I attended my first Tableau Conference in 2009 at the Hyatt Hotel in Seattle, quite an eye-opener for me. Speakers included the three founders, Christian Chabot, Chris Stolte, and Pat Hanrahan. In addition, there was a killer talk by Stephen Few, whose books are a bible for the visual representation of data (Show Me The Numbers, etc).
I was stunned by two things, which have not changed in the ensuing ten years: the unbridled enthusiasm from the customers and the stunning originality and quality of the product. As analysts in 2009, we sat around a conference room table with the principals; in 2019, there were over fifty of us there.
Before Salesforce acquired Tableau in 2019 for $3.7 billion, it's stock ticker was DATA, not VIZ. The founders understood from the beginning that the stunning visualizations were the vehicle, but data analysis was their mission. From business intelligence to data science to machine learning, the supposition was always, "Once you have the data, you can do this."
Tableau took that bull by the horns a few years ago as "Project Maestro," which became their data preparation offering (creatively named "Prep") as well as purchasing the Hyper in-memory database to make it performant. Two new features, Ask Data and Explain Data, are further features of their recently-announced Data Management offerings:
As the amount of data increases and the pace of decision-making accelerates, the need for data management has never been more critical to foster a thriving data culture,” said Francois Ajenstat, Chief Product Officer. “With Tableau 2019.3, we’re integrating data management directly into the analytics experience, making it easier for customers to curate and prepare all the data needed for analysis and improving visibility and increasing trust in the data for everyone within an organization.
My colleague Jon Reed at Diginomica.com has already written about his observations. I am going to focus on two new features: Ask Data, which was introduced earlier year, but it is substantially improved; and Explain Data, which is new.
Tableau's Ask Data - query and analysis with NLP
The major impediment to self-service analytics is the complexity of data, data models, and their relationships. With Ask Data, analysts have the tools to ramp up their analyses without researching the nuance of the data, and no need for programming, using Natural Language such as “What are the latest figures on damage estimates.” Ask Data understands the context of “latest” uses synonyms for “damage” and “estimates.” Further, without starting over, the user can evaluate the output and ask, “Just for the last 18 months,” and additional questions by adding follow-up queries or statements.
Beyond the kind of simple SQL-generating tools that are common in this market, Ask Data can handle more complex data such as time-series and spatial data. Natural Language technology is a superior approach to untangling the ambiguities and pragmatics of data.
Part of the magic is the use of a knowledge graph to organize relationships, contexts, and all of the other objects in Tableau's newly-released Catalog. We see rapid adoption of graph technology in data management because of its superior ability to capture semantic data and provide extreme performance when queried.
There is no special setup for Ask Data because it integrates with Tableau Server's existing security and governance capabilities, which makes it ready to use. Ask Data was introduced earlier this year as part of the 2019.1 release. I believe that is still the case with the current 2019.3 release.
Unlike data discovery or navigation tools, Ask Data augments the user’s investigation using inference from background machine learning algorithms that automatically profile and index data sources; providing suggestions that connect to relevant data and content.
Adding to the list in Tableau's latest release are Alibaba, SAP/HANA, Kyvos, LinkedIn Sales Navigator, and Qubole.
Another subtle but powerful feature is when a web page with Ask Data is embedded in the load, Ask Data opens the data source without a viz, waiting for users to ask a question.
There are too many to explain here, but this link provides a condensed review.
Tableau Explain Data - from the what to why
Tableau's latest release, 2019.4, will bring automated statistical analysis to visualizations via a feature called Explain Data. Explain Data is based on technology acquired from the June 2018 purchase of Empirical Systems. As Kyle Wiggers writes in Venture Beat:
Explain Data leverages statistical methods to evaluate hundreds of patterns across all available data, and deliver potential explanations in seconds. Users select the data point they wish to analyze, after which they’re able to view the results within interactive visualizations.
Explain Data is built directly within Tableau, and people can click through an icon to get automated analysis without data modeling or data science. Explain Data is also based on a set of algorithms that analyze all available data to surface statistically relevant items on a data point.
Francois Ajenstat is the consistently upbeat (why shouldn’t he be?) Chief Product Officer at Tableau. I don’t know of another senior executive in a technology company as open and clear as Francois. Recently, he said, “Explain Data is designed to evolve ‘what happened to why it happened, to surface something statistically significant that is not obvious.'”
Paraphrasing Ajenstat, users can select any data point in a visualization and utilize Bayesian statistical modeling (Why You Need To Put Bayesian Nets in Your Tool Bag) to evaluate patterns and explanations across multiple data points. As an added benefit:
Explain Data reduces the risk of error from human bias, Ajenstat says, “By taking into account every dimension of corpora. That’s as opposed to conventional solutions, which are typically constricted by predetermined hypothesis. Explain Data uses AI-driven statistical analysis so that, regardless of expertise, anyone can quickly and confidently uncover the ‘Why?’ behind their data. Explain Data will empower people to focus on the insights that matter and accelerate the time to action and business impact."
What are some of the aids Explain data provides?
- Anomaly detection
- Expected Values
- Calculation with Models (finally - see my article 10 Rules for Business Models)
- What if?
- Sensitivity Analysis
Here is a diagram that depicts the Explain Data process:
Existing Tableau clients who use third-party tools for prep will increasingly see the value of using Tableau-native tools. Prep, Data Management, Ask Data and Explain Data are compelling tools to drop the label BI from Tableau finally. It is far more than that now.