4 min read

New Directions For OLAP

As OLAP reaches its second decade, it's being extended beyond data to text and geospatial information.

The online analytical processing tool arena is now in its second decade, and the focus of business intelligence innovation has shifted to dashboard displays, applications such as performance management, and operationally embedded analytics. That's a hallmark of the acceptance of multidimensional data models, pivot tables, "slice-and-dice" analysis and interactive "online" data exploration as essential ingredients of modern business intelligence programs. It's this acceptance that has enabled the most exciting innovation in the OLAP category.

A number of vendors are extending the OLAP paradigm in new directions. The goal is the same as always: to help users detect patterns in data that explain and predict. The thesis is that visual presentations — charts, maps and diagrams — help users gain insight that doesn't exactly jump out of tables, canned reports and static charts. These vendors are promoting new ways of interacting with numerical data, and they're extending the OLAP model to both text and geospatial information.

Multidimensional Views

Tableau, for example, is a tool that adds interactive visualization to familiar pivot tables. It can generate tables from standard structured OLAP and relational data sources. But Tableau doesn't just present data tables; it also lets you embed charts within the cells, with each chart presenting a multidimensional data view. Like other trellis displays, the resulting grid of charts can boost the user's ability to identify trends, but unlike other trellises I've seen, Tableau's display is interactive. You can drop a dimensional variable into one of the table or charting axes, or into a "shelf" for rendering using different chart symbol shapes, sizes and colors. Manipulating the hybrid table-chart is as easy as working with the pivot interfaces offered by many tools that present only numbers, though maybe with a bit of color coding added. The result is that you can work productively with a higher level of detail than is feasible with conventional data tables, and can explore the data in ways that respond to our visual pattern-recognition abilities. (If you're having trouble getting the picture, read Stephen Few's "Data Analysis At the Speed of Thought".)

The JMap Spatial OLAP extension from Kheops Technologies gains its edge by delivering multidimensional data through interactive maps (in addition to tables and diagrams). You can aggregate and drill down on cartographic elements — not only geographic areas defined by polygons, but also on named places and features such as roads — that are associated with conventional categorical dimensions. Kheops' JMap software provides Web map navigation and manipulation with a modular architecture that lets Spatial OLAP do the rest, including the usual OLAP crosstab composition, drilling operations and calculated measures. And like Tableau, Spatial OLAP overlays the display with data-derived charts, as well as symbols and lines of various shapes, types and colors, albeit on a map rather than on a data table.

Text OLAP is a third hybrid that builds on the OLAP paradigm, and it's just one analytic engine within Megaputer's PolyAnalyst suite of data and text mining tools. In a departure from typical text mining, Text OLAP offers interactive, structured analysis of textual documents. It uses extracted terms and concepts ("entities") as dimensions. Computed quantities ("measures") displayed in table cells indicate the strength of the relationships among entities rather than quantities such as the sales figures you'd study with a conventional OLAP tool. Text OLAP provides a matrix display of entities with linked charts and a color-marked display of relevant extracts from the reference texts. Because we're working with text, you also get a rich and integrated pattern-definition language, but the overall feel of the Text OLAP interface will be immediately recognized by any OLAP user.

The products I've cited are great examples of cross-pollination. They apply familiar, tried-and-true OLAP techniques to problems where few available solutions offer OLAP's accessible interactivity. OLAP's structured dimensional models have proven their value for sophisticated end users, and these hybrid approaches that meld OLAP with charting, mapping and text analysis promise an even greater analytic advantage. Let's hope the larger analytics market sees it that way, too.

Seth Grimes is a principal of Alta Plana Corp., a Washington, D.C.-based consultancy specializing in large-scale analytic computing systems. Write to him at [email protected].