Vela Systems cut cloud spend 38% in a quarter

The invoice was too late

Vela Systems had already invested in cloud observability, but the monthly invoice still arrived as a surprise. Finance could see the total, engineering could see individual workloads, and product could see usage growth. No one had a shared view of how those three movements related.

Cost conversations therefore became reactive. A large increase triggered a search through services, environments, and deployment histories, while the original decision that created the spend was already several weeks in the past.

Connecting cost to product behavior

DataWatch began by building a cost model around the questions Vela actually needed to answer: which products were growing, which environments were underused, what portion of spend was structural, and which increases were directly connected to customer value.

The work also clarified ownership. A platform team could own efficiency for shared infrastructure, a product team could own usage-driven growth, and finance could own the baseline used to measure the trade-off. Each group could see the same cost without being asked to use the same workflow.

Finding the avoidable layer

The analysis separated essential scale from accidental scale. Idle environments, oversized development resources, unbounded retention, and duplicated processing were treated differently from the compute required to support real customer demand.

That distinction kept the initiative from becoming a blunt cost-cutting exercise. Vela could reduce waste while protecting workloads that supported reliability, experimentation, and the product experiences customers had already paid for.

A 38% reduction with a new baseline

Over one quarter, Vela reduced cloud spend by 38% and reached a new baseline in eleven weeks. The savings came from a sequence of owned changes rather than one dramatic migration: rightsizing, scheduling, retention decisions, workload consolidation, and a clearer review of new infrastructure.

Every recommendation was paired with an expected effect and a way to check the result. When a saving did not appear as planned, the team could distinguish a missed implementation from an incorrect assumption instead of treating the entire program as inconclusive.

Keeping efficiency close to the work

Vela now reviews cloud movement alongside product and operational metrics. A cost increase can be accepted when it reflects valuable growth, challenged when it reflects drift, and investigated when the evidence does not explain it.

The lasting change is that cloud efficiency is no longer a monthly finance topic. It is part of how product and engineering decide what to build, how to run it, and which forms of complexity are worth carrying.