Getting Multivariate Landing Page Testing Straight!

These days, you can’t go anywhere without hearing about multivariate testing
so I thought I’d walk through a landing page test. Along the way, I’ll
provide testing pointers and leave you with several testing considerations.
I’ll highlight the importance of multivariate testing and show it can yield significant improvements in
pay per click (PPC)
accounts.

This case study centers on NXPowerLite, which sells compression software
for PPT, Excel, and Word documents. While too many details are
involved in multivariate testing to cover them all, the following information should
provide a good basis for your own testing.

Multivariate Testing – Pre-Testing Considerations

Consider the following before starting a multivariate test:

Overall page quality
Initiate tests with quality content pages to yield more significant
results. Consider all aspects of a landing page, including images, page layout,
and copy for quality. In terms of NXPowerLite, the old
landing page didn’t do a good job of spelling out specific product features
and benefits, nor did it a have a suitable call to action. As a result,
visitors were “undirected” and didn’t know what action to take on the
landing page. On the new pages, product features and benefits were clearly
displayed and had clear calls to action.

Solid call to action
In our analysis, we determined that 45 percent of the 30-day free-trial
product customers became paying customers. As a result, we decided to
include two buy options on the page – one to purchase the product directly
and the other for a 30-day free-trial product. This was not to give people
choice, but to make the free trial version of the product look more
favorable. The “buy now” helped the trial version become the “no-brainer”
option as it subtly encouraged visitors to sign up for the trial version of
the product.

Solid campaign goals and appropriate metrics
In this particular campaign, we decided to track the following conversion
events:

• Direct sales
• Sales from the 30-day-free-trial product

What
you track in a campaign doesn’t need to be ultra sexy. The important thing
is to ensure you identify solid goals and determine appropriate metrics for
PPC campaigns (and any other campaigns, for that matter).

The Multivariate Test: NXPowerLite.com

With testing, NXPowerLite’s “compression software” campaign saw significant
improvements in campaign conversions. In our tests we tested a number of
page elements, but the ones where we saw the biggest impact were:

• Page copy
• Button size
• Button placement on page
• Other images (placement)

In our test, the page below converted at approximately 27.5% percent. This
page had five features & benefits and fewer graphical images. Note: the top
parts of the pages below (navigation bar) have been excluded from the
screenshot.

When we ran the test slightly longer, we found that a shorter page converted
at approximately 36.6 percent. The page had few features and benefits and a
more prominent image in the lower right-hand corner (see below). Note: the
page converted 100% better than the one in the old campaign. There is an
important lesson here – to get sound data, be sure to run tests for a long
period of time. In the above example, had we not run our test for a long
time we would not have seen that the second screenshot outperformed the
first one by approximately 33 percent. In this case, we had statistically
sound data in one month, but you may have to run tests for a longer period
of time (statistical significance is based on volume).

Multivariate testing considerations

Here are a few more testing considerations:

Before creating landing pages, think specifically about who your customers
are and try to gain some insight into how they shop. Look through your web
statistics for trends. Also, look to sales data or previous tests that have
been run for additional insight. The more you understand about your buyers
from the get-go, the more likely you’ll understand what your buyers “need”
to make a purchase. In our example, we knew 45 percent of trial users became
customers and were able to build on this success.

Don’t forget to keep testing. And don’t stop once you think you’ve found a
winning formula, as there could be a better formula out there! Even in
accounts that have been running for more than four years, we continue to increase conversion rates because of
continual testing.

Recently, multivariate landing has become a lot easier and
affordable. An obvious example of this is
Google’s Website Optimizer tool,
which enables you to allocate how much of your traffic you want to go to
your test (for example, 20%) and will show which elements
are negatively affecting your campaigns. You owe it to your campaigns to
give testing a try. You may be surprised at how much it can improve
conversions.

Mona Elesseily is director of marketing strategy at Page Zero Media, focusing on paid search campaigns and conversion improvement. She’s also the author of Page Zero’s Mastering Panama: A special report on Yahoo!’s new search marketing platform (August 2007). The Paid Search column appears Mondays at Search Engine Land.