Editing AI for Marketing Analytics (section)

== <span style="color: #FFFFFF;">Understanding</span> ==
Marketing AI operates at three levels: **measurement** (understanding what happened), **prediction** (forecasting what will happen), and **optimization** (deciding what to do).

**CTR prediction at scale**: Google, Meta, and Amazon each serve billions of ad impressions daily. Each impression requires a real-time ML inference: given this user, this context, and this ad, what's the probability of a click? Models must be fast (milliseconds), accurate, and handle billions of users and millions of ads. Deep learning models with embedding tables for categorical features (user ID, ad ID, context features) dominate. Meta's DLRM and Google's Wide & Deep are influential architectures.

**Attribution modeling**: The classic "last-click" attribution gives 100% credit to the last ad a user clicked before purchasing. This ignores the contribution of earlier touchpoints (awareness ads, email newsletters). Data-driven MTA models use path analysis on user journey data to distribute credit more fairly. However, privacy changes (iOS 14.5 tracking restrictions, cookie deprecation) have severely limited the data available for user-level attribution — driving a resurgence of Marketing Mix Modeling.

**Marketing Mix Modeling (MMM)**: MMM uses regression on aggregated marketing spend and sales data to estimate channel ROI. It doesn't require user-level tracking, making it privacy-safe and durable through platform changes. Meta's Robyn and Google's Meridian are open-source MMM frameworks incorporating Bayesian inference and scalability improvements.

**CLV prediction**: Identifying high-value customers early enables efficient marketing investment. Pareto/NBD models (non-parametric statistical models) and BG/NBD (Beta Geometric/Negative Binomial Distribution) predict purchase frequency and customer lifetime from transaction history. ML models incorporating behavioral, demographic, and product features further improve CLV predictions.
</div>

<div style="background-color: #8B0000; color: #FFFFFF; padding: 20px; border-radius: 8px; margin-bottom: 15px;">