First impressions count

Remember when your mother told you that first impressions count? Web design is finally getting to that point. Whether it is because Internet use has moved far enough across the chasm to attract more of the late adopters looking for a "user experience," or because sites have improved enough that the global frustration with poor usability no longer trumps the first impression of the site, site designers are beginning to appreciate that mom was right. First impressions count again. The shift toward emotional design reflects one of the most interesting and exciting trends in Web development so far. It will present new challenges. It will also bring old ones back into sharper focus.

One of the chronic challenges that will be highlighted by emotional design is site download speed. There are many sources of delay in Web site and application delivery. Some, such as increased Internet congestion or connection speed, are outside the control of the designer. However, other causes of delay such as page/graphic size, server-task prioritization, and incremental page presentation, are within the control of the designer and implementation team. Thus, one challenge for successful emotional designs will be creating sites that balance visually compelling and task-rich resources with quick delivery.

Speed of download is not a new problem for Web designers. Weinberg (2000) estimated that $4 billion in potential e-commerce revenue is lost each year because of download delays. But how, specifically, do delays figure into user frustration and site abandonment?

Hear what I say, or watch what I do?

In an attempt to find alternative approaches to compare usability testing and expert review, Muller, Dayton and Root (1993) reanalyzed the findings from four studies (Desurvire, Condziela, and Atwood, 1992; Jeffries and Desurvire, 1992; Jeffries, Miller Wharton and Uyeda, 1991; and Karat, Campbell and Fiegel, 1992). Rather than looking at the raw number of problems identified by each technique, their re-analysis categorized the findings of the previous studies on parameters such as:

• # problem classes per hour invested,
• # of classes of usability problems identified,
• likelihood of identifying severe problems,
• uniqueness of results, and
• average cost/problem identified for each technique.

Again, their re-analysis demonstrated no stable difference indicating that either usability testing or heuristic review (conducted by human factors professionals) is a superior technique.

An array of methodological inconsistencies makes interpreting the findings in toto even more challenging. The specific types of interfaces or tasks used in comparison vary widely from study to study. Many studies do not clearly articulate what the "experts" are expected to do to come up with their findings (much less, what they really DO). The specific heuristics applied are rarely specified clearly. The level of expertise of the evaluators is rarely described or clearly equated, although it is often offered informally as a factor in the diversity of outcomes. As such, it is possible that, among other things, the conclusions of specific studies falsely favor a method, when relative benefits really result from the broader and deeper experience of the individual implementing the method (Johns, 1994).

And they call this FAST food?

Actually, determining just how long someone will wait is an old problem in the service industry. Years and years of research have been dedicated to determining how waiting for service effects customers' perception of both the product and service provider. The fundamental finding of this work is reflected in Maister's First Law of Service:

Service = Perception – Expectation

In other words, customers develop positive feelings when the perceived service exceeds the expected service. With respect to delays, users are happy when the site responds as quickly, or more quickly, than they expected, and they become frustrated when the site is slower than they expected. But how long is too long to wait for a Web site?

Many groups of researchers have set out to define how delays impact users' perception of interactive sites. Not surprisingly, these studies show that users faced with long, unexplained delays during the course of interactions are dissatisfied. But how long is too long? And what is the real impact of perceived delay?

Waiting is frustrating, demoralizing, aggravating, annoying, time consuming, and incredibly expensive. — FEDEX Commercial

Study 1: Does delay impact perceived usefulness?

Sears, Jacko, and Borella (1997) created two versions of a Web site (text-only and text-plus-graphics) to examine the effects of content displayed and download delay on perceived usefulness. In this study, participants experienced delays ranging from 0.58 to 6.80 seconds and then rated sites on perceived usefulness, organization and quality of information. Users ratings reflected sensitivity to the delay for both the text-only and text-plus-graphics sites. However, ratings for text-only pages were lower than text-plus-graphics pages even at the shortest delay intervals suggesting that participants expected text-only pages to render more quickly than graphics intensive pages.

Study 2: Does delay influence how interesting a site is?

Ramsay, Barbesi, and Preece (1998) examined the effect of fixed-download delays, site type, and page style on perceived interestingness of content and ease-of-scanning. Along with delay time, they varied types of site content (scientific, business, advertisement, personal, history, instructional and entertainment) and page styles (text only, graphics with few links, balanced text and links). The delays in their experiment ranged from two seconds to two minutes. Sites with long delays were rated significantly less interesting and more difficult to scan independent of the content presented. In addition, page style did not influence the perceptions: participants expected graphics-heavy sites to respond as quickly as text sites.

Boredom results from being attentive to the passage of time itself. — William James

Study 3: Does incremental presentation shorten the delay?

Bhatti, Bouch, and Kuchinsky (2000) report a study, and related guidelines for server design, based on work exploring factors moderating Quality of Service (QoS) perceptions on e-commerce sites. For their study, Bhatti and colleagues developed an ecologically valid, yet quantifiable, task by artificially injecting download delays (ranging from 2 to 73 seconds) as participants completed a set of common tasks on a live Web site. They designed the study so that the primary task of configuring and purchasing a home computer system could be broken down into a series of subtasks by purchasing components separately. They evaluated the impact of download delay by having participants rate the tolerability of delay for each page after it had completed rendering. In a critical comparison, Bhatti and colleagues varied page presentation by having pages render either completely or incrementally. Pages rendered incrementally first displayed the page banner, followed by text and then graphics.

Consistent with previous work, Bhatti and colleagues noted that participants in the complete-rendering condition started to judge delays unacceptable at roughly 10 seconds. This is consistent with previous work in applied attention and cognition suggesting that after 10 seconds or more of wait time, the boundaries of the current 'unit task' are broken (Card, Moran, and Newell, 1983). Simply put, after 10 seconds we get bored with waiting and start to look for something else to do. As a result we lose our place in the task and sometimes even forget what we were doing.

Critically, however, Bhatti and colleagues observed a significant difference between the complete and incremental rendering conditions in their study. Participants in the incremental rendering group tolerated up to 6 times more delay. They suggest that incremental loading helps users keep their attention on the task at hand (rather than redirecting it to the task of reevaluating the Quality of Service.)

If it's a Full Professor, you wait 20 minutes. If it's an Assistant Professor, you only wait 10....

The studies outlined above converge with other findings to suggest that 10 seconds is about the edge of too long. However, this is not necessarily a hard rule. Willingness to wait is moderated by other factors. For instance, novice users and older individuals tend to be willing to wait longer for a computer to react (Schneiderman, 1998, and Selvidge, 2003, respectively). In addition, users who have little or no experience with high bandwidth connections are more patient (Selvidge, 2003). Users tend to be relatively more patient the first few times they visit a site (Bhatti, 2000). Finally, users will wait longer in the service of completing important tasks. However, these same users report higher frustration levels than counterparts who experience the same delay in completing less important tasks (Ceaparu, Lazar, Bessiere, Robinson, and Schneiderman, 2002). In this last case, willingness to wait may indicate that the user feels trapped rather than satisfied.

This brings us to Maister's Second Law of Service:

It's hard to play catch-up ball.

That is, any impression (or halo effect) created early in a service encounter will color the rest of the interaction. As such, since not all delay is under the control of the designer (e.g., user's connection speed), it is important to consider (and test!) download speeds even if you are designing for a more tolerant group. Indeed, the previously mentioned literature on perceived quality in the (physical) service sector clearly shows that the largest payback for effort and attention spent in improving the interaction occurs from improving the perception of the early stages of the interaction — reducing the waiting time (Maister, 1985).

Are we there yet?

In summary, perceived delays in site presentation undermines users' evaluation of the site. Users systematically rate slower sites as less interesting (Ramsay, Barbesi, and Preece 1998) and having lower quality content (Jacko, Sears, and Borella, 2000). In addition they report that delays interfere with task continuity, their ability to remember the site, and use flow (Shubin and Meehan, 1997). Exceedingly slow sites can lead users to believe an error has occurred (Lazar and Norico, 2000). Finally, users correlate site performance and security: Chronically slow sites are considered to be less secure resources for purchase (Bhatti, Bouch, and Kuchinsky, 2000).

In the face of increased pressure to create visually compelling designs, these findings highlight the importance of balancing performance factors with emotion.

Reader comments