MVT using Publishing Targets

steliodibello Personalisation, Sitecore, xdb, , , ,

I have been involved in a few enterprise projects where unfortunately I cannot easily enable XDB architecture and fully leverage Sitecore capabilities for MVT testing, therefore, it comes to the idea of getting creative with what I got and find an alternative way of helping the marketing team testing the effectiveness of their new content and the new functionality that they have in mind….

I guess that it is pointless to mention, the advantages of the Testing culture and the idea behind fail quickly… however all these principles & features are covered with Sitecore MVT and in case you have XDB enabled in most of the cases I would strongly suggest you to use the built-in features…

In the specific project circumstances, I have been asked to test the effectiveness of some new “label” within the Booking funnel to see the impact on conversion….

For the Tracking, we were not using XDB and Sitecore analytics but Google Analitycs do a good job tracking Conversion and events…

For trying the different variants, I have split the sitecore change across the 4 Publishing Targets that I have for scalability reasons…  you can read more about publishing target here

My architecture requires a CDN / Proxy (Akamai in my specific case) to split the traffic across my 4 Content Delivery servers and have a publishing target different for each of the Content Delivery server…

Each Variant can be assigned to one or multiple CDs so that you can have a simple A/B testing or a more complex MVT….

With this architecture and this setup, I was able to allow the marketing team to test their content experiment and help them to find the best variation and the winning experiment even without having to configure Sitecore analytics, XDB and so on, but just leveraging the existing architecture….

The change that you would need to implement is to send via a custom dimension to GA the name of the Content Delivery server for each page view / event / transaction…

the code to track the CD, would be as simple as:

ga('set', 'cd1', 'Level 1');

obviously, this approach comes with some trade-off….  since Marketing users cannot really know which test is running and risk to override a running test doing a simple content publishing…. Also from the reporting point of view, you would need to keep track of the publishing that your content team is doing to be sure that anything else is affecting your experiment…. Also, need to remember to switch off your experiments and align all the content on the publishing targets to have a consistent web site….

 

Untitled Diagram

3 thoughts on “MVT using Publishing Targets”

  1. Thanks for sharing, it’s interesting approach that you’ve done here.

    Given the caveats that you mentioned at the end of the post, do you reckon it was worth it given the complexity it brings to have an MVT capability?

    I imagine the content authors needs to be quite savvy to be able to avoid making those mistakes and every MVT scenarios they want to implement will be heavily dependent on the implementation partner support?

    Like

    Reply

    • Hi Rey, in my opinion nowadays Content Authors and Marketeers in enterprise scenarios should be always testing content variations to see the impact on customer journeys and conversion… obviously, this complexity increase the workload & skill set for Content Authors & Marketeers…. in the environments that I have been working MVT is not done from the implementation partner but either from the internal authoring team or from creative agencies… Obviously, if the end client is not too big, they would just not bother about it, but have one version of the content and hope it is the “best one” but all these considerations are outside of the IT scope and it is a wider business consideration if the customers got capacity for digital optimisation….

      Like

      Reply

Leave a comment