Pages

Sunday, March 29, 2009

Testing & targeting - an interview of Omniture's Nicolas Mériel

A while ago I had a chance to swim in the snow with Omniture's Nicolas Meriél. And for sure, we talked little bit about web analytics and especially testing & targeting.
"Testing can be really simple and cost efficient way to get great insights and to learn online customer behaviour fast."
In this post you can find the following questions of mine and answers from Nicolas:

What is testing?
How to do testing and how to start?
What choises there are for testing (A/B & multivariate)?
Why companies should test?
What is your best testing case ever?
Why do you like testing?
What kind of tools there are for testing?
What is the hardest part?
Other tips for testing?

What is testing?

Splitting your traffic (for example A/B versions) and showing each part a different experience or content on one or more pages, in emails or display ads. Back to content

How to do testing and how to start?

It typically starts with a hypothesis! Analyse and diagnose your web analytics data and try to formulate a data-driven hypothesis. For example: would more content for new visitors increase my conversion or will showing different content based on gender segmentation make more sense?

Once you have a hypothesis you will need to run a test to prove it to be right or wrong. This is where your online business optimisation starts. Back to content

What choises there are for testing?

You can run A/B..N tests, Multivariate Tests, testing ads i.e. banners as well as emails, landing pages and even run tests for mobile devices. However, the real power of testing lies in combining your test results with targeting. Once you have found a winning version you want to make sure that you serve i.e. target this to specific segments (for example new/returning or male/female visitors etc.) which will make your visitor’s experience much more relevant and increase your conversions. Back to content

Why companies should test?

Because “one size does not fit all” and different people like different things. This is true offline and as well online. There is a lot of evidence that shows that relevant content provides a richer user experience and allows to increase conversions. If you know that your customers are into Lordi, you probably would not play Katri Helena to them. Then again who knows, so make sure you test it, as any kind of test is better than guessing. ;-) Back to content

What is your best testing case ever?

This depends and can not really be generalised. We have many great testing case studies (many available free at Omniture.com). On some tests we’ve seen conversion lifts of over 1000%! Back to content

Why do you like testing?

First of all, it puts fun into your work! Beyond that it allows to eliminate tedious (often ego-driven) discussions about design, colours, beliefs, tastes etc. and allows to let your visitor decide! Ultimately and if done well, it can achieve what marketing is really about: providing the right product/content, at the right moment, to the right person at the right price. Back to content

What kind of tools there are for testing?

Obviously there is Omniture Test&Target. Besides that, there are different solutions from other vendors such as for example Interwoven Optimost, Maxymiser as well as Google’s Website Optimizer. Back to content

What is the hardest part in testing?

Having good ideas and developing dramatic different designs! Is there something you should definitely avoid? Avoid trying to squeeze in all into 1 test. Testing is a process. Apply a crawl-walk-run strategy. Multivariate Tests have become pretty en vogue as people believe this is the “coolest”. However, quite often a well designed A/B..N test with sophisticated segmentation and smart behavioural targeting might be the better approach as well as providing faster results.

Do you have any other tips for testing?

Start simple and learn to be wrong fast!
Back to content

Thank you Nicolas for visiting us and sharing your thoughts! : )