Sunday, 18 September 2016

Iron Viz III - Device Specific Dashboards

If you scroll back through my blog you will notice one missing item. Iron Viz Qualifier II. It was all about politics during a pretty dramatic time. The Presidential candidates were getting chosen in the US and the UK was in the midst of voting themselves out of the EU. Visualising data on politics was the last thing I wanted to do.

So why this enter this time? Quite simply, why do I enter any of the Iron Viz competitions?
Well it isn’t to get up on stage and viz my little heart out. It’s to actually investigate subjects and techniques I am interested in. But this time was different. I actually could develop an app that was useful for me and improve my chance of getting better at something I do a lot; namely Cycling.
I have been sitting on a data set of all my rides for the last two years but I didn’t just want to visualise them for a vanity project of “oh look how much I ride”. I wanted to save the data set for a time that it would actually teach me something and aid my improvement. Tableau’s release of device-specific dashboarding gave me exactly that opportunity.

The data set shows that I already capture the data from each ride manually, but now I can just add it to a google sheet and get instant feedback on whether I am riding as much this year than last, whether I visited some cool places and I shouldn’t forget, or whether I ride more if the weather is better etc.
Keeping a running total of the distance I do, whether it is inside on the turbo trainer or Spin Class or on training rides or tours, can give me an idea of whether I am improving and riding more distance with more confidence. The mobile dashboard can be easily checked to see this.

But what about my friends who don’t keep up with every ride? Well they can check out the normal dashboard that will give them a view on the foreign adventures and how I am getting on with my overall distance for the year (a little peer pressure goes along when it’s raining outside and the last thing you want to do is hit the roads).



Tableau techniques
I don’t often build individual callout numbers but in this dashboard they certainly had their places. I didn’t want to create multiple graphs with similar trend lines for different metrics. Time on the saddle and overall distance would always follow the same pattern so calling these overview numbers out was an easy design choice to make. To do this, just drop the value you want to show in to the middle of the visualisation (or on to the text part of the marks card). You can then edit the text (click on the text part of the marks card) to put the value in to a description to help position the number.

Running Total - In The Information Lab we obviously use Tableau to visualise our sales numbers and seeing them evolve overtime is useful. Comparing similar time periods against each other is a great way to show your progress so I decided to take a monthly look at the distance I rode and how it adds up. Giving myself lots of monthly targets rather than always trying to make a new personal best can be a lot more motivational and helps to break big targets down. The way Tableau handles dates is perfect for this so splitting out the months in to individual running totals is really easy as you can use separate date parts (day on ‘Columns’, months on ‘Detail’).


Shapes as filter – Rather than using a quick filter, Tableau actually performs better using a sheet with a dashboard action filter affecting the other sheets. This means that users need to be guided to interact with the worksheet that you want to act as the filter. A fun way to do this is to use custom shapes to make these filter sheets more interesting. You can load images and your own shapes in to Tableau by adding them to the ‘Shapes’ folder in your ‘My Tableau Repository’ (you’ll probably find it in your ‘My Documents’ folder if you are a Windows user). Rather than just having four icons for the seasons – I thought four different images of me cycling during the different seasons would illustrate this differently. 


Wednesday, 15 June 2016

#data16 - My Summary

Every year I mean to write this post. In the seven Tableau conferences I have been to, I think I hit this aim once. Let's make this two shall we? So here's a small collection of thoughts on a range of subjects whilst they are all 'fresh' in my head:

The Product

  • Tableau 10 is going to rock. 10,000 people are on the beta programme and I love everything that I have seen and tinkered with so far. Speaking to Francois Ajenstat this week, I admitted that the volume of 'newness' can be overwhelming at times but woah does it make the jobs we have exciting. 
  • The developments are needed. I love the tool but it is still quite new and has grown so rapidly so there are definitely things that need to change. People are still struggling with Blending and need more Enterprise level control to appease more of the stakeholders who are blinking demanding. v10 is delivering a lot of these requests. 
  • Some demanded developments aren't being delivered. There is a lot of demand for sunburst, sankeys and network diagrams. I'm still glad that non-best practice charts are being held back. I like to highlight that if you really want to build these you can but there are reasons why you shouldn't. This challenge will never go away.
The people
  • Great to chat to so many familiar faces. I couldn't walk anywhere for more than a minute before I'd stop and chat to some great people who I only get to see a few times a year. I love the community and the conference is a chat to talk to people in more than 140 characters
  • Meeting lots of people who I have trained and are loving the tool. Special shout out to Claudia who I taught in the Netherlands but so many more from clients and Tableau Public training sessions. It really motivates me to keep trying to make my training sessions better 
  • Putting a few of those Twitter profile photos to real faces. You know who you are!
The Tableau team
  • Absolutely cracking job by the conference team to run a conference for a 1,000 people. This is no small event so great job to everyone involved. You guys allow us to just enjoy and that speaks volume
  • Great to meet so many new Tabloids.
  • Bethany Lyons maintained the rule of 'Attend any session at a conference that Bethany presents'. Freaking amazing and thought provoking work that keeps me humble (even if you did offer a me Women's sized t-shirt)
The Information Lab team
  • I have so much fun working with this team of passionate gurus who clearly make such an impact on the community. I love what this team adds in and also how much we still have to learn.
So thank you all for making it a great couple of days and see you in Munich or Austin soon.

#data16 - Artilize This - Bethany Lyons

Balancing Art and Analysis in Tableau

Bethany takes some crazy challenges including creating Matt Miller's shoes in Tableau. Mission successful.



This session is not just about Matt's shoes, it's about the mix between quantative analysis and attractive design. 

if you have four measures to create a weighted score for a country-by-country weighting, then a bar chart is easy. But it's not effective. If two countries are the same weighting, it doesn't mean they score the same in each of the individual scores. 



Averaging loses a lot of the context - it's Bethany's Quartet now and no longer Anscombe's! 

Share price data is also true for this across a lot of stocks. A histogram of the daily rate of change is a nice way to look at this.




Stop fighting with the calc to form a certain type of chart and change the chart to make life easier.

Bethany shows an example of overlapping compaign dates and actually when talking through the way she fomulated the calculation logic, she broken it down bit-by-bit. A great lesson of how to solve the seemingly impossible / improbable to solve but do you need the level of precision the calculation is asking you to get to. 

No way I can blog the individual vizs without the data so hopefully Bethany will post these great examples / walkthroughs up! [Hint, hint]

#data 16 - Maria Konnikova - The Confidence Game

1951 in Canada and there is a huge shortage of doctors as no one wants to volunteer as they will be taken to be part of the Korean War. The Recruiting Officers were thrilled when an experience surgeon turned up out of the blue. The ship was on passage to Korea when they came across a ship of Korean injured soldiers. Ferdinand Demarra was actually the great 'imposter' and didn't even have a high school diploma - just a few text books on medicene. He actually undertook the surgery and somehow no one dies on the operating table. The Doctor Seer was actually in Canada and had his identity stolen when at a Monestary from a Brother Monk.


The real Dr Seer gets prosecuted! Demarra actually gets discharged with honours. He actually carried on practising medicene into the 1980s and nearly built a bridge in Mexico. He had an biographer who wrote a book on him but then the biographer had his ID stolen by Demarra. 

So what is it that makes us give our confidence (trust) to others?

Maria thought that sceptics wouldn't be caught out. The research proved that as a species we are pretty bad at working out if we are spotting lies. As most people are not out to get you, you get an implicit trust if you are not looking at them. Every single day we lie and we actually accept it as pure truth telling with be pretty miserable. "How are you?"... "Well let me tell you..." Uh oh! People with higher IQs actually trust peoople more. Stronger social connections often lead to longer lives. We have not evolved to spot them. Maria basically wanted to lock herself away as questioning the truth telling lead her to be more sceptical about society and all those around her. 

How do you take in data from the world around you and judge yourselves?
The 'above average effect', the 'exceptionalism bias' People think they are naturally above average on desirable traits. Being bad at 'understanding people' is not something we expect.

Al Capone - one fine day Victor Lustig (fine conartist who sold the Eiffel Tower - twice) Al Capone met with Lustig even though Capone knew his background and 'skills'. Lustig gets $20k to double Capone's money. Lustig just used a safety deposit box to hold the money but just gave the $20k back to Capone and said his ideas didn't hold out. Capone expected to either 1. Double or 2. Lose it all. Capone actually just gave him $6k to get him out of his financial difficulties. 

To avoid a con 'if it seems to be good to be true, it probably is'. But it's very easy to tell this if you are judging others. Not so easy when we assume we are exceptional ourselves, so wouldn't get caught out. When you are experiencing something about you and emotion - you are no longer as logical. If you get someone involved within the story, they critique the data less. Yale researchers proved this by comparing the same story in a newspaper compared to a stronger narrative. 

Hacking is about 'Humanint' - Human Intelligence - and we are sharing lots of our Human Experience that are giving hackers the keys to what we do and therefore making data available to bad actors. 

The best magician is the best storyteller and distractor rather than the best technical skills.

The reason why we don't have good data on cons as we don't want to admit we have been conned. 

Data16 - Storytelling Through Comics - Irina Porodnova



Comic Theory:
1. Sequential Art
2. Visual Grammar / Visual Language
All to apply Comic Theory to Visual Analytics

Comics add single images together to add context to that singular event. This is an obvious link from a single worksheet / chart to good storytelling with dashboards, pages or Storypoints.



McClouds transitions:
A. Moment to Moment transition - very simple progression. The pages shelf allows you to do this in Tableau simply and you can walk through the story step-by-step or moment-to-moment.
B. Subject to Subject - A series of changing subjects within a single scene - Multiple panes within the same view but of the same chart.
C. Aspect to Aspect - different aspects of a place, idea or mood - multiple filters or storypoints to investigate the different angles.
D. Non Sequitur - No logical relationship between the panels - a lot harder to find the story as there is no logic here. Classic for someone's first Tableau data visualisation!
Key takeaways from Transitions: A create logical relationship. B/ Images develop meaning when placed sequentially. C. Can serve as a replacement for written words

Cohn - Visual Language of Comics
Frame Types:
A. Establisher - setup of the story
B. Initial - build up of the story
C. Prolongation - increase tension and increase anticipation
D. Peak - the moment of the story
E, Release - exit - maybe the punchline

Great framework for putting together storypoints in Tableau.

When a comic is in the wrong order, comic storys become very difficult to follow yet we do this with data visualisation often.

The neuroscience of comics - in research, out brain expects a story to flow so when it doesn't the brain kind of gives up and doesn't try harder to understand. Therefore, avoid suprising your audience as the hardest the brain works is at first in setting up the basic understanding of the story in the Establisher phase.

Will be interesting to try the different structures of stories:


Great takeaway: Simplicity creates a Great Impact all on it's own

#Data16 London - A Laguage for Visual Analysis - Jock MacKinlay keynote


In the 1980s Jock started his research in to data visualisation and lots of this is built in to Tableau. 

Visual Analysis uses the power of visual language and visual analysis to combine to be very powerful. 

Jock reading Orwell's 1984 gave him the idea that to protect yourself from Big Brother was to use Data. Playfair's Balance of Trade visualisation gave him the idea to build a programme to allow people to make these kind of charts. 





Bertin's 'Semiologie graphique' (translated in to English from French) became the spine of Jock's phd. The book focuses on the syntax. The main figure allowed Jock to understand what the key aspects to visual analysis were.
Selection: 
A. Association 
B. Selection
C. Order
D. Quantity
Jock joins Association and Selection in to 'Categorical' data The horizontal axis of size, value, texture, colour, orientation and shape allows you to make the association. The language aspect comes from taking visual components and their use is the equivalent of turing words in to sentences.



Bertins actually started out as a Cartographer and hand drew his maps. Understanding the visual system allowed him to make them as clear and impactful as possible. Bertin's built physical tables with rods to allow for sorting (Permiatations Matrices)


At Bell labs, Cleveland and McGill were scientifically researching human performance that validated Bertin's work. Here is there findings:



And Jock's combination of the two:


Jock managed to turn this thinking in to software. Here's a photo of the recording of his 'apt' tool.





Jock got to work at one of the top technology parks (Xerox Park) on the west coast. Dr Stuart Card developed the first mouse as a pointing device.



And some of the first developments the team made:


Human Visual analysis pipeline was created in the late 1990s and forms the basis of normal visual analysis now and how the tools are designed this flow. Tableau uses this flow that there is complex backtracking as the process isn't purely linear as you make analytical decisions:


When Pat Hanrahan transfered to Stanford, Jock got to spend time with Pat and Chris Stolte at Xerox Park. Pat turned Chris to Jock's dissertation. Jock's work was missing the link to databases and that was what really lead to Tableau. 


Show Me is Jock phd research in action. Here is the logic behind the scenes:


Jock's demo of wildlife strikes is great. He experienced a birdstrike that damaged the front radar and therefore his favourite dataset to demo became a lot more pertinent. 

The present - Research. Tableau has a strong research team to do fundamental research as it is really trying to understand a lot of visual language aspects. Here is some of the research that is currently happening. 

Humans and computers are different in their skills. Language makes Humans cooperative. Humans are intuitive, computers are not but they are great at diving deeply in to the data. A few researchers have dived more deeply in to this. The 30-40 categorical data fields can be analysed a lot faster than humans so could Show Me be developed to find the signal within the data? 


In organisations, we need to cooperate. This can be tough to do so can computers show who has what data more effectively? 

...more to come

Tuesday, 14 June 2016

#Data 16 - Brian Cox keynote

How data helps us study some of the biggest questions of our age?



The Theory of General Relativity shows how thinking about fundamental questions can lead to massive examples. Einstein challenged the way we think about gravity and the forces in the universe.


Freefall is a state of rest and a rewrite of Newton's laws. We see this in the International Space Station where everything stays still when released. It's why sitting is hard work because we are under force from gravity.

Einstein replaced force with geometry. Newton's law of attraction was replaced with Geometry (Space Time) - Einstein's Theory of General Relativity. Using Einstein's theory about gravity, you get the idea that the Universe is moving and therefore there was "a day without a yesterday" (Lemaitre). Using the equations therefore predicted what was in the universe before it was discovered and proven.


Huble taking pictures of Andromida in 1923 and understanding behaviours of types of stars proved there were other galaxies to our own. Huble went on to use 'Red Shift' to show the expansion of the universe in 1929.


This data is essentially wrong so he was wrong. The data was represented properly but his data was actually wrong (disproven with modern data)! Basically, Huble was measuring the galaxies were too close.

There are 350 billion galaxies in the observable galaxy. There seems to be patterns and therefore, what lead to that sort of pattern is one of the big unsolved problems. If you held a 5p coin to the sky, 75 feet away you would still get 10,000 galaxies in that tiny piece of sky. That means there is a lot of data to analyse.

Gravitational measurement of the stretching and the squashing of space /time. One detector is in Louisiana and one in Washington State. There was a detection made of distruption because of two large black holes merging. The actual event happened over 0.05 seconds. It went from 1/3rd to 2/3rds the speed of light in a 1/10th of a second. It left an object the mass of 60 times the mass of the sun. This is the triumph of Einstein's Theory. It's the first time that light wasn't used and gravitational waves that were.

The idea of the moon before we started exploring our Solar System was of a barren moon (due to our own). This is far from the case. The rings of Saturn are just 2 meters thick but 100km wide. One small moon of Saturn hass water and is very active. These water vents (pictured) is very much like our own start of life on Earth.


The Large Hadron Collider is producing huge amounts of data and is mimicing the initial collisons that were present at origin of the universe. Dr Brian Cox works on the Atlas detector. It captures Proton / Proton collisions and this actually captures a Higgs particle (Muons are the red lines)


And extra findings like (not statistically proven yet) but there are particles that are not currently recognised that could be dark matter or signs that additional dimensions actually exist. My mind is offically blown right now so apologies for the words after this part...


The 90 billion light years across can be measured back to 10 x-22 of a meter across. This distribution helps to predict with precision about the distribution of particles. The eternal inflation model shows that there could be multiple big bangs as part of a longer cycle so there could be an infinite number of universes. Welcome to the Fractal Multiverse!!