Sponsored: How Coastline Cloud cut a $60k AWS bill in half

Sponsored content. This case study was produced in partnership with Coastline Cloud and reviewed by the LosAngeles.IT editorial team for factual accuracy. Coastline Cloud is a paying sponsor. Our editorial coverage, reviews, and Awards methodology are unaffected — see our editorial disclosure.

The situation

A Santa Monica-based streaming analytics company (name withheld at their request) was spending approximately $60,000 per month on AWS across production, staging, and internal analytics workloads. Growth had flattened for a quarter, but the cloud bill had continued to climb — a familiar pattern. Leadership engaged Coastline Cloud, an El Segundo-based FinOps consultancy, for a six-week engagement with a single mandate: cut the bill without cutting features or reliability.

Six weeks later, the run-rate was $29,400 — a 51% reduction — with no production incidents and a modest improvement in p95 latency.

Here is how they did it.

Week 1: Instrumentation and tagging

Coastline's first move was not to change anything. They deployed a cost-and-usage report pipeline into a dedicated analytics account, enforced a mandatory tag policy (team, environment, product, cost-center) on all new resources, and back-filled tags on the top 200 cost-driving resources by hand. Within five business days, the client could see — for the first time — what each team and product actually cost.

Finding: 34% of spend belonged to a single internal analytics pipeline that no active product feature depended on.

Week 2: Compute right-sizing

Using two weeks of CloudWatch data plus Compute Optimizer recommendations, Coastline resized 42% of EC2 instances downward by at least one size, and moved 18 workloads from m5 to m7g (Graviton) with no code changes beyond a base-image update. They also converted three always-on staging environments to schedule-based auto-shutdown between 8pm and 7am and on weekends.

Savings: ~$11,000/mo

Week 3: The analytics pipeline

The 34% pipeline turned out to be doing full-table scans on 90 days of raw event data every hour to produce a dashboard that four people looked at weekly. Coastline rebuilt it as an incremental daily job writing to a small aggregated table, cutting compute by 96% and storage IOPS by 88%.

Savings: ~$9,500/mo

Week 4: Storage lifecycle and egress

Object storage was the second-largest line item. The team implemented lifecycle policies moving objects older than 30 days to Infrequent Access and older than 180 days to Glacier Instant Retrieval — a change the application already tolerated because nothing read the old objects on the hot path. They also moved a chatty internal service off cross-region calls, eliminating ~$3,200/mo in egress.

Savings: ~$6,800/mo

Week 5: Savings Plans and Reserved capacity

With the workload right-sized and stable, Coastline modeled a 1-year Compute Savings Plan at 70% of steady-state usage. Committing at that level (rather than 100%) preserved flexibility for continued optimization while capturing most of the discount.

Savings: ~$3,300/mo

Week 6: Guardrails and handoff

The final week was about making the savings stick. Coastline installed:

• A cost-anomaly alert routed to a Slack channel with named owners.
• A weekly cost review dashboard readable by non-engineers.
• A pre-deploy CI check that flags any new resource lacking required tags.
• A quarterly review cadence, with the next check-in booked before handoff.

Results

What made it work

Three things, according to both Coastline and the client. First, executive air cover — the CTO cleared calendars for the analytics-pipeline rewrite in week three. Second, instrumentation before action — no changes shipped in week one, which built trust and prevented cargo-cult cuts. Third, guardrails after — the weekly review has kept the run-rate flat for two quarters since.

Learn more

Coastline Cloud is listed in our directory under AWS Consulting. Editorial coverage of LA cloud providers is separate from sponsored placements and follows our published methodology.