Getting Good BI Without a Single Version of the Truth
A single version of the truth has long been a business intelligence goal. But should it be?
Remember that scene from "A Few Good Men" where Tom Cruise wants the truth and Jack Nicholson tells him he can’t handle it? The same could be said about business intelligence (BI), according to Forrester Research analyst Boris Evelson.
"BI has overestimated the need for a single version of the truth for decades," he said. "If it costs far more to get a single version of the truth, maybe it's wiser to take a cheaper version which is 80 percent good."
According to Forrester’s data, 44 percent of business intelligence users complain of not having all the data they need. Only about 25 percent of unstructured data and 35 percent of structured data is currently available within enterprises for analysis by BI and analytic systems, Evelson said. Enterprise business intelligence systems are often siloed, and further organizational insight might exist in spreadsheet form. This places major limitations on analysis accuracy and ability to get a single version of "truth."
Traditional information siloes can be quite agile, Evelson pointed out. Data can be analyzed right away, although it offers only a limited view of the truth. While systems of record such as ERP were an attempt to gain a single version of the truth, they didn’t succeed. In fact, such systems can slow analysis down. As a result, said Evelson, people gravitated back to their own silos and spreadsheets.
Hadoop and Hub-and-Spoke Architecture
The transition toward Big Data is one part of the solution, he said. "Hadoop is great for sifting insight from huge piles of data but not as good for immediate queries. You have to set up faster data query sources."
Forrester envisions greater agility from so-called systems of insight that tie everything together. Forrester believes this is best achieved via a hub-and-spoke architecture, with an uber-portal lying on top of underlying business intelligence systems so the user doesn’t need to enter each one separately. This would entail different policies for customer-facing and non-customer facing BI applications. The dashboard available to users would be able to access various parts of the business intelligence network.
"You need some systems set up on more expensive and fast query systems, and then use Hadoop for large amounts of data to be analyzed cheaply but not as fast," Evelson explained.
Business Intelligence and 'Certified' Data
Wayne Eckerson, an analyst with Eckerson Group, is another who believes business intelligence has had trouble handling the truth. He said the pendulum swung too far away from distributed BI into centralization, and that it is likely to swing back again. Standardization vs. speed is a key challenge.
He believes business intelligence should be able to deliver certified data and reports – a validated version of the truth in other words. Rather than trying to capture and analyze and take responsibility for everything, he sees a future where BI establishes a brand as the certifier of truth across a particular swathe of data. That might only be one data warehouse, or it might encompass a wider zone which includes Hadoop.
"Management will see that seal and know that data is trustworthy," he said.
How about the problem of business units setting up their own tools to analyze data? Eckerson said that preventing them is not likely to work, so the best approach is to offer them help and do so using standardized systems that offer some degree of integration with other enterprise systems.
"Rather than saying everything goes through me, help them use tools and set them up correctly so you stay in control," he said.
Like Evelson, he favors a federated approach with multiple tools, including self-service options set up for senior business analysts with more technical expertise.
"The key to being agile is getting to know the business from the people who live the problems and know what is needed," Eckerson said. "Get ahead of the demands of the business by trying to recruit people with deep business knowledge onto your BI team."
One Domain at a Time
Jake Freivald, vice president of Product Marketing at Information Builders, sees things transitioning from people using tools to the establishment of truly self-service business intelligence.
"We’ve moved from Excel hell to Excel hell with a pretty face," said Freivald. "Problems with data quality are starting to surface, and they want true data consistently. But the BI systems can’t handle the truth."
To achieve the fabled single version of the truth, it is best to focus on one particular area rather than trying to make everything accurate across the enterprise. Freivald gave the example of a health care provider which could choose one domain and go for achieving a single version of truth within that limited zone that is considered of most value.
He thinks the old bottom-up approach – in which you install something like Oracle and add BI apps to try to gain the insight you desire -- was faulty. That’s backward, Freivald said.
"Don’t try and model the entire enterprise; instead model domains," he said. "Ask questions about one domain at a time and build out from there."
Align Business Intelligence to the Business
Cindi Howson, an analyst at Gartner, quoted another Tom Cruise line, "Show me the money."
"Most organizations use BI to improve efficiency, but that is not enough," she said. "They should be tracking hard business benefits like money and profits, and aligning BI to that."
Lyndsay Wise, an analyst with Enterprise Management Associates, echoes that sentiment. Her advice is to tie business intelligence to a specific business goal to ensure it delivers measurable value.
"You can be vague in your implementations or try to do too much too soon and not be focused enough," she said. "You end up with the scope becoming expanded and more requirements being added. It’s best to figure out a goal and limit your scope to tackling that one thing."
Drew Robb is a freelance writer specializing in technology and engineering. Currently living in Florida, he is originally from Scotland, where he received a degree in geology and geography from the University of Strathclyde. He is the author of Server Disk Management in a Windows Environment (CRC Press).