Example Blender Allocation Diagrams
System diagram of how Blender allocation rules combine relevancy, performance, and advertising objectives.
Allocation rules get complex, particularly when mixing delivery systems like ads and organic items and combining objectives like engagement, profitability, semantic relevance, and ad revenue. A simple "quality score" ranking is insufficient to capture these nuances.
Promoted's Blender system uses multi-dimensional sort per allocation slot configured using the Blender Domain Specific Language (DSL). The parameters to Blender rules themselves can be optimized using the Hyperloop system to optimize for user responses like lifetime value or optimal ad load.
The "sort order" abstraction is easy to reasonable (in comparison to a nest of conditional allocation logic) and gracefully falls back to lower priority items when a higher level condition cannot be filled (e.g., no ad is available, so fill with the next highest value organic item). It can also be expressed in Promoted's spreadsheet-based Introspection Tools
An simple example of multi-dimensional sort order that first sorts by semantic relevance, then by expected revenue. A few reserved slots have an additional higher dimensional priorities for novelty or promotion. This is a typical "semantic relevancy trimmed" example.
A more sophisticated diagram of how different Blender rule dimensions can interact to compose the optimal allocation.
Updated about 2 months ago