Web Objects and User Schemas

Design Standards and Popular Sites

In 1999 our web guru Jacob Nielsen published a classic article on his usit.com site called “When Bad Design Elements Become the Standard”. He argued that the most important thing in web design is consistency, and, when one considers the web over all, this consistency is defined by the most popular sites, which create “de-facto” standards. (Note that these design "standards", have to do with the appearance of the web page and are quite different from the coding standards advocated by Jeffrey Zeldman and the Web Standards Project).

More specifically, Nielsen states that "

He is also quick to note that these standards do not necessarily represent the most usable design. In fact, he goes on his article to discuss some example standards, most of which he criticizes for one reason or another. These examples include: navigation tabs, left justified “navigation rail”, and breadcrumb trail. The article provides more information on what these standards comprise and what their weaknesses are.

User Schemas

The term“schema” has been around for a long time in Psychology. It refers to a mental construction, or representation, often representing some prototype or typical situation or scenario. One of the basic tenants of schema theory is that we all learn basic schemas for how things work in the world and then we store (remember) these schemas and use them as guides to make our behavior more efficient. The notion of schemas is implied in Nielsen’s article. What he is suggesting, in effect, is that a user develops a schema for how web pages and sites are normally designed and this user then utilizes this schema to work more efficiently on a page. These schemas are presumably learned from experience on the web, which is best represented by the most popular sites.

Nielsen makes an implicit but important assumption, which is that user schemas are formed from users experiences with the most popular sites, so one can simply look to these sites to get an idea as to what is probably going on in a user's head. Though this is almost certainly true to some extent, the complication with this assumption is that memory and schema formation are tricky things. Certain salient characteristics of an object or procedure are more likely to be remembered and stored, and many things affect schema formation beyond a single type of experience. For example, the schema one develops for shopping online may be partly influenced by experiences shopping off-line. This would not necessarily be represented in the most popular e-commerce sites.

Research on User Schemas

One possible solution to allow us to identify user schemas more directly is through research with actual users. Such an approach was used in two studies, one by Michael Bernard and published at Usabilitynews.org. (Usabilitynews.org is an excellent source provided by Wichita State's Software Usability Research Lab for empirical research on web design) and another conducted by my student Jeremy Markham as a part of a graduate class on research methods. In both of these studies users were presented with a grid that represented a web page and were presented with a series of web objects. Web objects can be thought of as individual entities that represent some functional unit on web page/site (e.g., links of different sorts, search forms, banner ads). Users were asked to indicate where on the grid they would expect to find a given web object.

In the Bernard study the users viewed a paper representation of a web browser divided into 56 quadrants (7 x 8). They placed the web objects where they expected them to appear on a typical web page. The objects that were used were: “back to homepage” link, links to internal pages, links to external pages, internal search engine, and advertisements. He also compared the novice and experienced users, the former consisting of those who use the web a few times a month or less, and the latter those that used the web a few times a week or more.

Bernard found that users were quite consistent in their expectations and he also found that the novices and experienced users differed very little. He found that the 1) Back to homepage link was expected in the upper left corner; 2) internal links were expected on the left side of the page; 3) external links were expected on the right side, or bottom of the lefts side of the page; 4) internal search engine was expected the upper and bottom-center of the page (this was less consistent); and 5) advertisements were expected in the upper middle of the page. (Users expectations are represented graphically in his article – scroll down to results).

Jeremy Markum’s study (Markum & Hall, 2003) borrowed the basic methodology from the Bernard study with the following differences: First, the geometric grid method differed in that the task was simplified. Markum’s study was completed online with participants viewing a real web page divided into quadrants. In addition, he included nine quadrants (3 x 3) rather than fifty-six used in Bernard’s research, and did not allow web-objects to overlap, nor did he allow the objects to cover more than one quadrant. A second, important difference was that his focus was specifically on e-commerce sites, rather than generic web sites. While the Bernard study included five generic web objects, Markum extended this to ten objects and included objects such as “shopping cart” and “order button”, which are unique to the e-commerce environment. Third, he also asked users to rate the importance of each of the web objects in terms of their impact on purchasing. So, not only did he identify user expected locations for web objects, he also identified those that they rated as most important in making an online purchase.

The importance ratings are illustrated in a Figure in the paper, which is reproduced in Figure 2 below.

Importance Ratings
Figure 1. Importance Ratings as a Function of Web Objects

With respect to the importance of web objects results, the article's interpretation is as follows.

“… the importance of a given web object is primarily a function of how important the object is for allowing a user to make a specific purchase quickly. The links to merchandise, shopping cart, and order button were rated as the top three most important objects. Such a rating fits well with the scenario that occurs when a user comes to a sight knowing what she wants with the intention of buying and completing the transaction. … It’s also interesting to note that these three top rated objects have traditional brick and mortar shopping analogs. Even though logging in to register is necessary with most e-commerce sites, both merchandise links and order button were rated significantly higher. Logging in to register is, of course, not a part of the traditional bricks and mortar shopping experience. This suggests that the traditional, non-internet, schema for buying is very powerful even for online shopping. It’s also interesting to note that links that are not directly associated with the one-stop shopping task, including internal and external, and even help links were rated quite low by the users, which further supports the picture of a shopper on a mission, as opposed to one who wants to explore related sites, or even learn about how to use the site. The low rating of the help button indicates that users expect the site to be usable and straightforward, they do not expect a help link to be an important part of the purchasing experience.” (Markum & Hall, 2003)

In terms of preferred location, the results of Markum’s study were consistent with Bernard’s in that the users were in relatively strong agreement as to where given objects should be located. Statistical tests indicated that with eight of the ten web objects one location was preferred significantly more than others. Also, as with Bernard’s study, the link to the home page was in the upper left corner, the other links were primarily on the left side (which, incidentally, is a convention that Nielsen mentions in his article). Further, the banner ad was expected at the top center of the page in both studies.

The “prototype” page that emerged from Markum’s study is depicted graphically in the article and reproduced in Figure 2 below. Those items in parentheses are the two objects that did not have a single statistically significant preferred quadrant. (The top three preferences for each object are illustrated graphically in the article – scroll down to the results section).

Prototype Web Page Based on Users' Expectations
Figure 2. Prototype Web Page based on User's Expectations

In summary, it appears that users have well established expectations as to where they expect objects to be located on a generic or ecommerce web page. In other words, users do appear to have well-established schemas and these are reasonably consistent across users. Designing according to these schemas, at least with respect to a page and common web objects appears to be straightforward.

References