kentoh - Fotolia
Several business intelligence managers who spoke at the 2016 TDWI Executive Summit in Las Vegas this week offered advice born of experience to their peers: Stop worrying and learn to love self-service BI technology.
Deployments of self-service software are often end runs around central IT and BI teams by business units with unmet analytical needs. That frequently leads to disjointed analytics efforts -- and considerable ill will between the two sides. But in companies tired of the conflicts, self-service BI has come in from the cold and been accepted into the IT fold.
At software vendor SAP's Concur subsidiary, for example, data analysis now centers on officially sanctioned and fully supported self-service BI tools. Over the past two years, the developer of travel and expense applications has rebooted its internal BI program through a combination of Tableau's BI and data visualization software, Agile development processes, and end-user outreach and community-building initiatives, said David Han, senior manager of BI at Concur.
Han, who joined Concur in April 2014, eight months before it was acquired by SAP, said a new approach was definitely needed. He cited a laundry list of problems and pain points preceding the changes. Among them: The BI team was slow in developing requested functionality, partly because of a lack of skills on the IBM Cognos software that the company had previously adopted as its BI platform. That often forced business analysts and other users to use Excel spreadsheets to analyze data, which led to different business units reporting "wildly different numbers" on the same metrics, according to Han.
As a result, only a limited amount of data analysis was being done -- and the two sides were at loggerheads with one another. "It was like a food fight," Han said. "It was really tense in meetings."
A better way of doing BI, business
Things have improved significantly thanks to the new self-service BI technology and development processes, he added. The number of active BI users has increased from about 200 to more than 1,000, with nearly 900 of them using the Tableau software. Han said the effort has also produced some "quick wins" to demonstrate the business value of the new approach. For example, Concur attributed about $10 million in added revenue to a BI dashboard that combined billing and customer relationship management data, improving the up-selling and cross-selling capabilities of sales reps.
Chip giant Intel's sales and marketing operations encountered the same kinds of problems that Concur did in their first take on a BI program. David Schaefer, chief BI architect in Intel's sales and marketing IT group, said it initially tried to keep a tight lid on data, controlling the entire process of creating reports and dashboards. But that created long backlogs of BI to-do items: It typically took the IT unit six months to deliver on even small requests for information from business users.
Impatient and frustrated users came to view the IT group as "layers and layers of bureaucracy," Schaefer said, recalling that one business analyst memorably described various teams set up within IT to manage data and vet BI requests as "powdered wig societies." Increasingly, business units responded by circumventing IT and using Excel or new BI tools that they deployed themselves to directly access the company's data warehouse, resulting in localized spreadmarts containing inconsistent data.
Data doors open wide for self-service BI
Intel began changing its ways in 2008, according to Schaefer. His team adopted Agile BI methodologies, and set up a new sales and marketing data mart with what he described as a "lean data governance" model. And since then, he said, the IT group has opened up access to the data mart to any self-service BI technology that business analysts and other users want to work with. IT officially supports two self-service tools, which he declined to identify -- but business units can deploy and manage other ones themselves, or pay the IT group a chargeback fee to support the software for them.
"We used to try to enforce what tools were used, but we decided not to anymore," Schaefer said. "We told the users to pick whatever tool they wanted to use. We're just trying to embrace the change."
David Schaeferchief BI architect in Intel's sales and marketing IT group
On the back end, the IT group ratcheted things up in early 2014, deploying a new Hadoop cluster to host the data mart. The cluster, which is based on Cloudera's Hadoop distribution, supports different levels of data preparation and cleanliness, giving users additional options for analyzing information. Raw data is processed and made available for self-service BI access within 24 hours, Schaefer said. Cleansed versions of data sets become available within three days, and fully conformed and consistent data within two months.
Internal tracking at Intel has shown that more than 75% of the business analysts in sales and marketing are now actively using self-service BI technology, and that nearly 85% of all queries are being run against data in the Hadoop cluster and an in-memory analytics platform that the IT group also supports. "And it's not just people running something once and then going away," Schaefer said. "They're running things over and over."
Read our guide to deploying self-service BI tools and data discovery software
Consultant David A. Teich on IT's continuing role in self-service BI initiatives
Gartner's take on the shift toward self-service software in BI technology