As Nielsen says in a 2005 Alertbox column, Putting A/B Testing in Its Place<http://www.useit.com/alertbox/20050815.html>, a large part of why A/B testing is successful or not comes down to the metric you're measuring to define that success. If there is no agreement on the value of that metric, it's unlikely to get adopted... Examples, however, of where it might work: Consider the suggestion to *add* a second link, or move a secondary link, to check account information on the homepage. If more people click that link (as measured via referrals or JavaScript) than use slower methods to get to the same page, you could win. If you time how long people take before they get to the account page, and it's shorter, then you've another win. But this assumes there's always an alternative to the change that's nearby, or that the user could ignore it if they choose. (E.g. color, left-or-right, spacing, order/ranking, image...) If not, I suggest restricting changes to a smaller subset -- for instance, those logged in to your site. Show certain accounts one thing, and other accounts another. This is especially useful for application functionality changes. You could even pull a Google Labs, if large enough, and allow people to opt-in to certain changes. Placing [x] buttons or collapsable heading arrows serves a similar purpose, and can tell you what features are used or ignored and when, if tracked. Ultimately, A/B testing, outside of advertising conversions, can be rather limited however, unless used in conjunction with other kinds of analysis and remote or local testing. Consider running a survey where participants opt-in to your questions and changes, so they know to expect things to shift around. People like surprises, if they ask for them and feel included in the end result. It's why people (like myself) run beta or dev copies of Google Chrome, so they can try the latest features and forgive you if things break. *Highly recommended reading:* (which I've bought as DRM-free PDFs and you can too!) 1. Driving Technical Change: Why People On Your Team Don't Act on Good Ideas, and How To Convince Them They Should<http://www.terrenceryan.com/blog/post.cfm/i-m-an-author-driving-technical-change>by Terrence Ryan (The Pragmatic Programmers) 2. Clout: *the ART and SCIENCE of INFLUENTIAL WEB CONTENT*<http://content-science.com/expertise/clout-the-book>by Colleen Jones (New Riders) 3. Remote Research: Real Users, Real Time, Real Research<http://rosenfeldmedia.com/books/remote-research/>by Nate Bolt & Tony Tulathimutte. However, don't expect much on A/B testing in any of these. As the third book says on page 127, it puts A/B testing (or what the nerds like to call multivariate testing these days) in a context of how to design your research, what tasks to perform, etc. For math, it recommends two books: * * *[By math, it refers to "number-crunchy regression/conjoint/factor/quadrant analysis, optimization, or any of the multivariate stuff you learned in AP Calc."]* * * *Measuring the User Experience* by Tom Tullis and Bill Albert (Morgan Kaufmann Publishers) "In this book you'll learn all about the kind of data analysis we sheepishly avoid." http://measuringuserexperience.com/ "The other book, which goes into fine-grained, advanced automated research techniques is called *Beyond the Usability Lab: Conducting Large-Scale User Experience Studies* by Bill Albert, Tom Tullis, and Donna Tedesco (also tech. ed. for Remote Research)." http://www.beyondtheusabilitylab.com/ Both of the above websites together have a ton of resources and examples. Ultimately, however, it's up to you to influence people and processes to improve both content and design in measurable ways. It all seems to come down to politics regardless. Good luck! Louis. On Wed, Jan 26, 2011 at 6:09 PM, Bill Dueber <[log in to unmask]> wrote: > I've proposed A/B testing for our OPAC. I managed to avoid the torches, but > the pitchforks...youch! > > On Wed, Jan 26, 2011 at 5:55 PM, Sean Moore <[log in to unmask] > >wrote: > > > There's a lot of resistance in my institution to A/B or multivariate > > testing > > any of our live production properties (catalog, website, etc...). I've > > espoused the virtues of having hard data to back up user activity (if I > > hear > > one more "well, in my opinion", I'll just go blind), but the reply is > > always > > along the lines of, "But it will confuse users!" I've pointed out the > > myriad successful and critical business that use these methodologies, but > > was told that businesses and academia are different. > > > > So, my question to you is, which of you academic libraries are using A/B > > testing; on what potion of your web properties (catalog, discovery > > interface, website, etc...); and I suppose to spark conversation, which > > testing suite are you using (Google Website Optimizer, Visual Website > > Optimizer, a home-rolled non-hosted solution)? > > > > I was told if I can prove it's a commonly accepted practice, I can move > > forward. So help a guy out, and save me from having to read another > survey > > of 12 undergrads that is "proof positive" of changes I need to make. > > > > Thanks! > > > > *Sean Moore* > > Web Application Programmer > > *Phone*: (504) 314-7784 <tel:+15043147784> > > *Email*: [log in to unmask] > > > > Howard-Tilton Memorial Library <http://library.tulane.edu>, Tulane > > University > > > > > > -- > Bill Dueber > Library Systems Programmer > University of Michigan Library >