Previous: Visualize your data’s usage and appearance in reports BEFORE you configure Google AnalyticsNext: Hello RJMetrics!

To Unify or Diverge? That is the question.

I fancy myself as having a “Jr Developer” badge.  It’s not a leap – I have a Master’s in Computer Science from a totally-challenging-and-enlightening-but-not-particularly-practical program (Stanford Univ) and I’ve spent the better part of the past 15 years in pretty technical environments leading product, marketing and/or analytics.  And I taught (and loved!) Databases & Algorithms during my graduate school years, to boot.

So it’s a surprise to me when problems in the arena of technical architecture and implementation that seem like they should be solved by now aren’t.  Case in point – to unify code across multiple landing pages or to diverge?  Any person who has spent a few months in the world of A&O knows that the answer isn’t either extreme.  All your good optimization principles say to customize your LPs based on ad copy or user target, etc. to get those precious lifts in conversion rates.  Great.  But then you end up spawning lots of LPs that win…and you forget to delete the ones that didn’t.  So then you have these “landing page cleanup” projects because, oh by the way, you forgot to apply data-validated lesson #12 to previously existing LPs, so they get out of sync even more.

And as part of this project, you aim to “consistify”.  (a word I think I made up in the late 90s)  There’s consistifying that which is seen by the user, which often (correctly) leads you to consistify (unify) the code.  But you don’t want to over-unify because well…you need to diverge based on ad copy or user target, etc.

Great, great.  You figure this all out, and do a build that cleans things up.  Yay.

And then maybe a month or two later, you want to launch a Google Content Experiment on just the one version of your “no form” landing page, and OOPS – a ton of different “no form” landing pages share the header code and suddenly you don’t know what to make of your data.

This is no one’s fault, and sure you ought to have good processes for asking exactly which LPs should be touched, and testing afterwards, etc.  But really, how many of you that have been doing this for years still run into this problem?



  • Tristan says:

    linkspam linkspam linkspam

    Just kidding, I have a real comment.

    I recently went through the process of “consistifying” (love the word) RJMetrics landing pages. It is painful, I agree. However, this time, in the process, I implemented the reasonably new Google Tag Manager. I made separate GTM tags for different types of landing pages, and then incorporated the appropriate code into each.

    While this certainly doesn’t solve all the problem, I’ve found (so far) that the additional layer of abstraction, and being able to work within a management tool as opposed to copying/pasting javascript snippets, has been very helpful.

    Maybe not a solution, but this is probably one of those headaches where there’s no such thing as a cure, just pain relievers.

Leave a Reply

Your email address will not be published. Required fields are marked *

Previous: Visualize your data’s usage and appearance in reports BEFORE you configure Google AnalyticsNext: Hello RJMetrics!