"We deliberately did not put BI in technology, because we were worried they'd focus on technology and not the business problem," said Ed Macri, who heads the Boston-based company's BI and marketing divisions.
The first two years, they get tired and kill it. You need visionary leadership to get through the first two years.
Still, when big data reared its huge head, Wayfair realized its crackerjack BI team didn't have the skills to write the algorithms. "So, we have subsequently spun up a practice within IT of engineers," Macri said. Some of the data mining is still handled by the BI group, but "when it gets too tough, we hand it off to this IT group."
What the structure of a BI and analytics organization should be -- and how it relates to IT -- was one of the questions before a panel of IT professionals at a recent CIO gathering of the Boston chapter of the Society for Information Management (SIM). As outlined in our companion story, "CIOs show that the road to BI and analytics is tied to company persona," BI and analytics remains an elusive target for many businesses. Even though the discipline has been around for a couple of decades, it remains highly subject to company culture -- ingrained business processes, the attitudes of top executives and the idiosyncratic ways employees get their jobs done.
At companies where BI and analytics is relatively immature, the IT team, as keeper of the data, often does lead the way on analytics, said SIM panelist Paul Barth. "But ultimately, analytic skills belong in the business," said Barth, managing partner and founder of NewVantage Partners, a Boston consulting firm.
At companies with mature BI practices, the marketing departments are doing lead generation analytics, the finance departments are doing profitability analysis, and the product and service departments are optimizing business processes. "The technology gets very sophisticated," Barth said," but ultimately, analytics leaves IT."
Competing with 'screen-scraping factories'
That is not exactly the case at retail juggernaut TJX Companies Inc., where the architecture group has taken the lead on getting an enterprisewide BI and analytics practice off the ground. The Framingham, Mass.-based discounter fields a large IT department stocked with "BSLs," or business solution leads, who work closely with lines of business, said Jon Geggatt, vice president for information architecture and business intelligence at TJX.
"But those folks don't have the skills they need to introduce concepts of master data management, data architecture, and administration and governance of BI tools," Geggatt said. "We need to get our arms around master data, our vendors, our product information -- that is key for us."
The task is all the more difficult because the $23.2 billion retailer is growing so quickly. "Our mainframe store application is having a tough time keeping up," Geggatt said. An upcoming implementation of an Oracle-based site-management and master data platform should help. Meantime, even as CIO Kathy S. Lane has given the green light to building an enterprise data warehouse, Geggatt's business partners continue to take BI and analytics into their own hands.
"They are not dummies," Geggatt said, explaining how the business partners will write macros to bring up mainframe reports, scraping one column of information after another and throwing them into Excel spreadsheets or a Microsoft Access database. "We have a lot of screen-scraping factories out in the business," he said. When he reached out to the business, he was told these mini-factories were spending 70% of their time "muscling the data" and 30% of the time analyzing it. "I said, 'Let us in IT do the muscling, so you can spend your time analyzing it.' We're not there yet," he added.
Analytics: Have patience and throw in a wildcard
One reason BI and analytics practices are hard to establish at large companies is that they require patient investors, NewVantage's Barth said. The first people in don't see a quantifiable return. "They invest in cleaning up the mess, but it is hard to make money back on that," he said. The second person gets to use the data and analytics and adds their own. That provides some benefit to the first group, but often it is only in the third year that payback is apparent. "The first two years, they get tired and kill it," he added. "You need visionary leadership to get through the first two years".
Even in mature BI practices, the unpredictable results that are part and parcel of an analytics project compound the problem of ROI, said panelist Jim Whalen, CIO at Boston Properties Inc., one of the nation's largest owners and developers of Class A office space. He makes sure his business cases reflect that. "I have to put in a 30% wildcard -- an unknown factor -- because these things have a life of their own."
More about analytics and big data
Allowed to thrive, however, BI and analytics produce dividends well beyond the reports on past performance and even the real-time insight that analytics can provide, the panel agreed. Thomas Davenport, President's Distinguished Professor of Management and IT at Babson College, and moderator at the SIM event, recounted his recent conversation with the head of global analytics at Hewlett-Packard Co. That division focused first on developing expertise in reporting analytics, then moved on to predictive analytics; it currently provides 29 types of analytics as a service to the rest of the company. "What they are thinking about now is moving some of those into automation as products, so they can have more analytic services," he said.
NewVantage's Barth agreed that the goal of BI and analytics is to build analytics into enterprise business processes -- doing what he called "instrumenting" the enterprise. He offered two examples from the social-network game industry. Rather than spend $100 million on producing an online game end-to-end, a company will spend $10 million on the first chapter, publish it to players, then analyze the passive and active feedback to write the second chapter. Another company takes this a step further: It simply advertises the game on the Web, saying it does this and that, and waits to see what the click-through rate is before actually developing it. "It sounds horrible, but they say, 'Well, we get enough data so when we build it, the first iteration is right.'"
Let us know what you think about the story; email Linda Tucci, Senior News Writer.