Spreadsheets: big-picture views.

Spreadsheets have become so ubiquitous that it's hard to imagine how people handled large sets of numbers in pre-VisiCalc days. But it's apparently even harder to imagine how to extend and improve the classical spreadsheet metaphor. In the last decade, we've seen at most four innovations--integrated graphing and charting, macros, graphical displays, and three-dimensionality--that genuinely transformed the market's definition of what a mainstream spreadsheet should look like.

So what's left to invent? Lately, we've been looking at a group of rather exotic products--Improv (Lotus), Compete! (Computer Associates), SpreadBase (Objective Software), eSSbase (Arbor Software), and Hi-Sheet (BASEC)--that offer clues about where next-generation spreadsheets might be headed. We're not sure which of these products--if any--will survive the brutal realities of the spreadsheet marketplace, but the concepts behind these products are probably important enough to affect the direction of the mainstream spreadsheet market.

First of all, we'd argue that all of these products make sense primarily as tools for analyzing very large, very complex data models that usually end up on mainframes, not desktop PCs. (For example, consumer products companies like Procter & Gamble have to track the dynamics of sales promotions on market share for dozens of product lines across hundreds of local markets.) In theory, all of this data could be downloaded from mainframes and analyzed piecemeal, but that's a lot like trying to follow a football game by watching one player at a time. Sometimes there's no substitute for a big-picture view.

The neo-spreadsheets we've seen manage big-picture data with the help of two related innovations--multi-dimensional and hierarchical data architectures. Both models solve kinds of problems that are relatively painful for traditional spreadsheets, but both have some shortcomings as well. Our guess is that the long-term winners probably will be products like SpreadBase and eSSbase, which manage to combine elements of both architectures.

* The multi-dimensional model: Traditional spreadsheets have an almost unlimited capacity for building complex models--for example, economic forecasts, corporate budgets, and market analyses--that are based on elaborate data interdependencies. But these models tend to be restricted to a single perspective on the data, which is defined by the user's choice of rows, columns, and 3-D layers. Even though the worksheet may contain all the relevant data and variables, there are often questions the model can't answer without restructuring all the relationships.

To help explore this problem, Lotus began experimenting with a "multi-dimensional" spreadsheet called Improv, which was developed first for the next machine and is now promised for windows. Improv users can view data and relationships from a variety of views simply by switching row and column choices; for instance, an Improv user might look at weekly sales trends for a line of breakfast cereals by product type, box size, price, geography, supermarket chain, coupon program, sugar content--or any combination of these variables.

Improv solves a potentially important class of problem, but Lotus apparently hasn't hit on a solution for how to sell Improv itself. (The original developer of Compete!, another multidimensional spreadsheet, ran into the same problem and ended up selling the product to Computer Associates, which hasn't done much better.) Part of the problem, we suspect, is that developers of multi-dimensional spreadsheets tend to gloss over a critical question: Where does the worksheet data come from? Our breakfast cereal model only works if the user has access to literally hundreds of thousands of store-level data points, which don't just show up magically in a SQL file. To make this external data meaningful, users also need control over some fairly complicated consolidation and filtering processes..

* The hierarchical model: Control over external data is exactly the problem that hierarchical spreadsheet architectures help to solve. Unlike classical spreadsheets that rely on static linking and consolidating, a hierarchical spreadsheet understands that data rollups are usually dynamic. Companies add new products, realign sales territories, or decide to "drill down" to details that once were discarded. Thus, the three hierarchical products in this group are designed to manage big-picture data that exists outside a user's personal spreadsheet. Hi-Sheet (created by a group of Latvian developers) sits on top of a chain of Excel worksheets, managing lower-level linking and consolidation relationships. eSSbase (created by a group of mainframe veterans) puts large amounts of tagged data on an OS/2 server, which users can access through various client spreadsheets. And SpreadBase (created by a pair of ex-VisiCorp marketers) applies category-based parent-and-child logic to external data files.

For the moment, the big-picture spreadsheet category consists almost entirely of vaporware: The only shipping products are the NeXT version of Improv and CA-Compete! Everything else is still in various stages of testing or in limited distribution. But the spreadsheet giants are already reacting. Borland and Microsoft have promised that they'll find ways to integrate multi-dimensional and hierarchical data models into existing products, and Lotus seems reasonably serious about turning Improv into a mass market product. So big-picture data management may start showing up in mainstream spreadsheets surprisingly fast.

James Dorrian, president, Arbor Software Corp., 3211 Scott Blvd., Santa Clara, Calif. 95054; 408/727-7166. Ron Scott, president, BASEC (Baltic American Software Corp.), 1 Constitution Plaza, Boston, Mass. 02109; 617/242-4444. Marc Sokol, director of product strategy, Computer Associates, One Computer Associates Plaza, Islandia, N.Y. 117887000; 516/342-5224. Jeffrey Anderholm, group manager/Improv Product Line, Lotus Development Corp., One Canal Park, Cambridge, Mass. 02141; 617/693-7733. Richard Melmon, president, Objective Software, 248 Homer Ave., Palo Alto, Calif. 94301; 415/324-3333.