-
-
Marketplace
-
Channel Resources
Articles from this Site
-
One brand, one Web site! DM Review is now the home of all the content you're used to at BIReview.com and much more. If you are registered at BIReview.com, you're already registered at DM Review. If not, take just a moment to sign up for all the free services we have for you at the new DMReview.com.
Good Science For Business
It's hard to know what to expect when you first meet Usama Fayyad. A former NASA scientist who has won a medal for space discoveries, a top expert in data mining - one of the most complex undertakings in business intelligence - Fayyad might quickly jump into mathematical esoterica beyond pedestrian understanding. He can do this if you wish but the chief data officer at Yahoo is also a SVP who manages people, and is equally interested to talk about turning business intelligence into value.
From NASA, Fayyad joined Microsoft where he built the data mining products for that company's server division. A couple of startups later, Fayyad's data-mining consultancy DMX Group found a new owner in Yahoo. By this time he had personally seen flaws common to most monolithic business intelligence projects. "There are all these things described as the basic infrastructure," he says. "You have to build the right layer of data, it needs to capture all this stuff and it's big scope, big hardware, big software and big consulting. I look at it as a scientist and say it's a failure because we don't have an engineering recipe that tells us how to build a warehouse or a BI system." Since there is no way to automate a business-centric BI function, Fayyad says organizations are left to explore blindly, grasp for people who understand data, and then try to tune that into the business. Making these projects ambitious doesn't make them better. "Most data warehouses are write-only, you put the data in and it rests in peace. The building of these huge stores is an engineering feat like the pyramids, people brag about their warehouse being bigger than someone else's when it's nothing more than a data tomb."
You'd expect an outfit like Yahoo - which processes something on the order of 10 terabytes of data per day - to actually require a monolithic approach but Fayyad didn't see it that way. "If you approach it with the attitude that you're going to solve all these anticipated problems down the road, you'll build something with amazing scope and you'll never hit your goal, you'll never finish. Even if you have a release, it's so late in the game that the business has changed on you by the time you released it."
Upon his arrival at Yahoo, Fayyad decided that for the first three to six months the warehousing and BI infrastructure would be treated as "a necessary evil" and not a goal. "I said, let's start with the two or three most important problems where the data can see action and move backwards." His first move was to change the group's name from Data Services to Strategic Data Solutions, a self-explanatory term to redefine and aim the mission at fundamental problems. For this he brought together teams of data experts and business analysts to form a human interface for what could not be automated or engineered. It was a shift from the old mindset, solutions for business users to leverage instead of systems to be grown and maintained.
The teams report to business unit and client groups and are measured by how well they understand the strategic day-to-day problems of the heads of those units. "The statement I make to a typical analyst at Yahoo is that the head of a business unit had a nightmare last night," Fayyad says. "By the time he wakes up you need to know what that nightmare was about." Under the analyst groups are engineering teams to interface on the data side. Fayyad acknowledges that the majority of these people don't think about the business, and he does not require them to. "We circumscribe their world by saying 'the analysts are your clients, satisfy them and you're doing fine.'" Supporting two teams does require some critical mass but at Yahoo Fayyad relies on this structure to encapsulate the process in a single entity.
When it comes to applications, the consultant in Fayyad tells him that the targeted approach is replicable for many companies. "Depending on the size and complexity of the problem, I think with between $100K and $300K you can build some very compelling applications where the payoffs can be huge." The approach sidesteps the enterprise data warehouse with data-driven applications that might rely on very ad hoc data marts, which provide the proper seeds of the warehouse down the road. Projects might address the top line or bottom line, but the important thing is that they uncover the value of data spewed out of operations for bookkeeping or logs. Bridging this gap first gets the whole organization smarter, gives hints to what's most important, and suggests a design guideline for the level of detail needed, all the things people today don't understand when they shoot for "the whole thing."
For those wishing to sell BI in the organization, the key is finding five important numbers - rather than 500 - that can have a profound impact on operations. "[As a consultant] my ideal meeting was to start with the chief marketing officer or CEO, somebody who worries about the business, doesn't think much about the data and just gets graded on how many more dollars he brings in," says Fayyad. "Now we could talk about how we can help that. The typical discussion doesn't talk about data, it talks about trying to find those levers and identifying one where you can see the relation to data."
At Yahoo, the guiding focus is introducing and implementing consumer-centricity, so Fayyad is always looking for more visibility and better information about the property and how people spend time on it. "We have great reports on ads served and impressions at this cost or that and all that. The notion of building something that can tell us how is the consumer experiencing the service, how are we communicating and the messages we are sending to them, that is a big culture shift in the view and the reporting of the data." Fayyad says he's lucky to work in an environment like Yahoo's where the levers and related payoffs are extremely sensitive. "This is a business where the numbers are huge yet you never touch the consumer directly. So data becomes the senses of the organization and putting transformation in place involves building the consumer-centric data platform and also having these outputs, the five numbers that are critical for every business unit that become part of the background." These KPIs are projected on walls across the company's headquarters, they are seen on login and they are second nature in the business of putting data to work at Yahoo.
Jim Ericson is editorial director of DM Review, a SourceMedia publication. You can reach him at Jim.Ericson@sourcemedia.com.
For more information on related topics, visit the following channels:
