There are however three more visualisation tools in the many eyes toolbox and I’ve tried them all out against Julia Gillards recent Address to the Universities Australia Annual Higher Education Conference.

First up we have a Tag Cloud, but a slightly different one from the simple version you see in this blog over on the right that shows the number of blog entries against a certain content category.  This one has a one-word and two-word version.  In the interests of space I’m just showing the two word version below but you get the idea.  In the real version you can hover over any word to find the actual word counts.

gillard-tc-1

Next up is my favourite, a thing called a Phrase Net which allows you to see the association between words and vary the connector.  Here’s the representation of the same speech with the top 44 phrases that have a space between them.  You can hover over the connecting line in the network diagram to learn more about the actual content that applies to that particular part – in the example below I’ve hovered over the link between low and SES. Its something that doesn’t come across very well in this static blog page but well worth a look at.

gillard-phrase-net-1

Finally, here is a Word Tree that lets you see the context around any given word or phrase used in the text.  I picked students in the following example which is found 25 times.  You can show this back-to-front also which is interesting and you can rank on various criteria.  You can also explore the text by clicking on particular words and launching independent Word Trees based on that word, then clicking the Back button and returning to the previous view.  The views change instantly so it is very easy to interact with.

gillard-word-tree-1

So that was just 3 different views of a single piece of text.  There are 20 different visualisation tools on the site, although only 4 that exist for the analysis of text.  I’m not quite clear how Wordle has ended up on here and the version that is here isn’t as flexible as the one you can still get on the original wordle web site – for instance there are only 2 fonts available on the Many Eyes version whereas I just counted 31 on the original Wordle site.

On the About Us section of Many Eyes they talk about democratisation of data visualisation, that sounds like a great idea, assuming we get democratisation of data sorted out first (I’m still working on that one here)

Going Public

March 5th, 2010

Updated 8th March (see below)

I see that the folks at Tableau are encouraging us to get competitive again.  This time round you can win a roundtrip ticket to San Francisco, 3 night’s accommodation at a Web 2.0 Expo Conference Hotel, free conference admission and $500.

tableaupubliclogo

What you have to do is use their new Tableau Public software to analyse and publish a visualisation of one of their datasets on your own website or blog.  This is a pretty major change in the landscape for data visualisations from a vendor because this allows you to embed real data in your own space that can be manipulated through an application at zero cost.   You can read more about the story behind Tableau Public.

I’ve read a few things about the service, mainly that the performance suffers a little bit when compared say to a Flash application, because your browser needs to go to the public server and get the results when you want to manipulate the data.  Having said that, I haven’t actually played with the facility myself so I should ignore the doom and gloom merchants and give it a go.

If anyone wants to create a university dataset of some kind then please do and I’ll be only too happy to put it up here.

Update: Here’s my very quick first effort at a visualisation – I couldn’t get it to embed in the blog page so will have to work on that a bit

Update #2 (8th March) – The folks at Tableau Support have been very helpful in diagnosing my problem.  It seems the version of WordPress MU that we have isn’t very happy with Tableau Public.  The code that is embedded is ‘changed’ when I preview the post.  If I put backticks (`) around the code then that stops but the published version still doesn’t look right.  So for now you’ll have to look at the examples on their website or have a go yourself!  Hopefully we’ll find a way around that soon.  Our wonderful blog admin is hoping to upgrade us to v3 in a month or two so that might be the answer.  The forum post on this is here if you’re interested in following it or have similar problems yourself.

More on Colour Defective Vision (CDV)

March 4th, 2010

This post is a follow-up to one I made last week – Don’t be blind to colour. You might have spotted a comment on this post from D Overton, a spokesperson for the Colour Blind Awareness and Support Group, Australia.  Mr Overton has quite literally enlightened me on several fronts and I encourage you to have a look at his website which includes a lot of interesting information and links to some fascinating resources.

Of particular interest to me was a link to some interactive java applets that let those with normal colour vision experience things from the perspective of someone with Colour Defective Vision or CDV.  Mr Overton points out that CDV is the terms we should be using and quite reasonably explains that Colour Blind is a term that implies the viewer “cannot see colour or is blind” which is of course not the case at all.

This extract from Mr Overton’s website hits home I think

Our sunrises and sunsets are not interesting; Our rainbows are plain; Our fruit is not ripe; Our meat is not cooked; Our  clothes do not go together; Our paint charts are a blur; Our printer/camera batteries are never low; Where are the flowers on the Christmas bush? We all have a box of 24 different coloured pencils we never use? Red and green dot clearance sales are great; Travel route coding and signs are a waste of time; We can not even see when we are sun burnt; I could go on, yet today, children and adults are still being told, when diagnosed with colour blindness, you will adapt and that’s it!

Dennis sent through a huge series of leaflets (24 I think) which took me a fair while to get through.  In amongst them was this list of suggestions and things to consider if you want your content to be more readily perceived and understand by everyone, including those with CDV.

  • Some colour-blind people may like bright colours because they can detect them more easily. Use blue, yellow, white and black
  • Use clear, recognisable and meaningful navigational clues on information to stop people turning off searches
  • Use textures or line shading instead of colour. Consider additional labels or the pattern function on the computer, especially for maps and charts
  • Text must be pleasant and easy to read.
  • Use small range of colours.  Associate colour choices with each message or piece of information.
  • Do not use washed or low intensity colours particularly on small bands, lines or text, as this will cause difficulty in colour discrimination.
  • Do not use red, green, brown, grey and purple next to / on top of / or changing to red, green, brown, grey and purple
  • Do not use light pastel colours in low light or bright light conditions.
  • Do not rely on colour alone to convey a message – give some other visual clue.
  • Do not use colours from the same part of the colour spectrum.
  • Beware some colours may be perceived to look differently when placed on top or behind some other colours.