Case Study: How a Bengal SaaS Cut Cloud Costs 28% with Spot Fleets and Query Optimization

Hook: A 28% sustainable cost reduction in 90 days — not by cutting features, but by smarter orchestration and query design.

This case study walks through a real project with a mid‑stage SaaS company headquartered in Kolkata. We document the assessment, experiments, technical changes, and organizational shifts that produced a persistent 28% reduction in monthly cloud spend while keeping performance and feature velocity intact.

Baseline: where inefficiency hid

The initial audit revealed three main issues:

• Underutilized VMs running non‑critical workloads continuously.
• Expensive analytic scans from dashboards hitting primary OLTP nodes.
• Query inefficiencies in the ORM layer that surfaced under sharded traffic.

Step 1 — short‑term wins (30 days)

1. Introduce spot and preemptible instances for batch workloads and CI runners.
2. Offload heavy dashboard queries to a materialized view cache.
3. Fix N+1 query patterns and driver misconfigurations identified with Mongoose‑style benchmarks (Benchmark: Mongoose 7.x).

Step 2 — structural changes (60 days)

We deployed a hybrid OLAP‑OLTP topology and introduced lifecycle policies to move older analytics data to object storage. We leaned on the patterns in Hybrid OLAP‑OLTP Patterns (2026) and implemented a read gateway to isolate analytic load.

Step 3 — governance and teams (90 days)

We created a cloud cost SLO and a small cost‑ops guild with reps from engineering, product, and finance. The guild enforced budget guardrails and required cost impact notes on PRs for infra changes. For running data‑informed conversations across the company, the Analytics Playbook (2026) provided a helpful template to map technical metrics to business outcomes.

Outcomes and lessons

• 28% reduction in recurring cloud compute spend, sustained over six months.
• No feature rollbacks — developer velocity remained stable.
• Improved visibility into long‑running queries and retention costs.

Operational notes

Spot fleets require automation and careful capacity planning. We used graceful draining, capacity buffers, and automated fallback to on‑demand instances. Authorization playbooks from Authorization Incident Response (2026) were useful while reconfiguring service accounts and read gateways.

What we’d do differently

• Start a cost guild earlier — cultural change matters.
• Invest in read replicas and materialized views before re‑architecting major flows.
• Run driver and sharding benchmarks during spike tests (Mongoose benchmark guidance again helped).

Recommended reading

• Benchmark: Mongoose 7.x on Sharded Clusters
• Hybrid OLAP‑OLTP Patterns (2026)
• Analytics Playbook (2026)
• Tax Technology Roadmap 2026–2028 — for finance integration on cloud cost treatment.
• Sustainable Packaging Options (2026) — not directly cloud‑related, but helpful for SaaS companies adding physical merch efficiently.

Author: Arindam Sen — CTO Advisor. I led the technical audit and cost‑ops guild for the project described.