RANDOM THOUGHTS. . .

on information productivity

Regular if infrequent musings on the critical issues involved in making information and information systems truly productive. . .from the professional staff of INFORMATION STRATEGIES, INC

FROM THE EDITOR: In today's brave new world of information, there are few subjects that cause more consternation than what to do with TABLES. This issue of RT looks at one possible approach to tables, both as a historically important device in the dissemination of information, and as a challenge for the electronic delivery community. BAS

RE-THINKING TABLES FOR AN ELECTRONIC ENVIRONMENT

copyright, September 15, 1993 by Barry Schaeffer

INTRODUCTION:

Tabular representation of data has been with us for so long that it is almost part of our way of looking at the world. As published information has proliferated, the use of tables to compress large amounts of reference information in small spaces has matured, with each industry or cultural group developing its own conventions. Tables covering hundreds of pages or representing data in tiny type are not uncommon as publishers have striven to make data quickly and economically available to their users. The tabular format grew out of two seemingly antagonistic characteristics of publishing:

1. Particular types of data in which users need only small, often random pieces at a time from a sometimes large potential pool. In such cases, users need to find the specific piece they want quickly and based on a simple, easily-learned formula for location.


2. The nearly exclusive use of the page for display of data. This fixed rectangular format made location of data on the basis of simple triangulation the most effective way of satisfying user needs.

Tables, then, are a way of placing data on a rectangular page, in a highly compressed format, with a method for locating any piece of data based on the intersection of two perpendicular lines. The use of rules separating rows and columns (cells), complex stacked column heads and other graphic devices have been developed to speed and simplify the user's task in finding what he or she needs.

Over time, these visual characteristics have become part of our collective consciousness, and tables have virtually become an art form. Indeed, an entire wing of computer publishing has grown up with the sole aim of formatting complex tables from coded data.

With this elevation from means-to-end to icon, tables have ceased to be thought of for what they are; a graphic device to make available data which has value in isolated bits and pieces. Indeed, the very nature of the information needs from which tables sprang has faded from most consideration of the finished tabular product.

Now we face a major change in one of the foundations on which tabular information display was built. The movement to electronic delivery of information, with all the potential benefits it offers, has one glaring deficiency when compared to paper pages: a screen simply cannot display as much data in as small a space as can a paper page. With a single stroke of technology, the entire tabular world has been forever degraded (it isn't likely that any screen at any price will equal the static display capabilities of paper any time soon.) This degradation has forced information publishers to begin a rethinking of tables, often painfully re-editing their existing tabular data for display within the constraints of whatever electronic display devices they have selected. There may be more tedious, time-consuming tasks but none come instantly to mind.

While this response to electronic delivery has allowed the transfer of paper published data into new electronic forms, it must be viewed as a stop-gap. Indeed, there may be no effective way to redesign tables for truly effective use in an electronic environment.

RE-NEGOTIATING THE TABULAR INFORMATION AGREEMENT:

While most of the publishing world is scrambling to rework tabular data for electronic display, the real answer may lie in another direction altogether. Tables, after all, were a culturally "negotiated" solution to a particular type of information need. The negotiation, spanning many decades somewhere in the distant past, included factors such as the random nature of data use, the cost and limitations of paper publishing and the ability of users to learn an employ the two-dimensional triangulation method of data location.

Perhaps what the publishing industry and its clients are facing is not the need to redesign tables but the need to renegotiate the solution to that original information need. While the nature of the information vehicles may have changed, the information need probably has not. This renegotiation must be broadly based on three factors:

1. the video terminal's degraded ability to display data.
2. the computer's enhanced ability to store and apply more complex data location methods.
3. the current level of sophistication in the user population (as compared to the 19th century when tables came into use.)

Within these new boundaries, the publishing industry must "ask" its user population the same questions that prompted the design of tables in the first place; questions like:

• Given a large amount of data not organized for linear use (as with narrative), what patterns of usage do you usually follow?
• How much of the total data are you willing to wade through linearly in order to find the piece you want?
• What characteristics of the total data are you prepared to use in locating the portions you want?

The answers are likely to be very different this time, in ways that cannot be inferred by merely looking at tables as they are now used.

A PLAN FOR DELIVERING "TABULAR" DATA IN ELECTRONIC FORM:

The following thoughts grow out of the author's admittedly incomplete ruminations on the possible results of the renegotiation described in the previous section. They are presented in the hope of widening the debate over what is to be done with tables in the brave new world of electronic information delivery:

• Given a large amount of data not organized for linear use (as with narrative), what patterns of usage do you usually follow? It appears that the users of tables are nearly always looking for a small number of discrete information pieces associated with some broader category. The organization of the table permits users to select data based on two "locator" categories, the row stub and column head, then find the intersection of the selected row and column (usually called a cell) to read the desired data. While some tables utilize spanned heads or "mini-tables" within cells to represent more complex data, the basic "find and use" pattern remains the same.


• How much of the total data are you willing to wade through linearly in order to find the piece you want? An oft-forgotten characteristic of paper books is their rapid "browse rate." The average human reader can flip several pages per second and still locate desired information, paying what might be called a low penalty for linear search of the data. This is why multi-page tables aren't viewed as unduly burdensome.In an electronic environment, however, especially one based on the relatively slow access rates of CD-ROM, linear browsing carries a much high penalty, so much higher that users often view products which require it as unsatisfactory. This means that users of tabular data must be able to quickly identify and find data they seek without laboriously browsing through many screens of a displayed table to find it.


• What characteristics of the total data are you prepared to use in locating the portions you want?Users have become comfortable locating data through two characteristics (stub and column). In developing a locating procedure for electronic data, we are on relatively firm ground so long as our basic method is not much more complex than this. In a paper table, the user may easily peruse all of the stub and column head values to aid in determining how to find desired data. The capacity of paper permits these finding aids to be quite complete and descriptive, often comprising almost a guide to the table.
  
  In the equivalent electronic display, the display software may have to generate and display a "finding aid" visual upon a user's entry into the data. In its most crude form, this display would be a single-screen listing all stubs and column heads (minimal scrolling might be necessary for view all values.) Having made a selection of data class to be retrieved, the user would be permitted to indicate the row and column values to be applied in displaying the desired data.
  
  This, of course, means that certain complex data would never be displayed electronically in tabular form. Indeed, no attempt would be made to electronically display all of the available data at one time, in any format. This may be a difficult cultural change for many people, especially those who have labored so long and hard to perfect tabular composition. Nevertheless, times are changing and the medium simply can't support the visual elegance that is an important part of their work.

THE INTERACTIVE LIST:

If we accept that a new way of presenting and finding tabular data is needed, we must look also at the way in which we tag and store the source material for processing and presentation. Tabular data coding, after all, was developed primarily for processing by paper composition systems and may not be capable of efficiently supporting the new processing. The concept of an interactive list was developed out of ISI's perception of a target functionality for tabular data in a CD-ROM or on-line service delivery environment. It aims to provide four main processing supports:

1. Display a virtual "table of contents" upon entry.
2. Allow the user to select categories (row and column) identifying the desired data.
3. Jump immediately to the target data without linear browsing but be able to browse the data around the target if desired.
4. Provide an easy path to classical tabular composition for paper-based output.

The following sample data and display approaches are consistent with these goals. The sample display could be generated by most commonly-available display software packages, either directly from the data sample or with a simple conversion. For paper-based tabular composition, most mid-to-high level composition systems could generate a standard tabular format from the data sample, directly or with a simple data format conversion.

From the preceding data, one might then generate the following data display, allowing the user to select the categories and specific elements desired, in a fashion similar to finding the desired stub and head of a table. The exact nature of the resulting data display as well as any advanced capabilities would be the option of the designer. Suffice it to say, however, that this method of determining and accessing data now displayed in tabular form has some distinct advantages in an electronic delivery environment.