TORONTO – It’s the kind of question that only a good business intelligence system, drawing from a solid data warehouse, can answer: How can an airline know with reasonable certainty whether passengers will actually get on flights for which they’ve booked tickets?
According to Teradata CTO Stephen Brobst, who posed it to the audience at the recent Think Big Data Warehousing conference, it has nothing to do with the distance between the passenger’s home and the airport. “It’s whether or not they ordered a vegetarian meal,” he said. “If they ordered vegetarian, they are going to be on that plane.”
Though it was an isolated example, Brobst said it showed the kind of value enterprises get when they create the kind of data warehousing architecture that will bring more value to the information sorted by business intelligence (BI) software. The problem, he said, is that many organizations don’t spend much time thinking about how to connect data warehouses, data marts and other parts of such a system. They focus on the “fun stuff,” he said, like buying the technology, and then try to figure them out later.
“You need a service level agreement for a data warehouse,” Brobst said. “I’m horrified at the number of companies that don’t have anything written down about it, let alone measuring it.”
Brobst and others at the Think Big event offered their own reference architectures that they said enterprise IT managers could customize to meet their needs. This included a model called the Corporate Information Factory, described by Claudia Imhoff of consulting firm Information Solutions Inc. Without a conceptual blueprint, she said, it will be difficult for companies to justify the costs of implementing BI and data warehouse products.
“I’m from North Carolina, and we have a saying that if you’re lost in the woods, any path will do,” she said. “You have to have an architecture so that you understand where the data flows between components.”
Toronto-based Adastra Corp., which hosted the Think Big conference, offered several examples of architectural options, which were graded according to an “information access scorecard” by the Bank of Montreal’s senior platform development manager, Richard Livesley. These included tactical data marts, which might be used by individual business departments to pull data from source systems and extract from analytical information that’s relevant to their task.
Nelio Lucas, Adastra’s chief architect, said many firms that use isolated data marts quickly realize the limitations of such an approach, including inconsistent information from one department to another. The next step up is pushing data through an enterprise data warehouse to the data marts, but even here there are roadblocks, Lucas said.
“You have to spend more attention to managing change every time the source data changes,” he said.
Livesley said BMO is still at this stage. “We have a lot of silos of expert BI. We’re not about to claim that we’re managing information as an enterprise asset.”
BMO’s board of directors, however, has created policies to support further development of its data warehousing and BI strategy, said Livesley, who admitted that cultural barriers continue to crop up in his organization.
Think Big sessions also looked at centralized data warehouses and federated data marts, but the best practice architectures were often a complex, highly sophisticated map of components. Brobst said his reference architecture put the data warehouse off to the side, as opposed to in the middle of the blueprint. “We always think we’re the centre of the universe,” he said, “but it’s the middleware connecting all these pieces together that’s important.”