Cricut
A leading consumer crafting technology company known for its smart cutting machines and cloud-native Design Space platform powering millions of creators worldwide through a subscription-driven digital experience backed by large-scale AWS data infrastructure spanning multiple accounts and analytics workloads.
As Cricut's consumer data platform grew across multiple AWS accounts and regions, the organization needed a structured FinOps approach to control cloud costs without disrupting the analytics pipelines powering product, marketing, and machine learning decisions.
Multi-Account Cost Sprawl at Scale
Cricut's centralized consumer data infrastructure had grown across multiple AWS accounts and regions. Redshift Serverless workloads were significantly overprovisioned, cross-account data transfers were generating avoidable NAT Gateway costs, and storage retention policies had never been tuned all quietly inflating the monthly AWS bill with no clear framework to act safely.
Without workload-level profiling and cross-account visibility, every optimization attempt risked breaking something the pipelines powering pricing, personalization, and growth simply can't go down.
Four root causes were quietly inflating Cricut's AWS bill every month:
Four-Phase Optimization Approach
Quper designed and executed a structured four-phase assessment and optimization of Cricut's AWS data infrastructure — moving from fast, low-risk quick wins all the way through deep architectural redesign, without ever touching production pipelines.
Migrated an out-of-network AWS account into a shared AWS VPN — eliminating cross-network data egress and NAT Gateway charges immediately.
- Removed manual snapshots not required for backups
- Tuned snapshot history limits on Redshift Serverless clusters
- Optimized CloudWatch retention periods to retain only relevant logs
Workload profiling revealed a significant mismatch between Redshift Serverless capacity and actual usage patterns — confirming major overprovisioning.
Simulated historical workloads across various Redshift node types and cluster sizes to identify the most cost-efficient and performance-optimized configurations.
Migrated suitable workloads to provisioned Redshift clusters with rightsized node types and counts — resulting in better resource utilization, reduced compute costs, and significantly improved query performance.
- Refined column encoding and data types for storage efficiency
- Implemented regular
ANALYZEandVACUUMoperations - Revised sort and distribution key strategies to enhance query performance
Rearchitected workload distribution to minimize reliance on Redshift concurrency scaling while maintaining SLA commitments — optimizing WLM queues, resource allocation, and scheduling policies.
Comprehensive audit identified and removed unused datasets and stale workloads — reducing storage costs and improving query performance across the platform.
Transitioned from full-refresh pipelines to incremental ETL jobs — significantly improving data pipeline efficiency and reducing compute usage per run.
Redesigned the warehouse with clear modular data layers — raw ingestion through to aggregated outputs — centralizing transformation logic, enabling decentralized team ownership, and simplifying stakeholder-facing models.
40-50% Cost Reduction, Zero Pipeline Disruption
With Quper's assessment, Cricut achieved a 40-50% reduction in AWS data platform costs across four structured phases. Phase 1 quick wins delivered immediate ROI through cross-account VPN migration and retention cleanup. Phases 2-4 unlocked deeper savings through Redshift rightsizing, performance tuning, and architecture redesign — resulting in 2× analytics performance with zero disruption to production pipelines.
Concrete Outcomes
By using Quper, Cricut moved from AWS cost sprawl to structured FinOps action — every recommendation backed by historical workload simulation and cross-account analysis, not assumptions.
A Phased Optimization Roadmap
Quper gave Cricut a clear, sequenced path — deliver quick wins first, rightsize the expensive clusters second, then optimize the architecture for long-term efficiency.
Cross-account VPN migration eliminated NAT Gateway egress charges immediately. Snapshot retention tuned across Redshift Serverless clusters. CloudWatch log retention periods optimized — all completed with no production impact and immediate cost savings.
Overprovisioned Serverless workloads profiled against historical usage patterns and migrated to rightsized provisioned Redshift clusters. Concurrency scaling dependency reduced through WLM optimization — achieving 2× performance at lower compute cost.
Full-refresh ETL pipelines transitioning to incremental jobs. Data warehouse redesign in progress — clear modular layers from raw ingestion to aggregated output, eliminating logic duplication and enabling decentralized team ownership for long-term scalability.
