Introduction

It would be expensive and space-consuming for users to own guidebooks from each city, reference guides for specific tourist attractions, and newspapers from various cities. In addition, people would need to make sure the information is accurate and up-to-date. The internet, however, provides a tool that allows for easy maintenance of information at a low cost and takes up very little space. Users can find a full range of information about cities and destinations without ever leaving their home or purchasing a guidebook. The internet has the potential to offer users of different backgrounds and with varied tasks to access information with little expense, space, or time.

The project of the Pittsburgh Restaurants Online team tackled the issue of users' choices in navigation. Because people are accustomed to searching for information the way that they feel most comfortable, we suspected a web site must be designed to provide many paths in navigation. A design with well-designed options will grant users flexibility in navigating the site to create fast and informative searches and will interest users in visiting again. To create such a design takes thorough analysis of current information sources, data, behaviors, as well as user testing. This report will explain the steps our team took to solve an information design problem.
top

Step 1: Analysis of Information Design Problem

This particular project was chosen because of the poorly designed travel and restaurant sites that we found. It became obvious that the current sites were not taking into account the differences in usersâ background knowledge. In addition to variations in research tool preferences, users differ greatly based on levels of familiarity. In this particular project, users are categorized by level of familiarity with the city.

Here are some examples of travel sites about Pittsburgh and explanations of their problems:

www.realpittsburgh.com
Because there are too many headings, it was difficult to find the information we were looking for. When we finally located the restaurant link, it led us to another poorly-designed page÷a survey about  preferred qualities in a restaurant and information about places that failed  inspection last year. There are no actual restaurant links. This site seems to cater to people who are very familiar with Pittsburgh, because the presentation of information presumes that you already know the restaurants you prefer. The confounding nature of this site makes a newspaper seem more appealing.

www.pittsburgh.net
Because this site is designed for both residents of and travelers/visitors to Pittsburgh, it has links for various locations in the city. Although it is clearly designed, the main section is divided by links to areas/neighborhoods with a sidebar for visitors. The links for visitors are somewhat better defined, but they lead to restaurants based solely on location. There is no way for the users to choose a restaurant by cuisine or price range. This site is very attractive but seems to be better suited for users who are more familiar with the city.

www.lonelyplanet.com/lp.htm
Although the Lonely Planet is a reliable and trusted source in its printed version, the on-line version provides no easy way to access particular information easily. Despite its aesthetically pleasing pages, it is difficult to navigate around the pages. Links named  ãoptic nerveä and ãthe thorn treeä may sound catchy, but donât give the user a clear sense of what lies behind them. Because trial and error is the only way to find information, it takes a lot of time to work your way through this site.
Our next step was to conduct direct observations that would allow us to view first-hand the differences in using reference guides by familiar and unfamiliar users. The type of reference guide, how it was used, the level of success achieved by users, and other means of referencing the information are all topics that we could learn during testing.
top

Step 2:Interviews
The interview and direct observation results helped us to better understand the different processes undertaken by users, and to identify that there is a difference created because of the level of background knowledge between familiar and unfamiliar users. We selected 16 of our potential users for interviews, and 5 of those for direct observation to analyze different ways that our potential users seek and use information needed, as well as to find out needs that are not met by current information providers.

Profiles
Interviews were conducted with people that had differing levels of familiarity with Pittsburgh. These interviews were given to people between the ages of 20 and 73 at varying locations, including the Pittsburgh airport and Carnegie Mellon Universityâs campus.

Methods
We conducted the interviews by asking specific questions, being careful not to put the interviewee on the spot or to lead their answers. We asked questions beginning with a general inquiry about their presumed level of familiarity with Pittsburgh that also targeted the specific type of information they would most likely search for. Because we focused on their recent searches for information on Pittsburgh and other cities, we felt they could remember specific instance as to how they searched for the information (i.e. navigation path).

While recording their responses to questions, we also took into account different success and failures that our subjects encountered. We could then try to include the advantages and avoid the disadvantages of information searching in designing our web site.

Results
We found that people look for dining information not only when they plan a trip to a city, but also when they are a resident of the city. Using this new information, we decided to create Pittsburgh Restaurants Online web site so that would help both those unfamiliar and those familiar with Pittsburgh to find a restaurant through searching by location, name, or cuisine. We intended to keep users from ãguessingä where to go, and instead, guide them to the information they need as quickly as possible.

Because many of the interview subjects described that they had accessed restaurant information very quickly through asking friends, we decided that our site also needed to provide quick access to information. The second prototype idea stemmed from another of our interviewees who suggested including a map that provides detailed driving directions to the restaurants. The map would be helpful especially while searching for a restaurant by location.

Our interviews led us to believe that most people do not resist using the internet to obtain information if they have a computer that provides Internet access nearby.  Recent studies have shown that PC sales will have a 13% growth in the US (NetNews, March 11, 1998) and that the percentage of households with internet connection is doubled every 3 to 6 months (NetNews, March 13, 1998). From these statistics, we concluded that there is a need to design a web site environment in which users can get the same information that they currently obtain from other information resources, such as asking friends, newspapers, guide books, etc.

Another of our interviewees mentioned how difficult it is to access information about restaurants. He described that the local entertainment weekly, InPittsburgh, categorized information by location, a hierarchy that did not fit his particular navigational task—trying to find a restaurant serving Italian food. We noticed that any of these paths could work effectively, but depending on the navigational tasks of the users, not every path works every time. Their different degrees of background knowledge of Pittsburgh directly affects how they are able to navigate the site. We concluded that our web site should try to discern which path(s) provide the greatest usability for our diverse group of users.
top

Step 3:Direct Observations
We felt that direct observations could help us understand which path(s) is most appropriate for guiding both familiar and unfamiliar users through a site quickly and easily while providing all the necessary information.

Based on our interview results, we found both familiar or unfamiliar users generally gather information about a city in four ways: through friends, travel guides, newspapers, and current web sites. Our interview results indicated that people who are the most familiar with the internet are those who have daily access to it. We must find out, however, the navigation path a frequent user of the internet would otherwise gather information.  This process is important to us because we must discover the problems of language and navigational path that are occurring.  We thus chose users familiar with the internet, yet asked them to use the above four ways to gain information.

Profiles
Subject 1. 29 year old woman, not from Pittsburgh, but has lived here for approximately ten years, a Master student, very familiar with the Internet. She even designed her own pages. Her tasks included accessing the web site of Pittsburgh Visitorâs Bureau.

Subject 2. 30 year old man, not from Pittsburgh but has been in school here for four years.  He is a PhD student, and used internet daily. His tasks involved interacting with a travel guide.

Subject 3. 33 year old man, not from Pittsburgh, but has lived here for 3 years, a Ph.D. student, very familiar with internet. He was asked to use a current Pittsburgh web site.

Subject 4. 23 year  woman, native of Pittsburgh, a Master student, somewhat familiar with internet. We asked her to gather information about restaurants by asking one of her friends.

Subject 5. 23 year old woman, has lived in Pittsburgh for 6 months, a Master student, use internet only for school.  Her tasks involved checking with InPittsburgh weekly paper.

Findings
Subject 1, we noted that at times, found it necessary to search by name; at other times, it was most helpful to search by area, depending on the task. We observed that different users require different ãentry points," however,  because of their a priori experiences with a given area. While designing our web site, this finding gave us the incentive to provide different navigation paths to meet different usersâ navigational tasks.

While the first choice of Subject 4 in searching for restaurant information is to ask friends, she realized that often, they could not provide all the necessary details. She pointed out that things such as location, business hours, cost, food, and reservation information are a priority for a person going out to dinner. We kept this in mind when designing our final prototype.

While the information sources we uses were varied, they all revealed a common problem for both familiar and unfamiliar subjects—language. Two of our subjects took an unnecessarily long time to complete their tasks because they did a lot of satisficing because of ambiguous link labels (Neuwirth, Topics & Links) This observation supported our effort to provide clear language to facilitate navigation.

One of our subjects indicated the importance of organizing and displaying information in a hierarchical structure. The interviewee could not complete the task because the link label used in the category did not match his expectation. In an attempt to please most users, therefore, we determined that it may be necessary to create a site with several hierarchically organized paths to allow our potential users test the site.

Results
We found that there are several requirements that we need to consider when designing a dining guide for every one:

• Information about restaurants should be "chunked" in an obvious way to our users while being broad enough to cover their expected navigation goals.
• Navigation time should be short.
• Icons should be recognizable by people who are both familiar and unfamiliar with Pittsburgh.
• Some users access from home with slow modems; we need to accommodate them with fast downloading time.

Step 4:Prototype Development
Our initial plan was to have two design variations: text-based and graphics-based. While both designs provided the necessary information for users, the development of each targeted ease of use and speed for users with different levels of familiarity. Both sites used a hierarchical structure to create paths that allowed users to navigate through the pages according to the task at hand. Because we ãchunkedä the information on the homepage, they could access different levels of detail, minimizing their chance of getting buried in the design. At this point, we were also considering if users would find a graphics-based design more useable than a text-based design.

Required Capability
After analyzing the required capability of our site, we felt that the new capability would save navigation time because there is a cognitive model of link selection that matches users' expectations, whether they are familiar or unfamiliar with the city. (Neuwirth, Topics & Links). Further, the graphics-based layout may encourage users' involvement, as well as decrease users' perception time by displaying locations and other related  information at the same time (Norman). Applying graphics, however, had the potential to increase costs since it took longer to download, especially for those accessing from home. Additionally, the costs in maintenance are higher since we did not find a way that we could automatically update maps when new restaurants were added to the database. The webmaster, therefore, may have needed the help of a program developer.

Organization
Because we wanted to test the users' preferences for a primarily text-based web site versus a graphic-based web site, we decided on the following variations for the Pittsburgh Restaurants Online site:

• Design 1  presents information using only text
• Design 2A   presents information using only icons
• Design 2B   presents information using text with mouseovers
• Design 2C   presents information using a combination of text and icons
• Design 3  presents descriptions of areas in mouseovers

Following the advice of David Siegel, we carefully considered our options for the home page because it "is the main entrance into the web site and can impress users, generating interest and curiosity, or, if poorly designed, cause indifference, or even repel users" (Siegel 130). Although the designs have different entry points, namely the home pages, the other elements of the pages remain consistent. (See Appendix A, Design 1).

Design Alternatives
Design I
The first design (Appendix A, Design I) we developed was a text-based web site which places equal emphasis on all four subcategories of the information: name, cuisine, location, and table talk. Because large, beveled, and labeled plate icons serve as links to the subcategories, this design places equal emphasis on all four methods of locating the information; the similarity in size, color, and shape of the icons and text indicate that these categories are at the same hierarchical level. The design recognizes the different levels of familiarity users will bring with them as they conduct their search. The ãnameä category for instance, will most likely serve the needs of people who are already familiar with Pittsburgh restaurants. The alphabetical listing that the ãnameä category contains provides a quick and convenient way of finding out further details of a restaurant the user has in mind. The ãlocationä category, in contrast, will most likely benefit users who are unfamiliar with Pittsburgh, although familiar users may also use it. Design 1 assumes the users' familiarity with the spatial layout of the city and knowledge of where the different areas in Pittsburgh lie. We suspected that although terms like ãNorth Hills suburbs and vicinityä and ãSouth Hills suburbs and vicinityä may intuitively help a user understand where the areas are, a person who is unfamiliar with Pittsburgh may find little distinction between North Hills and North Side, and South Hills and South Side.

Design 2
The second design (Appendix A , Design 2) we developed uses location as a tool for accessing information. Options to search by ãcuisine,ä ãname,ä ãtable talkä and ãlocationä are still provided on the home page, but this time the icons are placed in a column running down the left of the page. A sketch map of Pittsburgh constitutes the main graphical element of the page, which provides the user with a visual representation of how regions in Pittsburgh are spatially located. Unlike Design 1, this method of presenting information, we felt, does not assume any prior knowledge of Pittsburgh and its surroundings. We chose a sketch map instead of a real map because we felt that the details of a real map were unimportant and it would also be difficult to include all road names on a sketch map because of space limitations. More importantly, we didnât want to overload the users with excessive detail at the initial stages of their search.

We rejected a sketch map that we found in The Pittsburgh Herald Tribune because we felt that the map presented too much information, which hindered a clear distinction of where the boundaries of each region lie (see Appendix A). The orientation of the map also posed a problem. The east end appears at the top part of the map, an orientation which conflicts with the expectation of most users. We did agree, however, that the sketch map was a good idea and incorporated that in our own map. The map we used has also been repositioned so that the regions match the conventional compass directions.
We used the pared-down version of our sketch map in Designs 2A-C and 3. By using a map as a main graphical element on these home pages, we were anticipating the usefulness of providing multiple access pathways to the information. These designs evolved also from the assumption that the usersâ background (varying levels of familiarity) influences the way in which they would access information.

In addition, by using the sketch map as main tool of navigation, we were able to experiment with how icons and text labels help to direct the user to the relevant information in the shortest time. Design 2A relies only on icons to help the user locate the information. It assumes that a user who is familiar with the city would be able to identify the icons quickly and associate them with the area they represent. We suspected that the cognitive leap a person makes between the icon and what it represents would not occur as easily for the unfamiliar user.

Design 2B (see Appendix A, Design 2B) uses text labels to mark the eight regions of the map according to the subheadings used in  the next level of the hierarchy of information. Design 2C  (see Appendix x, Design 2C) uses a combination of text labels and icons. We tested this method of presenting information on both types of users (familiar and unfamiliar) to see there was a significant reduction in the response time.

Design 3
Design 3 (see Appendix A, Design 3) allowed us to explore the effect of providing information in addition to the labels used to categorize the information in the next level of the hierarchy. We planned to use mouseovers to avoid overloading the home page with too much information, using a pop-up menu that contains a short description of an area. Our intention was that these descriptions could be helpful for unfamiliar users because people who are unfamiliar with an area may want additional information if they are just browsing different dining options.

Consistent Elements
When we began planning our site design, we looked at similar existing sites and found that when the design changed from page to page, it became difficult to tell if we were still within one site, or had linked to another (see Appendix A). Laura Lemay also suggests that a single 'look and feel' for each page will be comforting to readers (Lemay 271). Because of these experiences, we wanted to consistently implement one format throughout our site that was aesthetically pleasing as well as functional.

Titles
Because the title of Pittsburgh Restaurants On-line is centered at the top of each page, our users will become familiar with the organization of the site. According to William Horton, titles should "summarize or preview the topic. Thematic titles summarize the topic and prepare the user to notice its key points" (Horton 120). The subtitle of each page, a description of its contents, also appears in the browser's title bar. We carefully subtitled each page so that when users bookmark a page or use the back function, they will be able to clearly identify a visited page.

Typefaces
William Horton suggests avoiding serif typefaces altogether for small type sizes because they may display irregularly (Horton 246). Because it is a clean-looking and readable typeface, we chose Arial Narrow as our first font choice and Arial as our second, using the <FONT FACE="Arial Narrow, Arial"> tag. This way, the browser will automatically default to Arial for those users who may not have Arial Narrow. Although we may have sacrificed appearance, aside from our homepages, all the text is left-justified. Fully justified text may be more visually appealing, but it "usually requires putting extra spaces between words, resulting in uneven word spacing and occasional rivers of white" (Horton 250). We chose to make our text as readable and scannable as possible.

Background Color, Text Color & Link Colors
While it was tempting to use a tiled or brightly colored backgrounds, we decided that a white background showed off our the title GIF the best. In addition, Laura Lemay comments that low contrast between the background and the foreground can be hard to read, giving us good reason to choose black text for easy readability. We were also tempted to change the text links at the bottom of every page to green to match our title, but since changing the color of links can be confusing to the users (Lemay 268), we decided that staying with the default colors (blue, purple, and red) would be most functional, especially to infrequent users of the web.

Navigation Icons
76-487-887 Icons recommends that icons should either emphasize one global feature (shape) to make the icon more distinctive or emphasize local features that made the image more "representational" (Neuwirth, Icons). When we discussed the possibilities, we felt that there is no straightforward icon that represents name, nor any global feature that is easily linked to the word.
As an alternative, we chose the simple image of a plate, an icon easily associated with food, combined with the textual labels. Neuwirth advises that "a well-designed icon will have a simplicity of shape and a distinct separation of figure from ground," so the icons are uncomplicated and employ a drop shadow to distinguish them from "unlinked" graphics. 3D buttons give a universally accepted visual cue that a graphic has linking properties (Weinman 249).

For the icons used on maps, we chose icons that icons that are representative of their respective areas. Although we suspect that unfamiliar users wonât be able to recognize them, familiar users will benefit from the quick association they can make between the icon and the area. In our user testing, we will assess the costs and benefits of using icons and text.

Ordering & Hierarchy
ãThe concept of hierarchy is familiar to most people, due to their ubiquitous nature in everyday life. Without consciously knowing it, we absorb information in units which are internally structured, organized, and ordered into subcomponentsä (Sand 100). We tried to organize the hierarchical structure of our site, therefore, in a familiar way to accommodate usersâ expectations so as not to disrupt cognitive processes (Neuwirth, Topics & Links). In our interviews, although we did not directly pose the question of which is most important to people, the name, cuisine, or location of a restaurant, we feel that putting name first will be most familiar to users. Since name, being the least complex of the three links, occurs first, we decided to order the remaining two links by complexity (location has the most complex framework underneath it).

The ãtable talkä link leads users to a interactive form page, which enables them to contribute ideas and comments about dining in Pittsburgh, while also giving them the option of reading othersâ candid reviews. This exchange of honest and unbiased information may be attractive to both familiar and unfamiliar users who need a starting point for their search. ãTable talkä occurs last on our list of options because it contains the least complicated information, and when designing the site, we felt that that it would be least likely for a person to search for a restaurant based on reviews only.

Invisible Tables
For Designs II and III, we wanted the name, cuisine, and table talk icons to remain on the page at all times. To achieve this, we first thought of using frames, but since frames  ãslow down surfingä and our image map already takes several seconds to load, we decided to use tables instead (Siegel 85).

JavaScript Mouseovers
If our user testing had proven that users prefer the graphic-based design to the text-based design, we would have employed JavaScript mouseovers to present more information about a place without cluttering the screen. In this way, information would only be made available when the person moves the mouse over the section; they can choose when or how much of the information they want to access. In addition, JavaScript mouseovers would allow the user to access the information on the first level and avoid  ãsatisficing.ä

Our next step was to analyze through user testing which of the two means of presenting the information is most accommodating to both sets of users: text-based or graphics-based. Designs 2 and 3 allowed us to explore the value of using graphical and textual elements in general, as well as within the map itself. With Design 3, on the other hand, we could explore the value of bringing two levels of information to the home page.
top

Step 5:User Testing
Because we felt that users' levels of familiarity with the city influenced the navigation paths they used to access information about their dining options, we presented the five iterations of the design to our users, ten familiar and ten unfamiliar, and observed how they navigated the site to access the information requested in six tasks. One of the things we set out to explore in our iterations was the usefulness of a sketch map in helping unfamiliar users find information with a ãlocationä parameter. This was the main motivation behind the graphics-based iterations Designs 2A-C and Design 3.

Methods
20 user tests were conducted, four for each iteration. For each iteration, the user profiles are as follows:

• student/familiar
• student/unfamiliar
• non-student/familiar
• non-student/unfamiliar

We wanted our convenience sample to match the actual population that eventually uses our web site as closely as possible, so we were careful not to rely only on the Carnegie Mellon University campus for subjects. About half of our users were people from outside the CMU community. Following Ravden and Johnsonâs advice, we tracked our usersâ background information and considered it carefully when drawing conclusions about the results (Ravden and Johnson, 86).

Instead of testing a subjectâs reaction to all the five iterations, we decided to expose each user to only one interface, thereby reducing the effects of growing acquaintance with the content of the site, which could affect the test result of later iterations.

Tasks
The tasks we chose were benchmark tasks, defined by Nielson (1993) as tasks that ãspan the systemâs expected use.ä The tasks also ranged from easy to difficult, the latter requiring the user to search for a restaurant which satisfied a greater number of criteria, which correspondingly gave the user more options in his/her search.

Procedure
An important part of the test administration procedure was to reassure our users that the purpose of the task was not to test their abilities in web navigation, but to observe their choices in navigating our site. In addition, to ensure consistency, the group members also established a standardized method and sequence of administering each test.
We kept records of click trace graphs to track navigation paths and timed every task, taking note of think-aloud comments and overall feedback about the site in the QUISöQuestionnaire for User Interaction.

Analysis of Data
We analyzed the data according to the following criteria:

• the time it took, in seconds, to complete each task
• the number of clicks it took each user to complete each task
• the nature of the task itself (i.e. how many variables and unknowns were included in each question)
• The usersâ familiarity with Pittsburgh and with the internet
• comments made by each user after the completion of the test
• results from each userâs QUIS test, which measured site difficulty and user satisfaction

Task One
Find Penn Breweryâs Phone Number. This task presented relatively few difficulties for users because they only needed to find one item of information. Unfamiliar users in all designs took more time, possibly because the map (which took up almost 80% of the page) became a distraction and misled users by giving unnecessary information. Overall, users in Design 3 took a longer time to extract the information because of the simulated pop-up menus. This task reveled the usefulness of the alphabetical menuö69%of our users used it to access information.

Task Two
Find a restaurant that most likely serves weinerschnitzel. This task revealed a vocabulary problem for unfamiliar/foreign users who struggled with the cuisine categories which didnât always match their own definitions of cuisine. The users found some of the categories inconsistentöboth regional cuisines (Italian, German etc.) appeared together with ãtype-of-foodä cuisines (Delis and Bagels, Steakhouses etc.). The difficulties encountered by our foreign users also alerted us to the fact that sensitivity to cultural difference was something we could improve upon in our final prototype. Design 3 was generally a difficult test to administer because of the physical difficulty of combining paper and screen versions. Not surprisingly, users took a long time to complete the task. Users in the other four designs, however, completed the task in an average of 2 clicks. This task also confirmed that designs with text and icons required  longer processing times.

Task Three

You are going shopping in the North Hills and meeting a friend for lunch. Pick a place to go. This task caused problems for users in Design 2A, both for familiar and unfamiliar users, which proved that icons without text labels hindered task completion even for familiar users. The findings dispelled out prediction that familiar users would make the cognitive leap between icon and what it represents more easily. Some of the confusion evolved from the fact that the area boundaries were not clearly visible on the image map (see observation and quote below) and using icons to mark out the regions just wasnât an adequate communicative device.
[Moved cursor around map for a while deciding where to click. Clicked on Downtown Icon, realized that didnât include the North Hills, then clicked on the North Side icon] ãIâm not that good with this map, itâs too vague.ä                                  öSubject 2 (familiar)

äI got the South Hills icon confused with South Side ö so the icon got me to the wrong place. Instead of messing with icons I went to ãName.ää            öSubject 4 (unfamiliar)

Although this design took the same amount of time for users to find the information as they did in Design 1, Design 2A users were required to make more clicks in order to complete the task. Design 3, however, succeeded in requiring the least number of clicks and the shortest time to complete the task, because the map enabled a second hierarchy of information (the eight area names) to appear on the map. Unlike Design 1, users did not have to click on ãLocationä before they could find North Hills. This led to the conclusion that designs that included a map were more useful for both familiar and unfamiliar users when the task involved finding a location, than if the question required the user to find a restaurantâs name or type of cuisine.

Task Four
You've heard of Margaritaville on the South Side. Find which block it's located on. This task gave users two variables (name and location) by which they could conduct their search. Because of the familiarity the users had with Design 3 by the time they reached this task, it proved to be most effective for locating information that satisfied both variables. The pop up menus ceased to be a distraction because users were no longer interested in reading what was in them since some of them tended to be task oriented.
The variability of test administration caused some inconsistencies in our findings. Users of Design 2C were tested on a smaller screen, which increased the time taken by the users to complete the task because they were unable to view the South Side icon until they scrolled down.

Task Five
You're meeting a friend for Italian food at Davio's. Find out where it is. These results were fairly consistent across all five designs, with 60% of all our users opting to search by ãname.ä In this task, we also had to consider the ãrecency effectä (an effect produced when users conduct their search according to the last piece of information they hear), which could have influenced the decision to search either by ãnameä or ãcuisine.ä This task also highlighted the importance of looking at both ãnumber of clicksä and ãtime taken to complete taskä to assess the effectiveness of a prototype because the alphabetical menu increased the number of clicks while simultaneously reducing the time taken to get to the relevant information.

Task Six
You are a lawyer with an office in 5th Avenue and you want to take a client from Texas out to lunch downtown. You know she's a fan of steak, and you want to land this account to impress her. Pick a place and find the address. This task was intentionally formulated to include more than two criteria so that we could observe the different methods users employed in their search. Our findings also confirmed that unfamiliar users had a harder time locating the information and completing the task, which proved that the map wasnât a really a useful tool for them. The tricky part of this task was in superimposing information found in two different sections. Users were generally  frustrated that they could only search under ãcuisineä or ãnameä but not both and those who could completed the task relied mainly on background information.

QUIS Results
With so many attributes to consider, we calculated mean values for each design in each of the five sections (Overall Reactions to the Site, Screen, Terminology and System Information, Learning, and System Capabilities) to facilitate comparison. Most of the results averaged an 8 on a scale of 9 (the mean value for each attribute was 8.025).

Users were generally satisfied with the overall site, with the suggestion that we should streamline our categories for the ãcuisineä section and look into reducing the size of the map so that people with smaller monitors are not left with only a third of it on their screens. Our findings in the Screen section revealed that users liked having a predictable sequence of screens (and thus a predictable navigation path), a feature which boosted their confidence in the site. Similarly, the users also liked being able to control the amount of feedback they received, suggesting that users prefer not to be bombarded by an overwhelming amount of information.

Recommendations
Our findings are consistent with Vora's research (Vora, Helander, Shalin) which states that a graphical representation of hypertext with labeled links will be most beneficial to users from varying backgrounds. Both familiar and unfamiliar users took a shorter time in Designs 2C and 3 (prototypes with labeled icons) than Design 2A (icons only) and 2B (text only).

The sketch map was helpful to users when they were completing a task which required them to pick out a restaurant based on location. However, our findings proved that the map became a distraction when the users had to search for a restaurant by ãcuisineä or ãnameä. Our recommendations for improving the final site are as follows:
 

1. Use Design 1 as the prototype for our final web site design. The map was included as a main graphical element in Designs 2A-C and 3 because we hypothesized that it would help unfamiliar users navigate the site. Our findings show that the map became more a distraction when users had to find information not pertaining to location.

2. Include the map as an alternative way of presenting the second hierarchy of information under the ãlocationä link. Thus the final prototype will be a modified version of Design 1. It will include a map option under ãlocation.ä

3. Revamp our database by including all pertinent information about each restaurant in our final web site to reduce the difficulty of cross referencing ãcuisineä with ãname,ä. Each list will include the name, address, telephone number, area of the city in which it is located, and cuisine. We also want to explore the possibility of cross referencing more pages in our site. A person who clicks on ãnameä to get to Ruthâs Chris Steakhouse would also have the option of browsing other similar choices in ãdowntown,ä or the ãsteakhouse,ä categories.

4. Modify some of the labels for our plate links (the ãtable talkä link created the most confusion because some people thought it was a search function). The current links may be modified as follows: Restaurants by Name, Restaurants by Cuisine, Read or Submit a Review, Restaurants by Location, Restaurants on City Map. These labels should guide the user to the correct category, thereby reducing the number of ãmissesä he/she makes during navigation.

5.  Include a ãhomeä link at the bottom of every page so that users can return to the front page with ease.

6. Differentiate regional cuisines from type-of-food cuisines, since some of the categories were not distinct (Steakhouse may also be a subset of American Cuisine).

7. Arrange the lists in 6 vertically instead of horizontally; users have been conditioned to scroll through lists vertically.

8. Reduce the size of the map by at least 20% so that it fits on most screens.
top

Adding a Map Link
In order to give users the choice of using a map, we will add a "Map" link on the index page. This link will be small so as not to distract users, because testing showed that the map on the index page often caused users to take a long time looking at the map instead of trying to complete the task at hand.

Improving Restaurant Information
During testing, several users suggested changes that would make our web site more effective and efficient.  One of the biggest problems that users had was trying to match location information from a cuisine list and name list when choosing a restaurant.  For example, if a user located "Otto's" by clicking on "Cuisine" and then on "German", they would find information about the specific address of "Otto's", but nothing about the general vicinity of the restaurant.  We have added, therefore, restaurant information that informs users about the approximate location of the restaurant, information that corresponds to that of the "Location" pages.

Refining Cuisine categories
Users also found the category names under the "Cuisine" link somewhat ambiguous.  For example, they did not know what kind of food would fall under the category "Continental".  We then edited our category names for cuisine clarification and categorized every cuisine into three categories: American Food, Asian Food, and Other.  Users also tended to click on the first link that seemed to match their search interests; for example, when asked to find a steak house, some users would click on "American" because they didn't see the "Steakhouse" link at the bottom of the table.

Increasing Letter Size
While the alphabetical target lists at the top of the "Name" page were useful to many users, they suggested that the letters be larger to increase their noticeablility in comparison with the rest of the page. This change may cut down on the time that people will have to spend getting information.

Improving Link Labels
Finally, the "Table Talk" link proved to add some confusion to users' search.  Because "Table Talk" is unclear as to what kind of information lies behind it, we have decided to change the name "Table Talk" to "Read or Submit a Review " so the audience knows this is a place where they may take action, rather than a link to information and reviews.
top

Conclusion
In designing and testing the Pittsburgh Restaurants Online site, our goal was to allow users familiar with Pittsburgh quick access to the information that they need without requiring them to search through many unnecessary pages. We also wanted to allow users who are unfamiliar with Pittsburgh to circumvent links that are unnecessary to them. We understood that specific location links or local jargon would not work because many people would use the site never having visited Pittsburgh before.

The site design, therefore, required us to create a number of pathways that would give options to users of various levels of familiarity and who would use the site to achieve different tasks.

We conducted user testing to help us learn which of the five designs would be easiest for users both familiar and unfamiliar to achieve tasks. From user testing, we learned that each of the sites had strengths and weaknesses and the final site may be a combination of more than one of the designs.

We tested our prototypes to observe the usefulness of combining different forms of information, as well as the possibility of combining information in different hierarchical levels. Familiar and unfamiliar users may have been predisposed to process information better either through using graphical elements or textual elements.

A QUIS report revealed that although the graphics-based sites were visually attractive to users, they were difficult for both sets of users to navigate through. In contrast, Design 1 appeared attractive, easy to maneuver, and had no unexpected outcomes. The site also allows users to research the given tasks quickly and with freedom to navigate as they wish.

We decided to incorporate the results of our testing into Design 1 by making the following changes:

• adding a map link
• improving restaurant information
• refining cuisine categories
• increasing letter size
• improving link labels

These added changes, combined with the easy maintenance of the internet, make the revised Pittsburgh Restaurants Online an effective web site design as it caters to the needs of an audience with various levels of familiarity. We hope that they will take advantage of the specific features of the site to obtain information in the shortest time.
top