How NovaBrew Coffee Tripled Email Revenue Using AI Segmentation
NovaBrew Coffee used AI-powered email segmentation and predictive send-time optimisation to triple email revenue and recover 67% of abandoned carts.
The Problem: One Email, Everyone
NovaBrew Coffee sold specialty single-origin beans and brewing equipment direct-to-consumer. Their email list had grown to 41,000 subscribers — but their email strategy had not grown with it.
Every campaign went to every subscriber. A first-time visitor who had browsed one product page received the same email as a customer who had bought 12 times. Cart abandonment emails fired 24 hours after abandonment regardless of the subscriber’s history or behaviour.
Email revenue was flat. Open rates had declined to 14%. Cart abandonment sat at 74% with a 6% recovery rate from their single follow-up email.
The founder’s assessment: “We were spam. Relevant to nobody, annoying to everyone.”
The Solution: Behavioural Segmentation + Predictive Personalisation
NovaBrew implemented an AI-driven email platform that replaced their batch-and-blast approach with behaviour-driven automation.
Segmentation layer
The platform analysed purchase history, browse behaviour, email engagement, and product category affinity to classify subscribers into dynamic segments that updated in real time.
Key segments included:
- Espresso enthusiasts (purchased or browsed espresso equipment or beans 3+ times)
- Pour-over practitioners (similar pattern for pour-over products)
- Gift buyers (purchased during November–December, one-off high-value transactions)
- Lapsed customers (no purchase in 90+ days, previously 2+ orders)
- High-CLV actives (top 15% by lifetime value, purchase frequency above monthly)
Predictive send-time
Rather than sending all campaigns at 10am Tuesday, the platform predicted the optimal send time for each subscriber individually based on historical open and click patterns.
Abandoned cart sequence
A three-email sequence replaced the single follow-up: immediate trigger at 1 hour, product-specific follow-up at 24 hours, discount offer at 72 hours calibrated to the subscriber’s purchase history.
Results at 6 Months
- Email revenue: increased 312% against the prior 6-month period
- Cart abandonment recovery: 6% → 67% recovery rate
- Open rates: 14% → 38% average across segmented campaigns
- Unsubscribe rate: fell from 0.8% per campaign to 0.1%
- Revenue per email sent: increased 8.4×
The cart abandonment sequence alone generated more revenue in month three than their entire previous email programme.
The Content Shift
The technology enabled personalisation, but personalisation required content. NovaBrew used a ChatGPT-based workflow to generate first drafts for each segment variant, refined by a part-time copywriter. Time per campaign variant: approximately 90 minutes total, down from a full day for a single broadcast email.
What Made This Work
Clean product taxonomy. The segmentation model needed structured product data to identify category affinity. NovaBrew spent two weeks tagging every product in their catalogue before implementation.
Resisting over-segmentation. NovaBrew started with 6 segments and expanded to 14 over six months as they developed content capacity to support more. Starting complex produces overwhelming complexity.
Treating the discount as a last resort. Their previous cart abandonment email offered 10% off immediately. The new sequence withheld the discount until the 72-hour email — preserving margin on customers who would have purchased anyway.