Information architecture IT

1

Data Analysis Method with Using Optimal Sort

Note: The following example is based on a website redesign project of an IT department

I. Identify areas of overlap in participant generated categories

This process makes it possible to compare categories created by participants based on their similarity

and difference and suggest new effective categories.

Step 1. Under Categories, identify similarities in titles and, more importantly, content of those

categories. Just because similar phrasing or words have been used does not mean that the content of

categories is the same. Be sure to expand the category by clicking on the plus icon to the left of the

name to see cards placed under each.

Step 2. Create a new category by using Standardized Selected Categories at the top of the screen (Pic 1).

Be sure you have highlighted ONLY the categories you want to combine before proceeding. You will

need to give the ‘new’ category a name (Pic 2).

Picture 1.

Picture 2.

Step 3. The data associated with cards placed in these new categories will now be combined (Pic 3).

2

A new category Total number of cards included

Cards placed Frequency Participant in category of association supplied label

Picture 3.

II. Data analysis

Sometimes, you collect data from different user groups. In Optimal Sort it is possible to view data

generally or by a user group. To do that, you need to go through the following steps:

Step 1. Go to Participants tab

Step 2. Click select based on questionnaire

Step 3. Decide whether you want to select or deselect participants who answered certain survey

questions with a certain answer, e.g., Select participants who answered ‘What is your role at MU’ with

‘student’ (Pic 4).

Options for selecting or deselecting participants

Options for Specific question you group you want question completion want analyzed the question applied to

Picture 4.

Step 4. Click Update.

Step 5. Scroll all the way down to the bottom of the page and click ‘Go’ next to the Reload results from

selection box (Pic 5).

3

Picture 5.

After you come up with a manageable number of new categories based on the similarity of the content

of participant generated categories, you need to validate the strength of relatedness among the cards.

To do this, you can use a number of tools, e.g., the similarity matrix, the dendogram, and the

standardization grid.

The Similarity Matrix shows how many participants agree with each pair combination of cards. For each

possible pairing of two cards in the survey, a count is provided at the corresponding point in the matrix.

The count describes how many times the two cards were placed in the same category by all participants.

The darker the cell, the more probability there is that these two cards should belong under the same

category (Pic 6).

Shows how many participants are Shows total number of represented participants

Picture 6.

This cell shows the

frequency of ‘Desktop

software support’ and

Tiger Tech being placed in

the same category

together

4

Standardization Grid shows the distribution of cards across the standardized categories you have

defined. Each table cell shows the number of times a card was sorted into the corresponding

standardized category (Pic 7).

Looking down a column allows you to assess which cards were placed into a particular category.

Looking across a row allows you to compare how frequencly a card was placed into different categories

Picture 7

Dendograms are used to demonstrate data clusters. Read more about interpretation of dendograms

here: http://www.optimalworkshop.com/help/kb/optimalsort/interpreting-optimalsort-dendrogram

Note: During the data analysis process, you probably have noticed that the card sorting is not a

completely objective and precise technique. Many times, you need to make some subjective

decisions. Different participants will sort cards in different ways, so the cluster analysis and

dendrogram won't produce the ideal information architecture for you. However, OptimalSort can

help you start examining the patterns and the strength of the relationships. In order to pull

together a suitable grouping of items, you need to refer to the Cluster analysis, the Dendrogram,

the raw sort results plus participants provided grouping names. In the real world, you also need

5

to under the business rules or constraints which imply certain items have to go in certain places.

Sometimes, organizational politics will also influence the decision of a good information

architecture.

Certainly, you can also use the data and results from your card sorting study to help convince

development and managerial teams to make changes of a site. That is like the example I gave

you in the lecture notes—you may propose new IA structure. Overall, you need to apply a

combination of knowledge from the cluster analysis, dendrogram, and participants’ comments

during the sort, in order to create a good start of information architecture. You also can use other

data you have, such as usability studies and web logs to inform your analysis. Please

remember—this statistical output is great reference sources that you backup your decisions, but it

is not the only sources that you can rely on.

Due to the limited time of this course, we only learn the open card sort. In future, you can run a

reverse sort/closed card sort to test your new site structure/hierarchy.