A customer starts buying less often and opening fewer emails. Machine learning (ML) helps marketers spot such patterns in existing data and predict what may happen next. Its value is not an automatic fix, but better-informed decisions about whom to engage, when, and with what offer. section 1.6.2
Common marketing uses include:
- Predicting which customers may leave or have high lifetime value. section 1.6.2
- Recommending products from browsing and purchase patterns. section 2.3.3
- Refining audience segments, ad targeting, and campaign timing. section 1.6.2 section 2.3.3
Retail retention
A retailer used ML models to detect subtle changes in purchase frequency and email engagement. The team then adjusted email timing and personalized subject lines using past purchases. The improvement came through a series of small retention changes, not an overnight transformation. section 1.6.2
ML differs from generative AI: ML is used here primarily to find patterns and make predictions, while generative AI can create content such as personalized emails or ad visuals. In a marketing data architecture, ML belongs in the insights and intelligence layer, where governed data becomes predictions and recommendations. First-party data and privacy practices matter to that foundation; secure clean rooms are also being explored for models trained on shared, anonymized data. section 1.6.2 section 2.3.3 section 8.5
