Page tree
Skip to end of metadata
Go to start of metadata
Your Rating: Results: 1 Star2 Star3 Star4 Star5 Star 100 rates

GREYProposal on serving content based on recency decay bias.GREY


Serving content based on recency decay bias, means dynamically choosing from different versions of the content based on preferences shown by users over the time. Example of such could be a marketing campaign with multiple versions of same advertisement that is displayed anywhere in the page (or just retrieved to be displayed outside of the Magnolia based web). The advertisement is accessed over fixed URL and system (Magnolia) decides which version of the advertisement to serve. The decision is weighted based on tracking user response over the time (i.e. users clicking on one particular version of the advertisement served, increases weight of this version and version will be served to users more often). Since the customer groups and tastes for advertisement tend to change over the time, increase in weight of any version triggered by the user responding to that version, diminishes over the time.

Main Features

While this functionality doesn't require any special handling or implementing anything new in Magnolia backend, using it can be greatly simplified by providing unified way to configure such feature in Magnolia. The necessary support should consist of:

  • configuration of location where the alternative versions of same content are stored (sub pages or types in data module or children in DMS folder?)
  • the selection of implementation of weighting implementation
  • default implementation of weighting algorithm (storing info in Magnolia by default)
  • paragraph or page template, that could take the paragraph of choice (for rendering of the content above), render it and add backtracking information to the rendered output (javascript?)
  • UI to show and alter preferences for existing versions for each of the created content groups (if those are stored in Magnolia)

Open questions

While such features sounds hype and would certainly distinguish Magnolia, are they important and interesting enough that we should care (even if to show it only in a white paper) and/or support it?


  1. My 2 cents: there are related things we need to do first - for instance, implementing A/B and multi-variant testing, and segmentation. I have the feeling that the variant-testing, segmentation (showing different content based on the segment a visitor belongs to) and multi-language could all be solved with the same approach, because basically, it amounts to choosing which content to display for a given request. Your feature concept here might also be added under that umbrella.

  2. Sure, I just thought I put the idea down, while I had it still fresh in my head since I had to describe it for the client. Feel free to move it around to more appropriate location if you already have an umbrella page for such things.