Update: This Tuesday Google officially launched the feature. Has it changed the way you search? Let us know in the comments!

We’ve been reading about it since October, but today is the first day that a few member of the BFM team came across the test being run by Google on search results, which show a preview of websites. Users can get a glimpse of what a website looks like before actually clicking on a result. It will be interesting to see what effect this has on click-through rates.

My guess is that better designed websites or ones with clear messaging will have an advantage – if searchers use the feature. They may just skip it all together, since it does not seem to work that quickly and requires the user to click on a small magnifying glass next to a result to activate the feature. Once activated, the user only needs to hover to other results to see previews.

Your website is getting traffic, but these visitors aren’t taking action. Is it time for a redesign? Before you scrap the whole design and start from scratch, consider if it’s only a few elements on the page that are hindering your success. And thanks to Google Website Optimizer, there’s an easy way to find out using real user behavior.

It’s a little ironic to talk about best practices when the whole point of testing is that websites are not one-size-fits-all; best practices for page layout, calls-to-action, and information architecture might not apply to your specific niche of users. Nevertheless, here are some tips for setting up experiments that should apply most of the time:

1. Don’t test too many things at once.

If you’re trying out different colored “buy” buttons, don’t try to test headline, text and product image variations at the same time. Unless your landing pages get an extremely high quantity of traffic, pick one or two variables at a time to avoid having your testing go on interminably to get a decent sample size, and then having to spend just as much time interpreting your mountains of data.

2. Get a decent sample size.

Don’t overanalyze in the first few days of testing. I’ve had variable combinations that looked like a sure bet in the first 48 hours, with a sample size of a few hundred, only to be left in the dust a few weeks later.

3. Be consistent.

While you want to do everything you can to improve your website, don’t scare users away by making changes on a landing page that will make them wonder how many designers you had working on your website if they decide to click through to the rest of your site.

If you’re already doing SEO or pay-per-click, provided it’s done correctly, you should have a steady stream of qualified visitors. You’ve already distilled an audience that is interested in what you have to offer; don’t squander this willingness to transact by leaving them to fend for themselves once they’re on your site.