What Makes Rich Media Ad Optimization So Difficult? A Quick Overview Of Top Complexity Drivers At Work

Loch Rose, head of analytics here at ShopLocal, wrote up an overview of what is easier and more difficult as it pertains to optimization within rich media display ads.  It was great, so I’d like to share it.

To get to that elusive point where every possible content variation is mapped to

The difficult challenges in optimization:

• Optimizing based on user level data (e.g., BT data) as it can be very diverse, contain many missing values, and all of the related publisher-specific integration challenges.
• Optimizing for a combination of listings, e.g., selecting the best several listings (can only be solved via a 2-phase process, e.g. first optimizing by category, then by listing within category).
• Combining rules/constraints with optimization, e.g. for rich media, jointly optimizing the banner and the panel by selecting the best listings for the expanded panel that are consistent with the content of the unexpanded banner.
• Setup integration which encompasses incorporating multiple data feeds and creative elements.
• Optimizing around goal orientated performance metrics due to the overwhelming amounts of data to crunch through complex models. This requires efficient coding and/or a sample-based approach (i.e. only crunching a subset of  the overall available data).
• Getting all of the mechanics to work which involves serving all of the experiments (e.g., permutations) and then serving storytelling up the experiments that were more successful, in near-real time.
• Providing comprehensive measurement and reporting, including reporting on how the optimization itself is working.

What’s not so difficult in optimization:

• Optimizing on advertiser-controlled variables such as creative types and list of publisher.
• Optimizing for a single listing.
• Overall campaign tracking.