olly - Fotolia

Self-service analytics needs strong data architecture foundation

The adoption of self-service BI and analytics tools continues to grow. But understanding how to support their use is a must -- and that starts with a solid architecture for analytics.

Consultancy Gartner Inc. expects a majority of businesses to put in place some kind of self-service business intelligence...

and data visualization software by the end of 2016. But simply implementing a set of self-service analytics tools isn't enough to fully realize the promise of opening up the BI process to more workers in an organization. It also requires lots of behind-the-scenes work to properly prepare data architectures so business users will be well-positioned to succeed with the new technologies.

For example, CDPQ, a Montreal-based financial services company that manages public-sector pension funds in the Canadian province of Quebec, began working in 2010 to beef up its data management architecture to give business users better access to data and equip them with self-service tools for analyzing and visualizing it. Before, data moved between systems through convoluted point-to-point connections, and the information needed to analyze investment portfolios and market trends was largely locked away in spreadsheets, according to Alexandre Synnett, CDPQ's vice president of data management.

"Everybody had their own spreadsheet," Synnett said during a session at the 2015 TDWI Executive Summit in Las Vegas. "It was really, 'My system, my data.' There was no way to manage it."

Synnett said CDPQ was about 80% of the way through the deployment of a layered architecture in which operational systems feed a set of data warehouses, which in turn pass subsets of information on to secondary warehouses and data marts. That has made data available throughout the organization, and users can manipulate and analyze it on their own through a combination of Tableau's self-service BI and data visualization tools and Microsoft's SharePoint and SQL Server Reporting Services software.

Measured approach on new architecture

Sébastien Dupré, director of information and analytics at CDPQ, said that to make the project more manageable, the company followed "an opportunistic deployment strategy" instead of trying to build out the architecture in one fell swoop. He added that the new setup adapts the principles of service-oriented architectures to the data layer, creating a portfolio of data services that reduce coupling between systems and simplify the extract, transform and load process for feeding data warehouses. CDPQ also built a Web-based data catalog to help users find what they need in its systems, Dupré said.

The ultimate goal is enable data scientists, business analysts and other users "to extract the most information they can out of data as quickly as possible," said Luc Veillette, CDPQ's senior director of modeling and business analytics. "For the business, we need answers now. The market is fixing the pace, so we have to give the best answer we can at the right time."

At LinkedIn Corp., supporting self-service analytics and data visualization involves taking extremely large data sets and whittling them down to digestible bites. "No one can use big data directly," said Michael Li, the social networking company's director of business analytics. "You have to make the data work in an application that's tailored for a specific purpose."

During another session at the TDWI conference, Li and Luke Baxter, director of insights at LinkedIn, described the company's development of a self-service sales analytics tool it calls Magic Wand. The tool provides a Web-based interface that lets LinkedIn's corporate sales reps view information about potential sales leads and the competitors of those companies. That helps inform discussions the sales team has with customers about how they're positioned on LinkedIn and what they could do to improve their ability to attract job candidates or to sell products through targeted ads on the site.

Boiling down data boosts analytics process

Running against an architecture that includes Hadoop and Teradata systems, Magic Wand produces reports that condense huge amounts of information from the LinkedIn network, customer data from the company's CRM system and external business data from Dun & Bradstreet into simple PowerPoint documents with visualizations generated with Tableau, Highcharts and other BI tools.

Li said the idea is to turn terabytes of raw data into kilobytes of relevant information at the end-user level -- and to do so in less than a minute, creating customer-specific presentations on demand. Baxter added that Magic Wand has become a core component of the sales process: The tool serves up 100,000 PowerPoints per year, about 80% of LinkedIn's sales reps use it on a quarterly basis, and it plays a role in 50% of all the new sales revenue the company brings in. "It's central to who we are," he said.

Mike Lampa, managing partner at consultancy Archipelago Information Strategies, said that self-service BI and data visualization tools have become more reliable and feature-rich -- and that enabling frontline workers to more effectively use data to drive business decisions is a must for organizations nowadays.

"We need to augment the decision process," Lampa said. "It's not to say that decisions made in the past were wrong. It's about having a higher level of confidence that we're doing the right thing now."

Ed Burns is site editor of SearchBusinessAnalytics. Email him at [email protected], and follow him on Twitter: @EdBurnsTT.

Executive editor Craig Stedman contributed to this story.

Next Steps

See where data governance comes in for unstructured data

Avoid these worst practices in data governance

Test your knowledge on improving data governance and stewardship

Dig Deeper on Self-service and collaborative business intelligence