Despite a continued push from vendors, self-service business intelligence (BI) isn't making headway at most organizations, according to a recent survey by Forrester Research.
Of the 82 IT "decision
That's surely a much smaller percentage than vendors like Microsoft and SAP Business Objects -- both of which have made extending BI capabilities to non-power users a priority -- would like to see.
In October, Microsoft announced plans to release a set of managed self-service analytics and reporting tools with the next upgrade of SQL Server in the first half of 2010.
In August, Business Objects debuted Xcelsius Present, an Excel transformation tool that lets users convert spreadsheet data into one of 10 preconfigured visualization templates, including charts and graphs, via a point-and-click interface.
IBM Cognos and Oracle also offer self-service tools as part of their BI suites.
Holding back self-service BI adoption, then, is not a lack of vendor tools but the rigidity of underlying data models at most organizations, limiting the types of self-service analysis users can perform, according to Forrester's Boris Evelson, who analyzed the survey findings.
Data models define relationships between entities, such as products and prices, or customers and products that support and allow more complex BI analytics and reporting. End users can create reports and perform analysis only based on the connections and definitions of the underlying data model, Evelson said.
If a data model says one customer always buys one product, an end user would be unable -- or it would at least prove too difficult for most non-power users -- to analyze customers that buy multiple products, for example. "It's now a many-to-many relationship," Evelson said, "rather than a one-to-one relationship."
"[So any self-service]BI tool is only as good or gives you only as good an answer as the underlying multidimensional data model," he said.
Adding to the difficulty, complex data models are not easy to change to meet self-service user demand. In a traditional row-based database with millions of rows of product data, for instance, changing the underlying data model "can take days or weeks," according to Evelson.
The increasing adoption of columnar databases could change that, however. As the name suggests, columnar databases store data in columns, rather than in rows. Individual data elements, like customer name, can be accessed in columns as a group, rather than individually row-by-row.
As a result, changes to data models in columnar databases are much easier and faster to make than in row-based databases, Evelson said. And the easier it is to change data models, the more useful self-service BI tools become.
A number of solid columnar databases are currently available on the market, he said, including Sybase's IQ Analytics Server and Vertica's Analytic Database. And recent research from The Data Warehousing Institute found "increasing adoption of columnar databases as the platform of choice for data warehousing."
Likewise, as columnar databases for analytics become more popular, making it easier to change data models, self-service BI tools will also become more attractive, Evelson predicted.