AMC playbook: Keyword clustering
In short: Keyword clustering groups search terms by the products (ASINs) they drive traffic and sales to. The playbook reads search term and purchased item from the AMC conversion tables, builds clusters in Python where ASINs overlap, and filters the most important ones with the 80-20 rule.
What it is about
The principle: certain search terms reliably lead to the purchase of specific products. Arranging them in groups makes large keyword portfolios manageable, campaigns more relevant and budget allocation more targeted. The playbook applies to sponsored products and sponsored brands.
The query aggregates KPIs by customer_search_term and the purchased item (tracked_item) with dimensions such as the marketplace.
Which questions the playbook answers
- Which search terms lead to the same high-revenue ASINs?
- How do you arrange large keyword portfolios into manageable groups?
- Which clusters deserve the biggest budget share?
Flow
- Collect data from AMC: conversion tables with search term and purchased item, through the workflow manager.
- Cluster keywords in Amazon SageMaker: start at the most relevant keyword, merge keywords whose ASINs overlap beyond a threshold, until all are grouped.
- Analyse with scatter plots and the 80-20 rule, filter out less relevant clusters.
- Upload the enriched data and act on it.
Prerequisites according to Amazon
- AMC concepts, AMC SQL and AWS basics, Amazon SageMaker is named.
- Proficient Python and a basic understanding of clustering methods such as K-means or hierarchical clustering.
Our take
The threshold for "overlapping ASINs" is a choice, not a constant of nature. In the evaluation it pays to look at how the number of clusters changes when you move the threshold.
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.