AMC playbook: Optimal frequency analysis
In short: The playbook measures how many impressions a user received and how many conversions followed. From that it derives frequency caps, finds audiences with too few contacts and the point where extra contacts bring little return. It uses the most recent 90 days.
What it is about
Without frequency optimisation you overexpose your audience (ad fatigue, wasted budget) or expose it too rarely so it forgets the brand. The playbook assesses the business value of raising or lowering frequency and shows where budget goes to overexposure that could be redirected to new audiences.
The example queries group by campaign ID; according to Amazon any tactical dimension works, such as advertiser, audience segment or entity. Frequency is measured daily, attribution follows the logic of the amazon_attributed_events_* tables.
Which questions the playbook answers
- How do you measure frequency in a user base for your use case?
- How do you find the optimal frequency point algorithmically?
- How do you find under-served frequencies?
- How do you find the point of overexposure?
Flow
- Determine and set frequency caps.
- Identify and activate under-served audiences.
- Determine the optimal frequency point.
- Derive actions from the analysis.
Prerequisites according to Amazon
Our take
The analysis needs enough campaigns and days for the frequency buckets to clear the aggregation thresholds. For small accounts it is therefore sensible to check the thresholds before deriving caps from the numbers. A cap is a hypothesis and should be tested.
Amazon documents the playbooks in technical English with SQL and Python. This page summarises what the playbook does; the queries and scripts are in the original.