How do you test phrasing of labels, links, and descriptive text?

Question

At my workplace, an academic library, we often have extended conflicts over exactly how to phrase labels, links, and descriptive text.

For example, one link on our home page goes to a list of topics, mostly corresponding to academic departments (Anthropology, Business, Criminal Justice). Clicking one of those links takes you to a guide to available library resources on that topic. For many years, the link has been labelled "Resources by Subject".

Lately we've been using a product called LibGuides to produce and maintain these things, and there's an initiative afoot to rename the link to get the word "guides" in there somehow, on the grounds that we need to better promote them. But there is disagreement on how to phrase it. Suggestions have included:

• How To Guides
• Research Guides
• Guides by Subject
• Research Guides by Subject

This particular phrasing problem isn't what I'm interested in; it's illustrative. It's just that I've noticed that this type of conflict recurs with dreary regularity. We get multiple competing suggestions for very small phrases on the web site (typically no more than one sentence), and then spend days or weeks engaged in passionate debate as to which phrase is best for our users.

So the question is, how do we test something like that? If I had some way of gathering solid data demonstrating how the assorted options perform with honest-to-goodness users, perhaps I could lay some of this internecine conflict to rest. But I'm having a hard time figuring out how you design a test that zeros in on just one very specific factor like this.

Answer 1

Strategy 1

Design a simple user test and administer it with real people. Methodology depends on exactly what you want to test coupled with what resources you have to conduct the test. Here are a couple approaches to testing the example you provide:

Print five versions of the page with the link in question, each version with a proposed version, plus the current control. Gather test subjects, say students at the institution. If you have enough subjects (this could something you do in the cafeteria, spending 2 minutes with each person), put a version in front of them and ask them where they would click "to find a list of resources by subject."

Only show one version to each person and record where they point. You will learn 1) how well people can do this on the control, 2) if any of the proposed solutions perform better than the control, and 3) if anything else on the page performs better than either the control or the proposed solutions.

This is called paper prototyping. It's fast, easy, and the more data you gather, the more statistically significant it becomes. Granted, the test is taken out of context from the regular user environment, but it shouldn't queer your results much.

Strategy 2

Look over your site analytics to determine what search terms people use that land them on the resource page in question. Do any outshine the others? Do any performers match the proposed terms?

Also look at the SEO value of the terms proposed for what you are targeting (if search traffic matters to you).

Answer 2

Card sorting exercises is an alternative. Try reading: http://en.wikipedia.org/wiki/Card_sorting

What you get from card sorting with users, is data on how they understand and classify the words in their (i.e. your) domain. In the "open card sorting" mentioned in the wiki reference the users create their own names for the hierarkies identified. Card sorting is very simple to do, and you can use it with any number of users/stakeholders