Trends for the Web in 1997

by Jakob Nielsen on January 1, 1997

Two major trends will revive the Web as a useful tool beyond the current hype and uselessness:

  • taking the Web seriously in business
  • a rebirth of semantic encoding

A little exercise before you read any further: take out a piece of paper and draw "a computer." It doesn't need to be a beautiful or highly detailed illustration; just sketch a rough outline of what a computer looks like. We will return to this exercise at the end of this essay, but you have to do it before you read my comments.

Taking the Web Seriously as a Business Tool

Despite the enormous hyperbole surrounding the Web, current reality is that almost nobody takes it seriously as a business tool. Even though the December 1996 holiday shopping season was the first in which a few merchants reaped significant sales from the Web it is still almost impossible to actually get to buy anything on the Web: one journalist who set out to do all his Christmas shopping on the Web only succeeded in acquiring a single present. The problem is that most companies view their websites as glorified brochures and not as a way of interacting with their customers. The goal of having a website should not be to brag to your friends that you are on the Web but to provide service to your customers (and thus make money).

In a recent survey of large British companies by JCP Computer Services, only 25% of the Information Technology Directors said that they had ever used the Internet to purchase goods or services. When not even the IT Directors have used the Internet for anything real, it should come as no surprise that their companies end up with failed websites and unvisionary plans for leveraging the Internet. My challenge to any CEO reading this: Has your CIO ever bought something through the Internet? The same question should also be asked of your head of marketing. Nobody should be in charge of a company's Internet strategy if he or she doesn't take it seriously enough to use for some personal business. (After you have checked into this matter, go and buy a few things on the Web yourself!)

When you do try to shop the Web, you quickly realize how most websites fail to provide true customer value. Sure, they may have all kinds of "cool" decorations and animations, but there is rarely anything real there. Full product listings are often missing, as are detailed specs and indications of whether the product is even in inventory and what the expected delivery time will be. If you place an order despite this poor service, you will normally be left in the dark as to whether any human ever saw your order until the happy day when the UPS truck comes by with the goods. Almost no commercial websites have bothered to tie their inventory and sales management systems into their email system to give customers immediate notifications of shipments (and any delays). This despite the finding that punctual delivery of Web purchases increases customer loyalty from 60% to 96%.

The problem seems to be that executives in charge of Web strategy don't think through the implications of having an interactive medium at their disposal. The lucky corollary is that anybody who does do so will have a relatively easy time standing out on the Web and attracting the customers. There is a the huge difference between seeing a cool demo of a Web site and actually using it for a real purpose. I predict that 1997 will finally be the year when customers (and their money) will talk loudly enough to get rid of superficial coolness and make websites into serious business tools: this year may well be the last chance for legacy companies to make their Web presence useful before customers irreversibly defect to new brands.

The Renaissance of Semantic Encoding

The original design of the Web and its underlying data format, HTML, was based on encoding the meaning of information and not its presentation . For example, the above heading is encoded as a level-2 heading (<H2>), meaning that it is the highest level of subheading below the level-1 heading that's the heading for this entire essay. This encoding style was chosen by Tim Berners-Lee because he wanted the Web to be a universal information system: thus, he could not know what computer equipment the various users would have (some might have high-resolution color screens and others might be blind and use a voice-only interface) and it was necessary to keep the details of the rendering of the information out of the file itself. The exact way pages would be shown (or read) to the user would be determined by the user's own equipment.

The notion of encoding the meaning of documents was temporarily lost as some browser vendors introduced proprietary tags to encode the exact rendering of the information. For example, many Web designers have been induced to use presentation-based text encodings like "18 pixels tall bold Garamond" instead of using the semantic encoding of a level-2 heading. The benefit of presentation-based encoding is that the page may display in a close approximation to the intended design if the user has a combination of hardware and software that is similar to that used by the designer. Thus, more sophisticated layouts became possible.

Presentation-based design only works as long as it is possible to predict the user's hardware, software, and preferences. This was reasonably easy in 1995 and the first half of 1996 when almost everybody was using the same browser software and almost all hardware was essentially equivalent. If you followed my instructions to draw a computer, you probably sketched something with a big box for a monitor, a keyboard, and a mouse, since that's how the typical computer looks today. In 1997, we will see non-traditional computers like WebTV and personal digital assistants like the Newton and the PalmPilot become much more prevalent. These devices have very different display capabilities than the traditional computer (typically a much smaller screen) and thus cannot display Web pages well that have been coded for a specific presentation that looks good on a standard monitor. Using semantic encoding allows the device to optimize the display to its capabilities.

I also expect speech-based browsers to be used more extensively in 1997. Enhanced accessibility for users with disabilities (especially blind users) is one reason for speech browsing. Seeing users also find themselves in hands-busy, eyes-busy situations (such as driving a car) where they could access Web-based information if it were read out loud. It is obvious that speech interfaces to the Web will be much better if semantic encoding is used so that the system can understand the structure of the page. For example, knowing which parts of the text constitute headlines would enable a speech system to read a summary to the user who could then easily chose what sections to have read in full.

A final reason for reverting to semantic encoding rather than presentation encoding is the increased diversity in software used to access the Web. If one compares screenshots of the same page in different versions of the same browser (or even the "same" version running on different platforms) it is very clear that the resulting presentations differ substantially. My site was accessed by 68 different browser versions in December 1996. Even the single-most popular browser version (Netscape 3.0 for 32-bit Windows) accounted for less than 20% of the use. As different browser vendors gain market share and as all of the vendors release ever more versions of their browsers, there will be so many versions that it will become impossible to test pages for all of them if page designers insist on tweaking display appearance. You never know what will come next, and since " Data Lives Forever", your only hope of page survival is to follow the standard.

Instead of embedding appearance-specifications in the content, a better solution is to separate content and display-specific instructions. Information relating to the presentation of the information should be kept in a separate styl esheet file that is linked to a content file that only contains semantic markup. Style sheets are a new development on the Web and currently not widely used, but they are the only solution to getting nice presentation with ever-increasing numbers of browsers and display devices. For example, a page could link to three different stylesheets: one for desktop computers, one for small-screen devices, and one for television sets. Style sheets should be linked rather than embedded in the file in order to facilitate maintenance and sharing of style across pages (two benefits: UI consistency and less download time due to caching - if only Internet Explorer would stop hitting the server every time the same style sheet is referenced).


Share this article: Twitter | LinkedIn | Google+ | Email