How To Test Landing Pages
Testing is an important step of evaluating effectiveness of online advertising. In this post, I look at how to test landing pages for optimal conversion rates.
Multivariate testing involves testing the same landing page but with different elements on the page. Essentially, you are swapping out images, buttons and text when you are certain you’ve chosen the best design layout from your A/B tests.
It is important to consider statistical significance when testing landing pages. You want to make sure that you have served the landing page to a large enough sample size in each cell before inferring any conclusions. The company I work with, C.A. Walker, provides a sample size calculator on its site that is helpful for this. First, you enter the confidence level – typically 90% or 95% – then enter your maximum allowable error – typically 5% or 10%. If you’re not sure which to use, go with a 95% confidence and 5% error level.
Population size is a bit more tricky. You can try searching online for a population size, e.g. “population size of [target] in [your city or nation],” and see what research surfaces that you can use. If you are in the U.S., you can try finding your population size on the Census Bureau’s Population Estimates page. For example, if I want to geotarget an ad to Los Angeles women, I can see from the Age, sex, race, and Hispanic origin estimates under Population Estimates that there are 9.8 million people living in LA and 4.9 million of them are female. Plugging this into the calculator, the sample size is 384. Therefore, I would need to test each landing page on a minimum of 384 women in order to infer conclusions.
Another way to look at this is using Marketo’s Landing Page Version Calculator, which allows you to enter the number of expected conversions per day and your confidence level to get back either the number of test versions you can have or the length of the test you need to allow. If I enter 100 conversions per day at a 95% confidence level and 30 days for the test, it says I can run 13 versions of a landing page to generate statistically valid results.
Google’s Website Optimizer is a free tool you can use for landing page testing. To do so, you setup a “new experiment,” choose either A/B or multivariate testing, create and publish different versions of your test page at unique URLs, and then also create and publish a “thank you” page that users reach after they’ve completed a successful conversion. You then add tags to your pages so Google can track. From Google’s Adding Tags help page, “If you’re running a multivariate experiment, you’ll need to add tags to both your test page and your conversion page. If you’re running an A/B experiment, you’ll need to add tags to your original page, all test page variations, and your conversion page. You can find detailed instructions for adding all the tags in our Installation Guides for multivariate experiments and A/B experiments, as well as information about how to deal with tags in relation to shared headers and existing HTML code.” Website Optimizer will then provide you with a report on your tests to see how well each combination or variation is performing.
If the idea of creating multiple landing pages and managing tags sounds about as much fun as a root canal, then consider your paid options. I recently demo’d ion interactive’s LiveBall product, which does much of the work for you and is fully hosted on their site. Their pricing starts at $1,295 a month for unlimited number of landing pages and 10,000 visitors per month, and goes down if you get more traffic than that. For that price, they start you off with 12 landing pages complete with forms and images relevant to your industry; all you have to do is add copy. They also include 15 templates that you can use to change the look-and-feel of your pages on the fly. The templates are designed around 1 theme for each brand, for each season, or for each client (if doing this for multiple clients) and include 1 stock image gallery for each business (or you can use your own images).
If you want them to design your landing pages for you, they offer that service for an extra monthly fee or on an ad-hoc basis. They say some of their clients opt for the extra service for the first several months in order to build up their library and then do it themselves from there.
When you log in to their service as a client, the first thing you see is a dashboard to monitor real-time results of your landing page effectiveness. Your landing pages are represented by bubbles on a grid: the larger the bubble the more traffic you’ve received for that particular landing page, and the higher on the grid the better the conversion rate. In addition, the bubble are “live” meaning you can click on them for greater detail on that campaign.
To use the LiveBall system, you first create a “portfolio,” then “campaign,” then add a “path” to the campaign that is either new – chosen from the optional “frameworks” provided – or you can choose to copy a path over from another portfolio or campaign. A “path” is simply the series of pages that a person sees during the conversion process, which can be as simple as Page A and then the “thank you” page, or you can add intermediate pages that tries to upsell other products. I’m not sure how exactly, but was told their system also segments users to deliver robust landing paths/experiences.
To add a page, you select from the pre-determined master pages provided, change out the nonsense text and remove areas you don’t want to use. In addition, 3 forms are provided for you but are easily editable.
The last step is selecting the traffic source, e.g. Google, and it creates a unique URL for each traffic source for you. Here you can also choose to “auto-optimize” your landing pages, meaning that the system will determine when the pages reach statistical significance and pull poor-performing landing pages for you. You can also set it up to send people to another page when statistical significance is reached for that particular traffic source.
They also provide for you “advanced rules” (conditional logic) on the page or content level – 20 conditions, 15 actions. They are useful for geotargeting – substituting content on a page based on the region coming from – or previous action, such as adding the user’s specific search query into the page text or sending them to another page if they didn’t convert the first time they visited.
For reporting, you can view how a campaign is performing on a portfolio level, go into an individual campaign, or into individual paths. The dashboard is dynamic, meaning you can easily switch from viewing your results as a tag table, a respondent funnel, a pie chart, or other options. You can see other views there as well, such as which keywords are working best for you.
So that’s my primer on how to test landing pages and what some of your free versus paid options are. Some other resources I recommend on this topic are:
- 7 Best Tips to Create Landing Pages for PPC Campaigns
- The Cost of Landing Page Optimization
- Landing Page Testing: How to test and what to test
- Autonomy’s Best Practices for Optimizing Landing Page Performance, who have their own landing page solution
- SiteSpect’s Best Practices for Site Optimization, who have their own landing pages solution
Happy ROI hunting!