Tutorial Details

Download: Schedule of Events (PDF) subject to change
By Day - Selected
By Presenter
By Track
View My Agenda

Card Sorting and Cluster Analysis for Information Architecture

A challenge of web site and application design is to organize large amounts of content so that a user can easily navigate. This tutorial will provide UX practitioners with practical guidance and hands-on experience with card sorting and cluster analysis methods and tools to create an information architecture.

Tutorial by Carol Righi (CarolRichi.com), Janice James (Perficient)
Usability Fundamentals (UF)
8:30am to 5:30pm on Tuesday, June 09, 2009

About the Tutorial


Time Topic or Event 30 Startup, Introductions 15 Overview (session objectives) 15 What is a card sort? 30 Why are card sorts useful? 120 Mechanics of performing a card sort (Exercise) 120 Analysis of data (Exercise) 60 Group presentations (Audience participation) 45 Tools 15 Limitations and caveats

DETAILED DESCRIPTION OF TUTORIAL 1. What is a card sort? This introductory segment will begin with a very brief demonstration of a manual card sort using index cards. A couple of attendees will be asked to sort items into categories, and name the categories. This exercise is simply intended to “get everyone on the same page” with regard to the essence of the card sort. We will then cover a brief history of the use of the card sort in information architecture. We will describe precursor methods to the method that will be described in this tutorial (such as Bridge, and physical card sorts using index cards or sticky notes). We will then complete this segment with a discussion of the use of card sorts in current practice.

  1. Why are card sorts useful? In this segment, we will discuss the various applications of this methodology, including the following: • Organization of menu structures for both web-based and desktop applications. • Content organization for information-rich web sites and desktop applications. • Organization of the content of documentation.

This section will include discussions of case studies of projects for which card sorts have been applied, such as Cabelas.com and UAB.edu (University of Alabama at Birmingham). The case studies will address the objectives of the web site, why a card sort was selected as a method to use to achieve those objectives, and at a high-level, how they were performed and applied.

  1. Mechanics of performing a card sort. This segment covers the step-by-step method for performing a card sort. We will use WebSort, a web-based tool, to give attendees hands-on experience with performing a card sort using WebSort and analyzing the data. We will provide the data that attendees will work with through this segment. The steps of the process will be clearly enumerated and explained in their proper sequence. These steps will include:

1) Define the audience. We will discuss why it is critically important to define the audience for the web site or application, and address issues such as dealing with multiple audiences.

2) Recruit. As with all aspects of UCD, it is necessary to find a pool of participants who are potential users of the resulting design. We will discuss issues such as how to recruit, how many to recruit, and reimbursing participants. We will show some examples of recruiting tools.

3) Inventory the content. We will discuss the need to develop a complete inventory of a planned or existing web site or application, and show some examples of how best to develop the inventory. We will address issues such as how to avoid bias when providing descriptions for each of the items, how to determine the depth of content to include, and how to determine the appropriate terms to use for topics to avoid reflection of the current content for existing sites/application.

4) Input the content into WebSort. We will demonstrate how the study administrator sets up a WebSort study. We will demonstrate how the “cards” are created, the mailing list is prepared, and other aspects from the point of view of the administrator.

5) Use WebSort. We will engage the attendees in en exercise where they will be the participants in a card sorting study. We will have prepared a set of “cards” they are to sort. All attendees will perform this activity.

  1. Analysis of data. This segment will address the second phase of this process: analysis of the data using cluster analysis. We will begin with a brief introduction to Cluster Analysis, which will include a conceptual discussion of the statistics behind the method. We will discuss how it has been used in the past, especially in marketing studies. We will contrast it to the “eyeball” method typically used with the data generated by manual card sorts.

We will then return to our case study in progress. We will demonstrate the WebSort cluster analysis tool and explain how to use it. Attendees will then have an opportunity to use the WebSort tool to analyze the data. We will form teams of attendees. The teams will manipulate the tool to generate categories. Each team will be asked to create a basic architecture (categories and labels). We will have each team present their results, sharing the thought process that led them to their design. We will discuss how the same set of data can potentially generate multiple versions of an architecture. Other issues from the attendees’ experience will be discussed.

  1. Tools. This segment will cover some tools that are used to facilitate the process of the card sort and cluster analysis. We will discuss the use of: • EZSort, WebSort, and others.

An overview of the card sorting tools currently available.

• EZCalc and the WebSort cluster analysis tool. An overview of the cluster analysis tools currently available. • Excel. How spreadsheets are used to organize and sort the information in a content inventory, track recruits, and display results. • Visio or other graphics applications. Simple logic flow diagrams used to depict the site architecture. • Wireframes and site skeletons. Interactive .html pages used to show the architecture “in action.” • Surveys. Recruiting, priority studies, etc., performed prior to a card sort can be performed using survey tools.

  1. Limitations and caveats of card-sorting. This segment will cover some of the current issues around card sorting, such as variations in the methodology, limitations to the current practice, and possible directions for the future. These include:

1) Open vs. closed card-sorts. Whether participants are allowed to introduce new items will change the manner in which results are interpreted.

2) Renaming content. Because terminology can affect usability, it is often valuable to have users name the content as well as the categories. We will discuss when this is and isn’t recommended, and how to alter the methodology to accommodate this step (i.e., getting user feedback on the labels of the items after they've completed the sort).

3) Users’ thought process. Asking users to articulate their rationales while or after they’ve completed the sort can provide more insight in to the user’s mental model for the content, and can be helpful in developing the architecture. 4) Multi-level sorts. Currently, card sorts allow only one level of categorization. A variation in the methodology can help formulate lower-level categories into higher-level categories. 5) Card sorting and cluster analysis for large-scale information systems. In addition to “classic” web and application design, this method can and has been used for large systems (Dong, 2002).

6) Remote vs. face-to-face studies. We will discuss advantages and disadvantages to each.

7) Individual vs. group studies. We will discuss the implications to multiple individuals performing the sort to obtaining real-time group consensus by having groups perform the sort.

8) How many participants are needed? We will discuss a recent study by Tullis and Wood (2004) that provides guidance on this topic.