Netflix, Spotify, and Home Depot Won Tableau Conference 2018Shelby TempleBlockedUnblockFollowFollowingFeb 4Wrangled together the three best presentations on data analytics from TC18.
After sifting through 439 sessions, I have found my three favorite presentations from Tableau Conference 2018 (TC18).
Netflix, Spotify, and Home Depot are my winners! For this post, I will dive deep into each presentation and provide the key takeaways.
Tableau is the leading Business Intelligence tool for data visualization and they hold a conference every year that celebrates everything data.
I would say most of the sessions fall into one of two buckets.
1) To teach you how to Tableau better.
2) A company sharing how they are doing amazing things in the data analytics space.
For this post we will focus on the second bucket— the company data stories.
I went through many, many presentations to provide you the best of the best.
So you can be brutal with your time and not waste it on urinal presentations (see the end of the article).
Also to crank up the post intensity I am giving away fake awards to the various companies that presented.
NetflixPresenter: Blake Irvine | Manager Science & AnalyticsAward for the Most Binge-able PresentationYouTube Views: 932 | Time: 54 minutes (includes QA)Not only do they kill it with their originals, Netflix tends to reliably put on a great conference session.
For a company that is hesitant to share high level viewership numbers, they sure do love to pull away the curtains at TC.
Key Takeaway 1: Netflix is a Big Data Heavyweight Champion but they are still humble about their challenges.
In my experience Tableau is best when you can extract the data.
One cannot simply extract 300 Terabytes of data and begin to build dashboards.
They cover many ways they deal with their big data, including an open source product called druid and an in-house solution that lets employees run SQL queries that will push the output to a server as a Tableau Data Extract.
To most people the amount of data Netflix has is absolutely amazing.
Fun test — book mark this and show people this slide in 25 years — probably will be as laughable as the space on a floppy disc.
Even though they do great work with big data they do acknowledge being bottle-necked and still having to decide between speed and size.
Key Takeaway 2: Netflix lives into the data heavy start up culture.
Measurement and experimentation are the backbone of this company.
Roughly 10% of the employees work on a dedicated data analytics team.
Roughly 50% of their employees (2,574 out of 5,000) have accessed a Tableau Dashboard.
Here is their data analytics stack and analytic ecosystem:Lumen and Big Data Portal are in-house solutions.
This presentation went heavy into the Big Data Portal (discussed more below).
I found it interesting that all their data is starting in AWS.
Key Takeaway 3: Netflix is secretly a ninja in Data GovernanceAt every Tableau Conference, there are 50+ stories along the lines of “Everyone was using Excel then we brought in Tableau and now the company is better off.
” The narrative has become so tired at TC I mostly decided to ignore these presentations.
However, this year I did notice a second narrative emerge which is, “Our company is siloed and everyone is doing their own thing and there are data and reports everywhere — with no help on what to trust.
”Blake mentioned the double edge sword of liberating the enterprise with data.
The result is many, many reports and tables.
They multiply like rabbits.
However, through their in-house solution, Big Data Portal, and other various efforts they seem to be way ahead of the curve on this issue.
The Big Data Portal allows users to first search all existing projects and tables in the company to see if what they are looking for already exists.
Then if it doesn’t they can go to step two and begin to write their query or build their project.
SpotifyPresenter: Skyler Johnson| Data Visualization DesignerAward for Best Data StoriesYouTube Views: 433 | Time: 28 minutesKey Takeaway 1: Spotify is doing something more companies need to do.
Which is share cool data stories with the public.
Skyler highlighted some of the cool stories Spotify has shared such as people playing songs about eclipses peaking in the region when the actual lunar eclipse passes over.
Key Takeaway 2: Text and language processing needs to always consider including emojis.
They are now a part of our language.
Did you know which Emojis are most commonly associated with your favorite artists?.Spotify took the Emojis that are posted with artists on their platform and turned it into a game.
Can you guess these artists with their top ten emojis going horizontal?Without the answersSpotify Interests; Tableau PublicWith the answersSpotify Interests; Tableau PublicYes, I am controversial and included R.
Kelly.
Just thought the emojis were outrageous.
Key Takeaway 3: Passion trumps years of experience.
For Skyler, doing something he was passionate about helped him land his dream job at Spotify.
Skyler loves music.
Skyler told amazing data stories on his own time using the publicly available Spotify API.
Eventually Spotify took him in-house as an employee.
He begins his presentation by mentioning he has only been using Tableau for three years.
Yet he has some of the most popular data visualizations on the Tableau Public platform.
Home DepotPresenter: David Berry | Leader of Business Intelligence & Chase Zieman | Leader of Analytics and Data ScienceAward for Best Simultaneous Use of Prescriptive Analytics & PunsYouTube Views: 307 | Time: 57 minutes (with QA)Not Tooling Around; YouTubeKey Takeaway 1: A great way to enhance a business relationship or feedback loop is through a dashboard.
Home Depot owns a company called Blinds.
com.
Blinds.
com does not actually manufacture the blinds.
They use numerous vendors for their product.
In order to have a better relationship and feedback loop they provide a Tableau dashboard to their vendors.
One takeaway here is that they mentioned their vendors aren’t always the most tech savvy.
They intentionally designed their Tableau dashboard as something that can be easily pushed to a static PDF and shared more like a traditional scorecard.
Obviously being able to interact with a living dashboard is preferred, however, it cannot always be expected that the audience will engage with a dashboard in that manner.
Key Takeaway 2: If you have a website you should be modeling off your click stream data.
Blinds.
com bridges this data set with everything they already know about the customer.
The algorithms are able to point out and uncover improvement opportunities for the website, customer experience, or if the customer might just be prone to making mistakes themselves.
Key Takeaway 3: At the end of the day judge a model not by its accuracy, but by its impact to the company.
When the algorithm picks up orders to review, orders that are predicted to have issues, it ranks them by expected cost to the business.
Chase mentioned there are so many different KPIs you can judge a model by, but the most important thing the model can do is reduce costs and help the company do better — so they judge their models by the reorder probability multiplied by expected reorder cost.
This gives them an actionable metric that the business can leverage.
ConclusionThese were my three favorite and most time worthy sessions.
I wanted to go ahead and drop the link so if want, you can explore the remaining 436 sessions.
Tableau Conference 2018 Session LibraryAlso since I watched a bunch of these videos I wanted to share some other company sessions that were considered but didn’t make the cut (yet still worthy of a fake award)Didn’t Make the Cut Awards:Johnson & Johnson | Award for Best Implementation of a Tired, “We’re Better than Excel Spreadsheets” storyGoogle — Nest | Award for the Session Everyone Wishes was Recorded, but Wasn’tFacebook | Award for 1st place in the “I Came to be Impressed but Instead I was Shown Slides of Urinals” Contest (they were the only one to compete)Boeing | Award for the Most Simultaneously Beautiful and Ugly DashboardAlberta Health | Award for the Most Vigorous Use of the Stop Light Color SchemeRun Disney | Award for the Most Beautiful Deep Dive into Mediocre DataPortland Trail Blazers | Award for Implementing Team Colors into Every Viz Possible.. More details