Mike McDonald of WebProNews is at SES 2005 in San Jose, and took in an afternoon session on landing pages.

"If your traffic isn't converting well, all the optimization in the world won't help you,” Mike quotes Alan Dick as stating.

The session on Landing Pages featured something straight out of Who Wants To Be A Millionaire?: attendee feedback. Attendees were issued a card-sized remote, and could respond to multiple-choice questions posed by the speakers. Mike acknowledges missing one question, but claims it was a trick question. “It was actually pretty cool,” Mike said.

Mike summarized his thoughts on the Landing Page session:
Tim Ash, president of Epic Sky and its SiteTuners division, started his talk by referencing Cost Per Action (CPA). That equals CPC and conversion rate. While CPC is out of the advertiser's control, conversion rate is not.

The first step in site tuning involves price tuning. That requires finding a pricing “sweet spot” by tracking conversion rates and related product or service prices. Once the site has an optimal price to conversion ratio, you have successfully tuned your price.

Site owners may have to sample several layouts and designs to find the right combination of elements. Ash cites a/b split testing, which is easy to implement and tests one variable at a time.

A second approach, multivariate analysis or the “Taguchi Method,” tests several variables at once. And the third approach calls for engaging a specialist in site tuning, to handle those issues for sites.

Site owners should avoid making five mistakes while testing out landing pages:

1. Ignoring your baseline. You can't ignore your baseline traffic. Baseline traffic has to be accounted for when evaluating the conversion rate of your new landing page.

2. Be sure you have sufficient data. Remember the margin for error and random chance.

3. Compare aged data to aged data. Be sure you take into account the time period for the tests.

4. Assuming testing has no cost. Testing will result in some 'failed' experiments. Don't freak out if you have worse performance in early tests.

5. Ignoring complex Interactions To illustrate this point, Tim asked the attendees to picture an image of a Ferrari speeding down the highway with the marketing message "Ferraris are fast." Now picture an image of a car crashed into a tree with the same message. The impact that has on the reader is dramatically different.

Ash commented on how the CNN web site displayed ads for making reservations with Air France appearing on a page discussing that airline's recent accident in Canada. Strange things happen; site owners need to expect and account for them before making evaluative judgments on test subjects.

"There is always a better way. No matter how many times you've done it in the past, you can always make a change for the better. The problem is finding it,” Jimmy Rosche of Offermatica said during his turn speaking.

Rosche restated the importance of multivariate testing. By testing lots of variables simultaneously, site owners can find out quickly which methods work and which do not. He cautioned that the site's default page should be served to a significant percentage of site visitors during testing.

Too many landing pages may be seen by the search engines as examples of cloaking, which will get a site tossed out of indexes. Testing should be an ongoing process in site design. Those dollars spent on CPC will be wasted if a site doesn't have a landing page effective at converting visitors into customers.