How do you like the colours? How do you like the navigation?

January 19, 2009 6:47 AM

How do I do some quick user-testing for my website?

Hey gang, in the process of building a new website for my employers at the moment, and want to do some user-testing before this thing goes live.

What's a good way to go about doing this - we're thinking about asking two smallish groups to play with a beta, one group familiar with the company's work and one not. Maybe ask them to find some specific things, and then give them a survey asking how they found it to use.

Anyone got any tips/advise for a good way to go about doing this stuff? We've only got a week or so, but we're just looking for some general feedback.

Thanks!

posted by Cantdosleepy to computers & internet (13 comments total)

You should try using Amazon's Mechanical Turk if you want to get some quick feedback. The people who complete Turks are incredibly conscientiousness and will probably give you amazing feedback for a quarter each. For less that $20 you could probably get more feedback than you could use.

MeFi mail me if you want more information about how to get started Turking. It's a wonderful service.
posted by eisenkr at 6:56 AM on January 19 [2 favorites has favorites]

If you can, it is better to watch people than have them fill out a survey: you know what you are looking for and people are not always reliable self-reporters.

But yes, basic user testing would involve getting some folks to come to you, giving them a few tasks to complete (without any prompting!) and seeing what happens. If you can put together a little camera setup, so the note taking is remote, it can help folks feel more comfortable, but another alternative is to write down what you notice between users.
posted by dame at 7:25 AM on January 19

Stop what you're doing, go to a local bookstore and get a copy of Don't Make Me Think which covers cheap user-testing. And the book is short, so you won't waste time reading it.
posted by Wild_Eep at 7:26 AM on January 19 [1 favorite has favorites]

I have done user testing on and off for decades for a sort-of-living, so here's some really quick and general advice. For more, send me lots of money. ;)

No matter who you use, don't ask questions like your title suggests. The data you would get would be tainted by the question, and of very low value anyway. Who cares if sixty-seven percent of users like your colors, really? I hate MeFi's blue, for example, but that doesn't make the website less effective or "good". The survey will also be a mess unless its built by pros, since people really can't be trusted on surveys.

One cheap way to measure the usefulness of a website: give your users a short period of time, less than five minutes, and then close the window and ask real questions, like:

1) All right, so what does this company/organization do, anyway?
2) Is this company/organization large, small, or medium size?
3) Name another company/organization that this one reminds you of?
4) In what city is this company's office?
5) OTHER GOAL YOU ACTUALLY WISH THE WEBSITE TO ACCOMPLISH
6) What information did you expect but didn't find?

More thorough testing would ask such questions of first-time visitors before they see the site, then show them the website, and then monitor how they find (or fail to find) to the answer. The paths taken by users can be eye-opening if you have not been testing during design, and humbling if you think you got it right. When there is a strong pattern, wherever they think the answer should have been... well, that's where it should have been, guys. This works for things like online banking and hotel reservation systems just as well as it does for simple brochure sites. Watch people try to use/find something, and pay special attention to when and how they "fail".

If you don't have fancy monitoring software, something like clicktale.com can be a good substitute, if you have set it up in advance.

You don't mention technical testing, but someone had best use about 40 different browser/version/operating systems on it too, to be sure you don't have some nasty surprises lurking. Validating is not a substitute for testing. (There are some automated websites to help you test dozens of browsers, though.)
posted by rokusan at 7:27 AM on January 19 [3 favorites has favorites]

Seconding the Mechanical Turk idea. I did this about a month ago for a site I was launching, and got great feedback. I probably overpaid ($1 each, max 20 turkers) but every response was thoughtful and useful.
posted by stupidsexyFlanders at 7:30 AM on January 19

I read about UserTesting a while ago; for $20, someone will record a screencast of themselves using your website for about 15 minutes, and also provide a written report.
posted by helios at 7:31 AM on January 19 [2 favorites has favorites]

Thanks fror all the tips so far, guys!

Rokusan - I'm not worried about technical testing - we're getting that done by a proper web standards agency. The site is actually part of a larger different site which has been tested already in that area; this test is about the information on the site rather than the mechanisms powering it.
posted by Cantdosleepy at 8:13 AM on January 19

Figure out your objectives first. What questions do you want answered through testing. Can you answer them? For instance, can the user find the logo, etc.
posted by xammerboy at 9:13 AM on January 19

In the past when I've done simple user testing, we've sat a tester down with the site, given him an example task to do that's somewhat relevant to his actual interests, and asked him to speak aloud as he went about trying to accomplish it. It's amazing how much information you can get from something just this simple.

Hardest part = not helping the tester.

UIE has a bunch of very practical articles that will help you do exactly what you're describing. They have a bunch of great information available on usability testing, for free, and I even let them send me their newsletter because I love the articles that much.
posted by amtho at 9:56 AM on January 19

Okay, then, OP, that underscores the objective-meeting approach I and others have mentioned.

Forget the site for a day and go way back to your original reasons why you wanted to have a website in the first place. Measure "success" by users quickly finding/learning/doing the things you wished them to find/learn/do before you designed it. If you can get eye tracking (or at least mousetracking) that will give you great data along the way.
posted by rokusan at 10:03 AM on January 19

Seconding enthusiastically Krug's Don't Make Me Think, specifically the last 3 chapters. I use this as the "usability textbook" in my web content development classes and it works a treat. It even has sample scripts you can look at to see how to run a test. My students found it immensely helpful and it is a very quick read.
posted by media_itoku at 10:35 AM on January 19

Seconding Don't Make Me Think. All the basics of usability in a very usable book.
posted by echo target at 10:35 AM on January 19

Yeah, nthing DMMT, but note also you don't actually have to have a website to start testing. We sometimes get very interesting results just from showing people mockups on sheets of paper. They can't click but they can put their finger where they would click. Way cheaper than building a website and finding it's got some major issues.

I think Krug also recommends more than one round of testing. You can't find some small issues if there are really big distracting issues. You fix the big ones and you find the small ones in round two.
posted by AmbroseChapel at 1:54 PM on January 19