Goodiebag is a food sustainability platform that connects consumers with surplus food from restaurants, cafés, and bakeries. The platform enables merchants to sell excess prepared food while offering discounted options to users. As the platform scaled, its notification system struggled to balance customer engagement and merchant outcomes.
Problem: lack of personalized push notifications leads to user fatigue and poor targeting
Goodiebag’s existing notification system relied on fixed logic and batch-based delivery, which resulted in customers receiving up to 30 notifications per day. This led to notification fatigue, increased complaints, and lower engagement. The system lacked personalized push notifications and did not account for individual behavior, preferences, or engagement patterns.
At the same time, merchants faced inconsistent sell-through rates. High-demand inventory sold out quickly, while low-demand items remained unsold due to the absence of targeted personalized push notifications. As a result, the platform experienced high notification volume, poor inventory utilization, and limited scalability.
Solution: AI-powered personalized push notifications system
GoML designed and implemented a machine learning-driven system using its Agentic AI accelerator to enable personalized push notifications across the platform. The system optimizes customer selection, timing, and frequency using real-time data and predictive models.
Customer prioritization for personalized push notifications
The system ranks customers based on likelihood to purchase using:
- Purchase history and order frequency
- Location proximity to merchants
- Notification engagement patterns
- App usage behavior
Dynamic scheduling of personalized push notifications
The system replaces fixed batch notifications with adaptive delivery:
- Runs automated pipelines at frequent intervals
- Determines optimal timing for personalized push notifications
- Calculates wave-based delivery based on inventory lifecycle
- Stops personalized push notifications when inventory sells out
- Ensures delivery within valid time windows and time zones
Spam prevention in personalized push notifications
The system reduces fatigue through strict controls:
- ML-based daily limits for personalized push notifications
- Rule-based cooldown periods based on engagement
- Exclusion of users exceeding thresholds
- Dynamic control of notification frequency per user
Inventory optimization using personalized push notifications
The platform improves sell-through rates by:
- Predicting demand for different inventory types
- Prioritizing new customers for merchant acquisition
- Distributing users across inventory using ML ranking
- Adjusting personalized push notifications as inventory approaches expiry
- Ensuring fair allocation across partners and listings
Real-time system for personalized push notifications
The system integrates with Goodiebag’s infrastructure:
- RDS database for customer and order data
- S3 for ML model storage
- FastAPI services for orchestration
- Cron-based execution every few minutes
- Parallel processing across partners
Impact
- 60% reduction in notification volume through controlled personalized push notifications
- 25% increase in notification click-through rates
- 30% improvement in sell-through rates for merchant inventory
- 20% better demand distribution across inventory
About
Before Gen AI and after Gen AI
“Goodiebag improved engagement and sell-through by implementing personalized push notifications powered by machine learning.”
Prashanna Rao, Head of Engineering, GoML
Key takeaways for marketplace platforms
Common challenges
- High notification volume reduces engagement
- Lack of personalized push notifications leads to poor targeting
- Static systems fail to scale
Practical guidance
- Implement personalized push notifications using ML
- Use customer scoring to improve targeting
- Control notification frequency per user
- Optimize timing using behavioral data
- Build real-time systems for personalized push notifications
Ready to implement personalized push notifications for your platform
Partner with GoML to build scalable AI systems that improve engagement, reduce fatigue, and optimize conversions with AI Matic.
