Self-service BI governance might sound like a contradiction in terms. After all, self-service business intelligence is all about giving users more access to data and the freedom to manipulate that data
For Autodesk Inc.'s Patrick Booher, however, a BI strategy first and foremost means establishing "trusted, governed and certified data." An effort to achieve that has been under way for the last five years at the San Rafael, Calif.-based software maker, he says, long before self-service BI was even considered -- not that he has anything against giving users more access to data.
In the many years that I've been doing BI, I've yet to find a good way to prevent people from messing up the data.
Self-service BI was introduced at Autodesk in 2010 to meet the changing information demands of users. "We had an increasing number of data sources and high demands on IT. Analytics and exploration were needed, rather than just [the production of] reports," Booher, the company's director of enterprise data management, said in an email interview.
Autodesk continues to use SAP AG's BusinessObjects for operational reports. It is retiring Oracle Corp.'s Business Intelligence Enterprise Edition, however, because it overlaps with BusinessObjects and with Autodesk's self-service BI tool, QlikView from QlikTech International AB. Using QlikView, about 1,700 users have created roughly 100 dashboards for all sorts of business analytics. For example, Autodesk customers can opt in to send product use data back to the company, and employees use QlikView to analyze this data and improve products.
But the biggest benefit "is the confidence people have in finding the answer," Booher said. "There are revenue, cost, security and customer experience benefits as well, but it's difficult to attribute a specific value to just self-service BI."
Also appealing is the speed with which users can create new data views without placing the burden entirely on Autodesk's IT or operations groups. The dashboards they create are published within two to four weeks on average. It used to take six to 12 months to get IT to deliver or change a report because the demand on IT was immense, Booher said.
As for self-service governance, a BI competency center was put in place that supplies users with standard data sets, design guidelines, design templates, security guidelines, legal training and, when needed, the expertise of external consultant. Business "super users" develop dashboards using the competency center's standard data sets and design guidelines. And before a dashboard is published to the production environment, the center must certify it.
Once the center recognizes that a data set is being used in more than one or two dashboards, it is incorporated into a standard data set. And with data governance in place to manage the self-service BI data gathered from BusinessObjects and from a Siebel customer relationship management system, Autodesk now is looking at incorporating "massive data sets -- 25 years' worth -- from Web traffic, product usage and text analytics," Booher said. "[Data consolidation] is the holy grail of CRM…a 360-degree view of the customer."
Preventing the creation of data silos from self-service BI
Shawn Spott, vice president of corporate intelligence at RBC Wealth Management USA (a 6,000-employee, Minneapolis-based division of RBC Capital Markets LLC), tends to be "very liberal" about allowing employees to offload data and take it with them. "In the many years that I've been doing [BI], I've yet to find a good way to prevent people from messing up the data," he admits.
To prevent data silos, Spott's team reviews reports and dashboards, and consolidates them when they are similar. "It doesn't speak necessarily to governance, but it's along the same lines, in that we're trying to go back and make sure that what we're representing to one group is the same data being represented to another group, regardless of what their up-front criteria may have been."
More about data quality strategies
Spott's team still controls who gets access to what in terms of permissions and the data fed into its self-service data analytics tool, the Web-based Tableau Server from Tableau Software Inc. That data comes from the company's mainframe, SQL Servers, a Teradata Corp. data warehouse and an SAS Enterprise BI Server from SAS Institute Inc.
Still, self-service BI has its booby traps. So-called spreadmarts, which are like spreadsheets but function like independent data marts, are a risk of self-service BI, said James Kobielus, a senior analyst at Forrester Research Inc. in Cambridge, Mass. "You can have chaos, you can have disparate, incompatible and inconsistent stores of data," he said.
To prevent data silos, users should build reports based on an official customer and financial record, Kobielus said. "Everyone needs to be working from cleansed data [that has been consolidated and merged and] managed by IT strictly and with specific access controls for users." Another factor to keep in mind is that IT should maintain version controls, so that when data is extracted from the master tables in the data warehouse, those are the official systems of record at that given point in time, he said.
Let us know what you think about the story; email Christina Torode, News Director.