With Kalido’s Drag-and-Drop Data Warehouse Customization, “Business Intelligence” Is No Longer an Oxymoron
I’d like to explain what’s cool about Kalido, a Burlington, MA, software company spun off five years ago by Royal Dutch Shell, but let’s start with a story about beer.
Labatt Breweries is Canada’s largest beer producer, brewing 60 brands of ale and distributing them in stores, bars, and restaurants across the great North. Because Labatt has bought up so many regional brewers over the years, and because every province in Canada regulates the sale and taxation of alcohol differently, the company has inherited dozens of different sales databases and reporting systems, making it extremely difficult to figure out exactly which Canadians are drinking what kinds of beer.
Typically, big companies get a grip on their sales data, customer information, departmental budgets, and the like by buying a “data warehouse” where all of the most important information from lower-level transaction systems can be organized for fast retrieval; then they use “business intelligence” tools to transform the information in the data warehouse into charts and graphs that give historical, current, or predictive views. Labatt was deep into all of these technologies. It had enterprise resource planning software from Oracle, customer relationship management software from Seibel and SAP, data integration tools from Informatica, and business intelligence tools from Cognos. But when Labatt itself became part of Belgian beverage conglomerate InBev a few years ago—meaning that it would have to start feeding business performance data up the chain to managers in Europe—the company realized that its various divisions were still counting the same things different ways.
That’s when Labatt turned to Kalido, which is the product of a similar but even more severe crisis at Shell. During the 1990s, the petroleum giant acquired some 75 companies using 110 different reporting systems, according to Kalido CEO Bill Hewitt. “The guys in Beijing would call a brand of oil ‘A’ and the guys in Cincinatti would call the same product ‘B’, and the process of rationalizing the data every year was taking far too long,” says Hewitt. “So they built a data warehouse and generated reports off that—but the problem was that every time they wanted a different report or brought in a new company they’d have to rebuild the data warehouse, which would take months. Finally they ended up separating the rules that governed the data warehouse from the data itself. Instead of building a new warehouse, they built a business model”—a graphical representation of Shell’s business, with active links back to the raw data.
With this new interactive tool, the graphical model didn’t simply depict but actually controlled the way data was piped between the company’s low-level operational systems and its business intelligence software. So adding a new subsidiary’s data to the mix would be almost as simple as drawing a new box in the org chart and connecting it to the proper parent divisions. Other changes in the organization could be replicated in the model simply by dragging boxes around and redrawing the arrows between them (more details on that below). And not only that, but because the data was now tied to a flexible model, analysts could store snapshots of the model representing different points in time, then run the clock backward and forward to examine different what-if scenarios.
Back to Labatt. In 2005, the company hired Kalido to build a new data warehouse that would bring together data from 93 transaction systems databases holding 16 years’ worth of records on products, customers, suppliers, employees, and so on. A real-life test of the Kalido installation and its flexibility came almost as soon as it was completed, when Labatt acquired yet another brewery. Labatt’s field sales staff wanted to integrate the new company’s product information into the sales database right away, so they could see how Labatt’s new sales totals would look. But the company’s accountants didn’t want to integrate the new information until the end of the fiscal year, fearing it would mess up the books. “In a traditional data warehouse you’d never be able to have it both ways,” says Hewitt. “But because we have the ability to recreate a view at any point in time, the data warehouse managers could give the sales people the view they wanted and still wait until the end of the year to consolidate the financial information.”
Okay, you may have a hard time getting excited about a new data modeling tool that reconciles incompatibilities in a company’s database infrastructure and brings out the full potential of business intelligence software. And admittedly, developments in enterprise software can seem deathly dull to outsiders who aren’t enmeshed in the complexities of running a big business. But if you follow the larger world of software engineering, then you may appreciate the novelty and power of Kalido’s idea, which builds on more than a decade of R&D on graphical models that define and drive underlying data structures.
The linking of graphical models to actual software code and data is the core concept behind the Unified Modeling Language first developed in the early 1990s by Rational Software, which went on to be acquired by IBM and is now at the epicenter of Big Blue’s software development platforms business. Suffice it to say, business data modeling is an idea that’s picking up steam, and that’s likely to become more important as … Next Page »