When I started using Tableau in 2006 Enterprise Data Warehouses were the sine qua non of Business Intelligence.
Common wisdom, backed by the opinion of experts and years of practice, held that the only meaningful way to achieve value from an organization's data was to collect, organize, homogenize, restructure, and present it in enterprise-spanning industrial scale data stores from which meaningful and valuable reports, dashboards, balanced scorecards, strategy maps, and their ilk could be developed to deliver business information to those few, those lucky few, who could make use of it.
Notwithstanding the obvious self-defeating biases of this hierarchically-stratified decision-making paradigm, creating these massive data cathedrals and their information-exposing front ends turned out to be one of the most difficult, expensive, and error-prone business undertakings. Failure rates of enterprise BI projects have historically been embarrassingly high. There are many causes for this, the primary being the almost inevitable disconnect between business needs for timely data-origin information and the inward-looking technology focus of the BI development effort.
Tableau, used appropriately, can rescue enterprise BI through the application of its fundamental ability to immediately access and instantly and effectively analyze and understand business data wherever it lives. The key being: data wherever it lives.
Tableau provides a previously impossible, and largely unrecognized, opportunity to change the Enterprise Data Warehouse-based BI model. It does this by enabling several fundamentally new approaches to applying direct and pervasive data analysis to the full spectrum of activities.
The Business View.
From a business stakeholder's perspective the whole point of BI is to deliver accurate information when it's needed, in a form that makes it easy to understand.
The business stakeholder should not be concerned with, nor adversely affected by technological factors—these should be safely hidden behind the curtains.
Using Tableau, it's possible to work intimately with the business stakeholders to analyze their data, in their local context, and achieve immediate, even instantaneous insights into the data they need to understand to make effective decisions.
This intimate coupling of people with their data observes the simple truth that all BI is local—people are first and foremost concerned with the data within their immediate business horizon; without understanding that data they cannot do their jobs properly. This is the truth that traditional enterprise BI failed to recognize.
Creating and delivering these analyses up front and continually achieves an immediate knowledge benefit to the business stakeholders, which invests them in the process and builds their trust in the ability of the initiative to deliver the goods, which is a tremendous incentive for them to remain engaged.These high-value business analytics are then the gold standard requirements for the analytics the enterprise BI system needs to deliver (if it needs to deliver them is a legitimate question). Having been already vetted by the business stakeholders who need them, they provide the bright line path that can guide the whole program, from the design of the ODS and Data Warehouse, to the final operationalized reports, dashboards, etc.
So do we need Data Warehouses?
Given the rosy scenario painted above, the elephant in the room is the question: do we still need data warehouses and the benefits of enterprise BI?
Tableau enables Business Intelligence Delivery.
Yes. We still need Enterprise BI and Data Warehouses.
The problems and failings with traditional enterprise BI are of practice and implementation, not of principle. Attempting to build the universal answer-anything universal data store and query engine is a doomed approach. But as long as an organization's data is collected by disparate systems, and this will likely always be true, the need to collect, organize, and align it into a common context so that sense can be made of it at the enterprise level will exist. This is the raison d'etre for enterprise data warehouses and the enterprise platforms that support them.
How then do we conduct Enterprise BI?
I'm glad I asked.
It's simple: build outward from the business stakeholders' information needs. As shown above, and in proven in practice, starting with live, valuable stakeholder-approved analytics works. Using these analytics, every enterprise BI element can be conceived of, designed, and implemented with a clear idea of exactly how it will support delivery of one or more of the business analytics. Anything that doesn't support this goal is unnecessary, even harmful in its pollution of the BI space.
But isn't Tableau just an eye candy toy tool, good for making pretty pictures?
It doesn't fit in the industrial strength tool arena where real developers build real data infrastructure.
It's a big surprise to many BU traditionalists schooled in the industrial production process approach, but introducing Tableau into an enterprise BI project can provide tremendous benefits in the velocity, accuracy, effectiveness, and quality of the activities across the project process spectrum. I've been successful in rescuing multiple enterprise BI projects that had gone off the rails by bringing Tableau into the mix. At first it was for my own use in understanding the data in play. It didn't take long for my ability to rapidly and effectively understand the data to be noticed, at which point other people would become interested and start either using Tableau for themselves or asking me to provide them with the analytics from their data. Eventually, Tableau found a place across the full range of activities.
This might seem strange, as it did at first to many of the technical people. After all, enterprise BI projects are conducted by "real" developers using "real" SQL-based data analysis tools, sometimes native to the technology in use, sometimes standalone tools like Toad. Which is one of the problems at the heart of traditional enterprise BI - the people who need to understand what's going on inside the data processing systems are hampered by their tools. SQL query tools are by their very nature low level mechanisms ill suited for the higher level cognitive activities involved in data analysis. This isn't to say that SQL query tools don't have their place—they do, but it's not as the primary tool for understanding bodies of data, and assessing them in terms of expectations and realities. That's what Tableau was designed for, and does so exquisitely well.
Here's a link to a diagram showing how Tableau fits into a typical enterprise BI project. It's too large to fit into a web page, much less a blog post. But you should look at it, preferably at poster size.
The diagram includes the Business View from above (which was really extracted from the main diagram), connecting that perspective to everything behind the curtain that's out of the business person' sight (or should be). When used effectively Tableau, can be used by everyone to help them understand their data, as shown in the shaded areas in the diagram.
Used in this way Tableau provides the opportunity to introduce enormous benefits and gains in the conduct of those processes. Since the entire process chain is at heart a series of data processing stages, Tableau's ability to instantly, flexibly, and effectively analyze any and all of the data means that the internal project people can lift their efforts above writing SQL queries or looking at table-based tools to the realm of Tableau's visual analytical approach. It's not unreasonable to expect that by using Tableau effectively enterprise BI projects can be delivered successfully, with greater value and stakeholder satisfaction, in 40% of the time, at about the same fraction of cost, as projects conducted in the traditional manner.
Tableau isn't a silver bullet.
Tableau provides the opportunity to improve the conduct of enterprise BI projects. It doesn't guarantee results. The effectiveness of any project rlies upon the skill and effort of the people involved. All Tableau does is provide the means whereby enterprise BI projects can be directed towards delivering what the business stakeholders want and avoid working on anything they doesn't contribute to that effort, while removing the internal data analytical barriers that can impede progress.
Whether, when, and how Tableau actually replaces data warehouses and enterprise BI for the delivery of valuable business information are questions undergoing vigorous debate. There are no clear lines segregating the realms—the boundaries are and will remain flexible. We'll cover at least some of this territory in future posts.