Wednesday 10 September 2014

Seattle Tableau Conference - 2014 Day Two

Keynote – Neil deGrasse Tyson – Science as a Way of Knowing


For the British audience Neil is an astrophysicist who is the American Brian Cox. He is the host on COSMOS on TV in the US and this is on Primetime

Pluto is still not a planet – according to Neil

The Supermoon (when the moon the moon is closest), is only 0.001 inches larger than a normal full moon. Data is useful.

Bad Maths – “Half the schools in the district are below average” (actual newspaper headline). Member of Congress “I’ve changed my view 360 degree on that issue”

There is an atlas of peculiar galaxies (interesting shapes) created in the 60s. From data, once the computing power was available, it can be modelled. 250 million years can be processed in 5 seconds and accurately modelled. Every star in the galaxies are modelled in their interaction against each other. Proves the galaxies are wrecked galaxies due to each one being influenced by the other.

LSST – Large Synoptic Survey Telescope – still fund raising. The most powerful telescope. Full image of the universe every three days. “Dim and Dark picture”. 3.2 Gigapixels – World’s largest digital camera. 30 Terabytes of data nightly. Will create the first Movie of the Universe.

Killer Asteroids – The Meteor Crater in Arizona is the remnance of a killer asteroid. The Asteriod that took out the dinosaurs was roughly the size of Mount Everest. The impact of the asteroid in Russia in 2013 was the equivalent of 25 Hiroshima bombs.

There are 1,500 asteroids that are going to cross the orbit of the Earth. In 1980 there were 9,000 asteroids that astrophysicists knew about. By 1996 it was 25,000 asteroids. 2010 there were 550,000 asteroids found (most in the asteroid belt between Mars and Jupiter). There are a lot of asteroids been Mars and the Sun and this means they are going to collide with the Earth at some point.




Facebook Jeopardy: the Hack Edition - Andy Kriebel & Bryan Brandow

Hacks will be shared on Bryan and Andy’s blogs and are nearly impossible to capture in a few sentences so here is a clue as to what was discussed

1. My ETL was delayed today and it missed my refresh schedule?

Triggering when the data is done – using a self built tool called Trigger. The code to be posted later

2. My extract has been failing for the last few days and I haven’t noticed

A self created tool that you can sign up to extracts so you can see when they error. Emails are sent for successes as well as failures

3. Extracts are taking too long to load

Tableau Data Extract API is used to do this. The Information Lab use Alteryx for this. Facebook made a 40% saving on their server performance that you can funnel elsewhere instead by putting the loads on a separate server.

4. I like drilling down, but I hate losing context by switching tabs

Add the additional chart as a floating option. Use dashboard actions. On select and exclude all values so only by selecting it appears.

5. I want to build a waterfall chart for my 37 sheet dashboard to render

Create a scaffold and structure the data to enable it to all appear on one sheet

6. I wish I could expand my lines in a row when expanding a hierarchy

Done through floating objects that are hidden by dashboard actions

7. When my extract refreshes the data range quick filter doesn’t change the maximum date

Change the filter to use all dates. Tableau’s automatic setting is a set date range

8. My users are creating spaghetti charts! I need this to stop

Use Tableau’s SIZE() function to be less than 5. Create a COUNTD([Dimension]) that looks for 6 or more

9. Replace Data Source is Killing me. Surely there is a better way.

Sets disappear and calculated field drop. Open twb file with a text editor. Cut and replace the second data source with the first. Save as a new workbook.

10. People complain my dashboards are slow and I don’t have a way to quantify it

Using the Javascript API – look for CUSTOM_VIEW_LOAD and identify the dashboard finishing the load. You can look at browser type, server used etc

11. I took my laptop to a meeting and went to show the dashboard and it refreshed in my face

Using the Javascript API you can set up a control click to prevent refreshing and you can set up where it refreshes to. Pause automatic updates after the time your server goes to refresh.

12. Load dashboards faster

Load a static representation of the dashboard first whilst everything else loads behind the scenes use the .png image from the server



Porn, Pokemon and Pop Culture

Jewel Loree

“This conference is the only one that is like a big family reunion”

There are many reasons to have a blog – comments and response ensures you develop. It’s a friendly critique.

Subjects that are fun for people, make it easier to communicate to friends and family what you do. It’s also a great way to get recruiters attracted to your skills

If you’re engaged by a subject, then you will find others who are too

Andy Kriebel

“Friends don’t let friends use pie charts”

Andy didn’t know if anyone would read the blog, he now works at Facebook – he got noticed

The most popular content is tips more than anything else but one post about how common is your birthday has 250k views as it got picked up by the Huffington Post

Blogging has been a great way to enable him to write which he enjoys

Peter Gilks

Started blogging in 2012

The desire to have a Viz of the Day post published meant Peter got steered to having a blog

Theming creates a lot of woah to add a great feeling and impact

A CV doesn’t cut it if you are applying for Tableau jobs. Therefore the blog link really lets people see what you can do.

The IronViz contests are a great kickstart because it helps guide you on a story.

There is nothing wrong with creating the data yourself




Michael Lewis – In conversation with Kelly Wright

Art History Major from Princeton who is attracted to characters where data plays a central role. The characters’ disrupt their environment with a tool and that tool is often data.

Dysfunction on Wall Street was caused where data was hidden and difficult to use

Michael’s wife critiques his ability to look at one data point and see a trend

The change in professional sports is breath taking (in terms of data)

Oakland A’s had no other choice to find a competitive advantage than data (they could afford to do it). It’s a sign of market inefficiencies that the team (company) with less resources can win against those with more resources

Time in the locker room meant that he saw the players weren’t in the best condition / perfect human bodies. It’s about getting past the subjective opinions and look at the outcomes. The key question is, where else is this the case?

Oakland did not generate new insight, the ideas were out on the internet. They applied those ideas and had the courage to do it. The A’s could do it as they operated in a market where they didn’t have much to lose.

The controversial reaction to Moneyball was similar to Flash Boys. It’s just that Hedge Funds and Banks had a lot more resources to fight back. ML is still surprised by important people still lying about the situation and trying to discredit it.

Entrenched attitudes in Financial Services makes change and innovation a lot harder to make the changes.

The consistent trait of those who build change is storytelling. Being able to create Narratives is absolutely key.

Taking complexity out of the subject is key to get the point across. Michael uses the ‘would my Mother understand’ what I have just written. Michael goes out with friends to a bar to test out the narrative.

Pictures are a great way to make things simple but they are not easy to create.

ML – “if you take the chance to appear stupid by making your audience feel smart [by making it simple] then you will succeed more”.

Eliminating what isn’t necessary makes everything more simple and makes the story telling a lot easier

The hole in Moneyball is that following a sports team you love doesn’t make following just a rational approach easy to join the two experiences together




Tableau like a Sith – Darth Flashypants and Lord Jinbar



A full intro video (including dancing Stormtroopers) was amazing!

Use the techniques are the focus of the session rather than the tasks

1. How to get more

Advanced table options (you can increase to 16). To get 17 or more. Go in to .xml and change the options to as much as you want (but only after you raise the number to 16).

2. Downgrade the version of the workbook

In the .xml go and change the version number in the opening lines

3. Tableau performance recorder

In setting and performance in the Help menu, you will find the performance recorder you can start and stop this to measure your processes

4. Gauages

First you need to create a copy of the data – one with Start and one with End in the calculated field – and in End, take all the measures out

You need to use sin and cos to work out where the line of the needle is. Use a path line to create the end point of the dial. Use radians and not degrees to work out the angle. Some nice techniques to dig in to here.

5. Images

Create high and width of image. Use background images. Match the images. Use a web domain. Click on options in background images and select a filter. Open file in text editor. Search for <mapped-image/> and take that and copy in to excel. Use Excel to concatenate the data with the new image information (ie mapping the image to the customer name)

6. Exploding pie chart

Overlay two pie charts using duel axis. Have one at a deeper level of the hierarchy and then increase the size of the lower level of granularity and it appears from behind

7. Agent Ransack

Find interesting things on your server. Ie the login box – change the wording on the login screen. Find the underlying files in the server and search for the wording and alter it to what you need.




The Plot Thickens – Using Visualisation to tell stories - Robert Kosara



The connected scatterplot shows the flow of data or time really smoothly. Current examples are temporal so far. Use the label to show year for example. Requires neat annotations

The designer needs to think about implicit vs explicit time. Explicit time can be created in Storypoints to convince the user to explore a visualisation in a set pattern

Why do we tell stories?

Research shows even back to the 1920s that stories maintain attention span of individuals for longer. Good tv adverts tease a story in the first few seconds to keep viewers’ attention

Stories

are also really good at making knowledge stick for longer (it’s not currently known why)

Simplification is mentioned yet again. Simplification = Effectiveness

Exploration of data is great for a user but to make it easier you need to use ‘data scent’ to find the additional stuff – whether it is tooltips, highlight filtering, trails etc

Exploration is not a story as you are not guiding the user

Hierarchy – from bottom base needs, upwards:

1. Story

2. Annotation

3. Visualisation


2 comments:

  1. Great summary! For this one in the FB talk:

    6. I wish I could expand my lines in a row when expanding a hierarchy

    This was done using containers not floating objects. You can't shift objects down using floating objects, but you can with containers.

    ReplyDelete
  2. Carl - thanks for the update as always :-)

    Andy - look forward to seeing more info on your hacks on your blog - there's a few points that interest me

    Donna

    ReplyDelete