Accelerating Real-Time Analytics with Amazon QuickSight and SPICE

Challenge

TargetBay’s BayEngage platform powers campaign delivery for over 4,500 e-commerce merchants. The platform captures billions of campaign events — opens, clicks, bounces, and conversions — within Amazon Aurora MySQL clusters. Internal teams rely on real-time insights to monitor campaign performance, detect failures, and drive product decisions. Three critical issues were undermining this capability:

• Slow, Costly Ad-Hoc Reporting: Direct SQL queries against large Aurora tables caused multi-minute response times, production database CPU spikes, and risk to transactional workloads. Complex analytical queries competed for resources with real-time campaign processing.
• Fragmented Data Architecture: Customer and event data was distributed across multiple Aurora clusters, preventing unified reporting. Teams could not easily join tenant, campaign, and engagement data across cluster boundaries for cross-cutting analysis.
• Delayed Error Visibility: Campaign or batch job failures were typically discovered only during next-day reporting cycles, meaning issues that could have been caught in minutes went undetected for hours, impacting merchant campaigns and eroding platform reliability.

Solution

FactualMinds deployed Amazon QuickSight with the SPICE in-memory engine to decouple analytics from production databases, providing near real-time reporting without impacting transactional workloads.

Data Ingestion and SPICE Configuration:

• Configured QuickSight to ingest data from each Aurora MySQL cluster independently
• Scheduled incremental SPICE refreshes every 5 to 15 minutes, ensuring dashboards reflect current state without continuous database queries
• Stored key tables — tenants, campaigns, batches, and events — in SPICE for rapid in-memory access

Unified Cross-Cluster Reporting:

• Joined tenant, batch, and campaign tables from multiple SPICE data sources within QuickSight
• Created unified dashboards spanning tenants and regions, providing a single-pane view that was previously impossible with direct database queries

Live Dashboards and Monitoring:

• Built dashboards covering campaign engagement metrics (CTR, open rate, bounce rate), batch processing success and failure rates, and top-performing tenants
• Configured threshold-based alerts and anomaly detection directly within QuickSight, enabling immediate notification when metrics deviate from expected patterns

Self-Service Analytics:

• Empowered business teams to build ad-hoc visualizations — funnels, heat maps, time series, and pivot tables — without requiring engineering involvement or database access

AWS Services Used

• Amazon QuickSight with SPICE — In-memory analytics engine with scheduled incremental refresh
• Amazon Aurora MySQL — Source database clusters for campaign and tenant data
• Amazon CloudWatch — Infrastructure monitoring and alerting
• AWS IAM — Access control for QuickSight dashboards and data sources

Results

The QuickSight deployment transformed TargetBay’s analytics capabilities from reactive and fragmented to near real-time and self-service:

• Near real-time analytics with 5 to 15 minute refresh cycles replacing multi-minute ad-hoc queries against production databases
• Eliminated reporting lag by surfacing campaign failures, anomalies, and performance degradation within minutes rather than the following day
• Reduced Aurora database overhead by offloading analytical workloads to SPICE, freeing production clusters to focus on transactional processing
• Self-service analytics for business users enabling product, marketing, and operations teams to explore data independently without engineering support
• Instant anomaly detection and alerts through threshold-based monitoring that proactively flags issues before they escalate to merchant-facing impact