A Special Report

The Basics

Technology, Economics and Ego Conspire To Make Software Difficult to Use

By William M. Bulkeley
The Wall Street Journal

May 20, 1991

YOU'RE ON THE ROAD. A friend has lent you his desk so you can dash off a letter. You sit down at his computer and face a screen that is bare but for the letter "C" and an arrow.

This shouldn't be difficult. You use word-processor software in your own office. But you can't even find his word processor. You type "dir," and an unfathomable list of names rolls up the screen. Like a stranded motorist shaking the spark-plug wires under the hood, you tap a few keys. The computer responds: "Bad command or file name."

Eventually, your friend sets you up in the word processor. But your troubles have barely begun. Hitting the F3 key on your own computer gets you electronic help. You hit F3 on this computer and it says "no define block." Haplessly, you start to type. But you can't make the words appear on the main part of the screen -- just along the top line. Your chest tightens, your fingers clench. You are ready to scream.

Why is software so hard?

That same question vexes practically every software user -- from the first-time technophobe to the code junkies who program deep into the night, fueled by pizza and candy bars. Mitchell Kapor, founder of Lotus Development Corp., and designer of the 1-2-3 spreadsheet, has said that "not a day goes by that I don't want to throw my computer out the window."

Some software designers insist that's changing. They talk about radical software designs that will simplify programs, and about new easy-to-use computers that are controlled by pens. But many software experts are unconvinced. They say a combination of economic factors, technical difficulties and the creative pride of software designers conspire to make software hard. Software, they say, will remain difficult for at least another decade, until computers understand spoken instructions.

If so, computer and software makers are missing out on a lot of money. The industry estimates that about 60 million computers are in use around the world. But that means that about three-quarters of the workers and homeowners who could use computers don't have them, says David Liddle, chairman of software designer Metaphor Computer Systems. Mr. Liddle thinks that the computer business has stalled in part because software developers continually add more complex features to the same basic programs. "We've gone from the baroque to the rococo," he says.

Proponents of easy-to-use software like to draw analogies to the U.S. automobile industry, which exploded after Cadillac introduced the electric starter in its 1912 model. Starting a car was suddenly acceptable for women, doubling the potential market. Similarly, simple software might encourage many users to buy computers for the first time. And it would lead current users to buy more software packages.

It isn't that software developers haven't tried to make their programs easier. The search for the better mouse has been an obsession throughout the short history of computer programming. For a while, Hewlett-Packard Inc. touted the touch-screen, which could detect a finger pressed against the tube. Users complained of smeared screens and tired arms. One developer came up with a foot-mouse as an alternative to the usual hand-held version: It was rolled around under a desk to control the cursor on a screen. Users were totally baffled. Another firm came up with the head-mouse, a headset that let a user control the screen by moving his head. Users snickered.

To some extent, the computer is bound to be complex. "It's a machine with 101 buttons on it. That's complicated," says Daniel Bricklin, co-developer of Visicalc, the first spreadsheet. Mr. Bricklin is now writing software for computers that are controlled by pens rather than keyboards. Michael Wilding, president of Productivity Point International, a Hinsdale, Ill., computer-training firm, says software is "becoming easier to use." Nevertheless, it will remain difficult because "software is so powerful. There are so many things you can do."

Compounding the difficulty is that for most software, the education never ends. Computer users don't mind learning how to compute; after all, they once took the time to learn how to type or drive. The problem with software is that most knowledge can't be transferred to the next program. Companies often find that training a user on a new software package takes a week of classes and costs $500. "People don't mind spending time on the first application," says Howard Elias, vice president, computer merchandising, for Tandy Corp. "They resent it on later applications."

W.E. "Pete" Peterson, executive vice president of WordPerfect Corp., the Orem, Utah, maker of the leading word-processor program, says: "When you switch programs, sometimes you get digital paralysis," fearing that finger motions that open a file in one program will clear a screen in another.

Apple Computer Inc.'s Macintosh forces developers to use consistent commands for many different programs. But in the International Business Machines-compatible world, no such strictures exist. Indeed, when competitors designed commands to mimic Lotus's popular 1-2-3 spreadsheet, Lotus sued for copyright infringement.

Even with consistent commands, individual programs are difficult, because developers have concluded that ease of use doesn't sell well. "We could make a word processor very easy to use if it just did what a typewriter does," says Mr. Bricklin, the software developer. But by adding spelling checkers, automatic footnotes, boldface, italics, and the ability to move and erase blocks of texts, the program gets more complex. "There's a level of difficulty we're willing to tolerate to get all that," he says.

"Ease of use is overrated," says Gordon Eubanks, chairman of Symantec in Cupertino, Calif., the publisher of Q & A, which many reviewers have praised as an easy-to-use data base. Mr. Eubanks says that since any useful software will take some time to learn, it's a mistake "to put energy into learning something simple and then finding out it can't do much."

Buyers appear to agree. Companies that have made more limited, easier-to-use versions of their best-selling products have found sales weak. And although Apple's Macintosh is easy to use, IBM-compatible computers outsell them by 10-to-one, because the IBMs are cheaper and much more software is available for them. In fact, the Macintosh didn't become a hit until a complex desktop-publishing program, Aldus Corp.'s PageMaker, came out, providing a significant capability that wasn't available in the IBM-compatible world.

At the same time, producers of easy-to-use programs usually can't charge as much for them as for complex programs, reinforcing companies' reluctance to push simplicity. A few low-priced, easy-to-use programs like Quicken, a $49 check-writing program from Intuit in Palo Alto, Calif., have become best sellers. But all the big software companies count on selling software for $300 or more, and emphasize power over simplicity.

The whole process of software design also deserves some of the blame for complexity. For starters, when companies test their designs, they go to their existing customers. "Designers design a lot of code for people just like them," says Roger MacNamee, who runs T. Rowe Price's science and technology fund. "That explains why the current market doesn't get enlarged. There's no design step that says, `Put an idiot in front of the screen.'"

Because there aren't many standards, each designer also develops the best system he or she can imagine. The effect is comparable to having each typewriter manufacturer lay out its own keyboard, as they did before they standardized the QWERTY keyboard after 1873. "Everyone is reinventing the wheel," says Jan Davidson, president of educational software producer Davidson & Associates Inc. The result: No user can easily guess what keys will do what.

Even with a mouse, it often takes considerable pointing and clicking to be sure how to print out a line in boldface or indent a paragraph. "We have to constantly rein in our programmers," says Linda Pollin, chief financial officer of Interactive Media, a Diamond Bar, Calif., software firm. "Otherwise you get 40,000 icons that only two people recognize, and it takes 27 clicks to get to the place where you want to go."

And then there's the issue of the installed base. Software developers are so concerned about making it easy for current users to switch to new versions of their programs that they maintain illogical systems that they think users are accustomed to. Mr. Peterson of WordPerfect recalls, "We knew in 1985 that F3 shouldn't be the `help' key. But we decided we didn't want to upset the {existing} customers. If we had it to do over, we would have changed." Since then, millions of people have gotten used to WordPerfect's unorthodox system, making changes even harder.

Moreover, accommodating past software and vast amounts of installed hardware takes programmers a great deal of time. A software programmer joke: How could God create the world in six days? He didn't have an installed base.

But accommodating the installed base frustrates many efforts to make software easier. Microsoft Inc. of Redmond, Wash., is pushing Windows, a Macintosh-like software package for IBM-compatibles. Windows uses a mouse and graphical icons such as trash cans to represent commands. But buyers of Windows find it can take as long as four hours to adjust it for particular networks, printers and video displays, says Corporate Software Inc., a Canton, Mass., software-sales company that surveyed its customers.

William Gates, Microsoft's chairman and chief executive, says, "even as we make products simpler, the variety of hardware configurations and things people want to do is extensive," requiring custom fitting.

There is some good news, however: Software is easier to use today than it used to be. WordPerfect gets an average of one telephone call for help for every package it sells, half the rate three years ago. David Reed, chief scientist at Lotus, recalls that when he started using minicomputers, "they were only for people like me. You used all your knowledge all the time, just to use the computer."

Consultants say that graphical programs like Windows and Apple's Macintosh do make it easier to learn software. Temple, Barker & Sloane Inc., a Lexington, Mass., consulting firm, concluded that novices using graphical programs were half as frustrated as people using older software designs -- and taught themselves 23% more. But many Windows users still have to deal with older application programs with disparate demands.

New operating systems may offer more hope. Mr. Liddle of Metaphor, who is heading up an IBM-sponsored project, says that future software will be written in small, reusable chunks that other programmers or even ordinary users can patch together to make the exact package they want. "Ideally, you want it as easy as a component stereo system to plug things together," says Larry Tesler, vice president, advanced products, for Apple Computer. That currently is impossible, because there isn't any consistent way for one program to tell another to perform a calculation or reduce the size of a chart.

Other experts think the solution will be to get rid of the keyboard. "The keyboard will be dead in five or 10 years," predicts Mr. Bricklin.

The first big step in that direction will be so-called slate computers -- flat, notebook-sized devices that are controlled by pens writing across the screen. Jerrold Kaplan, chairman of GO Corp., South San Francisco, says GO's Penpoint operating system permits "software designed for people whose lives are focused on their work, and not on the tools to do their work."

Programmers say that Penpoint isn't designed for composing novels, but it works well for editing letters, say, or scheduling time. The screen looks like a loose-leaf notebook and a user can move from program to program by touching the pen to the pages. "We're able to deliver compelling simplicity because it's the first time you're physically touching the computer program," says Mr. Kaplan. He believes that removes some of the abstraction that discourages many novices.

But the ultimate in simplicity will occur when computers become powerful enough to understand speech. Currently, even the best speech programs require the user to slowly train the computer to understand each speaker or pause between words and often repeat statements. But that's enough to be useful in some applications. Emergency-room doctors now routinely dictate information on their patients to a computer system developed by Kurzweil Artificial Intelligence Inc., Waltham, Mass., in order to print out diagnoses and prescriptions.

Even sophisticated software programmers are surprised when they see how easy computers can be without keyboards. A few months ago, Raymond Kurzweil demonstrated the system to an audience of Massachusetts software developers. After a few minutes of dictation, he ordered his computer: "Print report." The audience broke into applause.


Mr. Bulkeley is a staff reporter in The Wall Street Journal's Boston bureau.

Copyright (c) 1991, Dow Jones & Co., Inc.