Customer lifetime value with AMC: cost, revenue and 60 months of purchase history
In short: CLTV is a customer's revenue minus the cost of acquiring and retaining them. Amazon's playbook uses one year as the long-term window in AMC. The Amazon Retail Purchases dataset reaches back 60 months, so it allows longer cohorts, but its time window does not match the 13 months of advertising data.
The formula and its levers
The simplest form according to Amazon: CLTV = customer revenue minus the cost of acquiring and retaining that customer. The playbook assigns the levers to two sides:
| Side | Lever | KPI |
|---|---|---|
| Revenue | Increase duration of the relationship | average customer long-term value |
| Revenue | Increase upsell and cross-sell | average purchase revenue |
| Revenue | Increase buy frequency | purchases per customer per month |
| Cost | Optimise awareness cost | spend per customer towards video views, page views, impressions |
| Cost | Optimise consideration cost | spend per customer towards detail page views |
One year or five years?
- The CLTV playbook defines "long-term" as one year because according to Amazon AMC allows only one year as the maximum window. For a predictive approach it points to the year-over-year segmentation.
- Amazon Retail Purchases (ARP) holds up to 60 months of purchase history in the Amazon store, regardless of advertising. That makes cohorts over several years possible.
- Amazon's warning: according to the ARP page the other datasets hold only 13 months of advertising data. Joining ARP with them without aligning the time windows can lead to wrong results. Since November 11, 2025 Amazon has extended the ad traffic lookback to 25 months in some marketplaces (including DE); check the window of your instance.
Approach
- Define the inputs: define the cost and revenue model (this is at your discretion according to Amazon) and map ASINs to broader product categories.
- Determine the duration of the relationship and the distribution of customer value, derive audience segments from it.
- Query aggregated at
user_idlevel, analyse the segments, activate high-value segments as a rule-based audience in Amazon DSP. - Include the shopping insights datasets so non-advertising-driven purchases (for example Subscribe and Save) count too. Without them CLTV shows only ad-attributed revenue.
Frequently asked questions
Is CLTV better than ROAS?
They serve different purposes. ROAS measures the short-term effect of a campaign with ad-attributed revenue, CLTV the value of a customer over the whole relationship including costs. Amazon presents CLTV as a complement, not a replacement.