Wednesday, 10 September 2014
Seattle Tableau Conference 2014 - Day 1 Summary
After a good amount of feedback from those who couldn’t be there (and those who could but didn’t want to take notes), I will be writing summaries and updates through the Tableau Conference 2014 (#data14)
Keynote – Christian Chabot and Chris Stolte
5,500 customers, a bunch of Zen Master and goodness knows who else are here. This conference feels significantly bigger than even last year. If you have never been to a Tableau conference then I really can’t get over the collective buzz that is generated. It feels like a sports event when the home team is about to win the championship. That’s crazy for a tech conference…
Christian starts by highlighting the great work (and breadth) of the Tableau community that is shared on the web wherever you go. Matt Francis, Ramon Martinez, Ryan Robatile and Kelly Martin’s work is all shared on the big screen.
The next big strides in the technical revolution is to ‘expand your creative potential’ of everyone. Adobe, Computer Aided Design and now other creative tools are allowing everyone to explore and create great design.
‘Data Analysis is a creative process’ – most important role in the modern business strategy
Analysts and artists are both on a mission to create something new and explore. Same human characteristics are used in both.
To create a tool that allows for a great creative experience takes four things:
1. Encourage experimentation – fast prototyping needed in any innovation process. Taking film out of cameras allows people to explore images and the quality of photos has risen and improved as a subject area.
2. Speed – To iterate, you need things to be able to iterate in front of your eyes. Tableau working on performance a lot
3. Expressiveness – Great artists don’t use Paint by Numbers art sheets. Business has been tied / constrained Business Intelligence solutions. Tableau creating a canvas.
4. Control – giving people control. Something is in your mind’s eye, you need to be able to get it out. Di Vinci didn’t do his work by “filling out a painting request form”.
Tableau 200 customers by 2004, 2,000 customers 2008, 20,000 customer by 2014. Growth is funding R&D in Tableau in the next two years than the previous 10 years.
7 areas for this development is:
1. Visual Analytics
3. Data Prep
Chris Stolte - New features:
1. Visual Analytics
Now double click and type in to the column or row shelf instead of dragging pills. This is done like creating a calculate field. You can double click on a pill (existing data item) and you can run complex calculations too. You can drop blended data sources in to those calculations too. Called “Freeform Calculations”. You can drag the new fields and drop it in the dimension or measures list to create the data item.
New calculation editor is a lot more simplistic. Edit the calculated field and can interact with the view at the same time. You can drag in data from the dimension and measure lists.
There is a new side pane too – the Analytics pane. Where you currently have data lists, you can flick to analytical functions and trends.
Reference lines update with the data you are selecting on screen.
Tableau calculations – can be hard so Tableau making them easier. Views respond immediately as you flick between table calcs. Tableau highlights what is being computed.
Geographic search now built in to Tableau – so you can flick between countries, states, postcodes – just type it in and the map filters (all built in to the long / lat)
New selection tool so you can Lasso, select circular areas etc to select really what you want. Very cool and is going to be great on scatterplots!
Tableau can go anywhere with data in very many ways – makes performance a lot harder
Andrew Beers (VP Product Development) demos
Tooltip respond instantly as you scroll round the visualization. This is still the same Viz engine bought in in version 8.0.
8.3 is showing performance improvements of x2-4 times faster. Taking advantage of multi-core processing that is now available in most machines. Using more parallel processing is changing response times to seconds rather than minutes.
Doing more in the browser is continually requested. Tooltips are instant and maps panes appear seamlessly during panning.
Analytics at scale – persistent query caching – shared across all nodes on the server and all processors.
3. Data Preparation
In the data load window, you can now ‘Split’ one data column in to it’s parts. Split is now available when working in the visualization window too.
Poorly formatted Excel sheets are now going to be easier to clean up. Just connect and Tableau is now becoming smarter to clean those excel files. Transposing data through ‘Unpivot’ in the data load window (big reception for this). Handling survey data from agencies is going to be a lot easier
Web Services data – REST API and JSON – “Web Data Connector” now possible
Dr Jock McKinley
“The next chapter in storypoints”
View thumbnails are produced so you know what you want to drag in to the Storyboard on your list of worksheets and dashboards.
You can now format the storypoints navigator – background colour, font, style (numbers instead of word descriptions), size and position.
See Enterprise deployments as mission critical.
Scalable, Resilient and Easy to Manage – are the main aims
Secure – Kerberos, smart cards and permissions
Extensible – API improvements – Javscript, Data Extract and REST API. Publish content and assign permissions through the APIs to come.
Thumbnails of the workbooks, instant search on the server browser and all appears instantly
You can look at the Data Sources of the workbook on the details in the server (I love this little touch!)
New heatmaps for the permissions in the server so easier to see what is allowed for
Tableau Online allows On Premise as well as Cloud data storage and then deploy to mobile
Tableau Online allows a live-to-live cloud base querying and connection
Tableau Data Safe – allows Tableau Online to control On Premise feeds in a secure way.
Tableau Online can embed directly in to SalesForce
App performance speed increased as you scroll through the thumbnails of the viz.
Now allowing for calculations in mobile and web editing for the first time – this will be huge
Swipe through the visualizations to switch between them
Offline snapshots captured at your settings down to every 15 minutes
New tool!! – Project Elastic
From a csv attachment – you can open it in “Project Elastic” and it creates a basic visualization for you. Click and drag to filter the data. Switch categories (dimensions) by sliding your finger across the dimension title. You can change the aggregation by sliding up on measure name.
Email the new image in two taps
Date hierarchy explored by stretching the time series – stretch an individual record to see the underlying data – genius!
Alberto Cairo – The Island of Knowledge and the Shoreline of Wonder (visualization for a better world)
Professor from Miami University and Data Visualisation Author
New Book coming called ‘The Insightful Art’ (working title)
AC believes sketching with pens and papers is the most important first step of creating visualisations. Aligns to Christian Chabot’s thoughts on design.
“A visualization is a display of evidence” – convey the information in a way the listener / viewer can absorb easily.
“Information shaped as a graphic function as a cognitive aid” – what ever is on the screen aids the cognitive understanding and importantly isn’t the cognitive function, it’s just the help for that.
“Words alone are useless, but so are visualisations” – it’s the combination of the two that makes the magic happen
“Good answers lead to more good questions” and visualization can really help this process.
Learning stops when blockages / impairments add too much resistance – often happens with lies / political bias etc
With the larger Island of Knowledge, the Shoreline of Wonder grows too.
Data visualization tools that are free and open can perpetuate bad behaviours – misleading visualisations.
Visual.ly promoted an article about creating an infographic from start to finish and it articulates the creator should decide their story and then find the data – very dangerous and likely to mislead.
The Challenge of Clarity – AC argues there is a value that precedes clarity. Truth is not an absolute, it’s a continuum.
Good features a visualization should have:
1. Truthful – be honest, be aware of your own biases
2. Functionality – needs to work and smoothly
3. Beautiful – draws the audience to the work
4. Insightful – allows the viewer to learn and if possible explore further
Only a divine being can truly be truthful – hence truth is a scale. For objectivity you need to be open to argument and debate.
A habit of skeptical thinking can be transformed in to a habit of visualization design
When working you should increase BREADTH and DEPTH to fully explore what you are presenting
Data literature takes for granted the quality of data. The books do not protect individuals from lying to themselves with data
“The first principle is that you must not fool yourself and you are the easiest person to fool!” – Richard Feynman
The phrase, “It’s more complicated than that” should always be kept in mind – never feel satisfied in the quest for understanding.
Tableau - Tricks for becoming a Jedi
1. By dropping the new pill over the top of the old pill (measures) it maintains the reference lines and forecasts etc (time saver)
2. Summary card - good for calculating totals and other levels of aggregation as you change what you select on the visualisation
3. In the analysis menu, you can fix the visualisation to always use a table calc on a certain workbook. Also, the partitioning too!
4. When nesting table calc, the original set up of addressing and partitioning is carried through as a new default.
5. Tableau does some Smart Sorting when it absorbs data that just makes sense - temporal sort for days if week or month name (even when the field isn't recognised as the date). Also works for IP addresses too.
6. Trellising of small multiples done through the two calculations pictured using index and sqrt. You then have to change the table calcs to compute by region etc
7. Moving reference lines - you can action a worksheet to itself
8. You can blend on a dummy data item. Just type in "dummy" as a calc field in each data set.
9. You can highlight at different levels - ie region, state, county etc
Zen Master Tips and Tricks – Craig Bloodworth, Kelly Martin and Mark Jackson
To filter data set for a condition (but not only those items) you need to create a new field that is a Boolean looking at another new field looking at CONTAINS(product, product parameter search). Apply that Boolean to Conditionally filter the Customer name / id
You can create layers of these by working with the Context filters (google it if you don’t know what I mean by this)
Tableau calculations in filters will act after all others. Partitioning in Tableau Calcs Advanced settings is like Group By.
Tableau Server and REST API
API – Application Programming Interface
The Google Extension for Tableau by the Information Lab utilizes the API to give fast access to workbooks and data sources
The undocumented API (v1 – been present since Tableau v3) and it can do everything that Tableau Desktop and tabcmd
1. Download Fiddler – captures all the interactions between Desktop and Server
2. Look at /auth.xml
3. Choose a language or tool (Alteryx, cURL, Python, Java, C#)
4. Understand the Authorisation Model
Using /datasource.xml /users.xml and /workbooks.xml to get the raw .xml response back. These are the files that The Information Labs tool is using in Alteryx
Stop users clicking certain objects by putting a textbox over the top
Use a dual axis chart to put a click filter over the top of the bars that you do let people click on. It looks like a selection mark
Use a help icon or an ‘About’ tab to explain the data where possible
Data Driven Storytelling - Ryan Sleeper
1. Know your audience – understand what they want and need
2. Smooth your Excel Transition – Tableau and Excel are very different tools. You can force things but use the right tools for the right job
3. Do you always need the number? You can still show the pattern and that give the reader the information that they need to act on
4. Leverage Colour – use neutral colour palette, limited number of colours. You are making the user work harder if you do
5. Keep it simple – strip back the dashboards as much as you can without taking away from the story
6. Use the Golden Ratio – most important things in the top left
7. Don’t Neglect the set-up
8. Don’t use pie charts – cogitatively they just don’t work. 5 slices or less just about ok. Never in a time series
9. Corporate Chart Types – Sparklines, Small Multiples, Bullet Graphs
10. Use call out numbers – make the key numbers jump out by being isolated or significantly larger
11. Allow discovery – include something for everyone (ie even the small teams for sports analysis)
12. Balance Data and Design – depends on the audience you are developing for
13. Eliminate Cart Junk but not graphics – Only a picture can carry such a volume of data in such a small space (look at Tufte’s rules
14. Tell a Story – annotations are great and makes things more easily absorbed
You can go to evolytics.com/15Tips for full right up
...more to come