Personalization cases that beat the 'show recommended' baseline

Personalization cases that beat the 'show recommended' baseline

Six rollouts with lift numbers and the honest control group. Curated ruthlessly.

— 1. A retailer's segment-of-one homepage — beat a rules-based version by 8% on revenue-per-session in a clean A/B. Read if you're choosing ML vs rules.
— 2. A media site's content-recirculation model — lifted pages-per-session 14%, which raised ad revenue 9%. The recirculation logic is the takeaway.
— 3. A SaaS's in-app upsell targeting — showed upgrade prompts only to usage-qualified accounts; conversion 3x'd vs blanket prompts, complaints fell. Essential for PLG.
— 4. A travel brand's price-drop trigger — beat batch email by 21% on bookings using real-time inventory signals. The trigger-vs-batch comparison is the lesson.
— 5. A DTC's holdout-group discipline — kept a 5% no-personalization holdout, proved the engine added 6% incremental revenue (not just shifted it). Read this one for the measurement rigor alone.
— 6. A grocery app's substitution model — cut out-of-stock cart abandonment 17%. Skip unless you have inventory volatility.

That's the stack for this week. Forward to a teammate.