AMC playbook: Customer long-term value (CLTV)
In short: CLTV sets a customer's revenue against the cost of acquiring and retaining them. The playbook shows how to define these inputs in AMC, analyse the customer base by CLTV, build audiences from it and activate them in Amazon DSP. Here "long-term" means one year.
What it is about
Amazon contrasts CLTV with ROAS: ROAS measures the short-term effect of a campaign and counts only ad-attributed revenue; CLTV looks at a customer's value over the relationship, including repeat purchases and non-ad-attributed conversions, and accounts for acquisition and retention costs.
The playbook defines long-term as one year because according to Amazon AMC allows only one year as the maximum window. It recommends including the paid features shopping insights datasets so that non-advertising-driven purchases, such as Subscribe and Save, count as well. Without a subscription CLTV can still be calculated and activated as a rule-based audience, but then only shows ad-attributed revenue.
Which questions the playbook answers
- How do you define cost and revenue inputs for CLTV?
- How do you analyse the customer base using CLTV?
- How do you segment audiences by CLTV?
- How do you activate the segments in Amazon DSP?
Flow
- Identify the inputs: cost and revenue, duration of the relationship, structure of the audience segments.
- Query data from AMC: aggregated CLTV and segmentation at
user_idlevel. - Analyse audience value and activate the segments in Amazon DSP.
Prerequisites according to Amazon
- AMC concepts (aggregation thresholds, privacy limits) and AMC SQL.
- Basic Python and AWS knowledge, Amazon SageMaker and Amazon Athena are named.
- Basic Amazon DSP knowledge to activate the audiences.
Our take
If you need 60 months of purchase history you will run into the one-year window of the playbook. The Amazon Retail Purchases dataset covers 60 months, but its time window does not match the 13 months of advertising data in the other tables; Amazon warns about wrong conclusions when joining.
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.