Build1 publisher3 min readPublished
A three-week Elastic baseline stood in for the SIEM proposal at a 70-developer product shop
An anonymized consulting write-up says a product team of 70 to 100 developers got failed-auth, RBAC and container-exec visibility out of the Elastic cluster it already ran for troubleshooting. The post describes the sequence and generalizes the technical details.
The Engineer · Build desk

What happened
- A software product company with roughly 70 to 100 developers ran AWS, Kubernetes, application and platform logs, and an existing Elastic deployment used mainly for troubleshooting and operations.
- It had no consistent view of failed authentication, suspicious access patterns, Kubernetes warning activity, RBAC changes, unusual container execution or authentication spikes across services.
- The consultant instead spent roughly three weeks picking high-value sources, building a small prioritized detection set and dashboards, and writing a response workflow the DevOps team could own.
Compiled by The EngineerSomething wrong?How this is made
Why it matters
- cost The saving claimed is a first-phase licensing and implementation commitment the company had not made. The new bill is DevOps triage time, and the account puts no price on it.
- decision A team with logs and no security budget now has a documented first step to put against a vendor proposal. The choice it forces is who owns triage and escalation.
- constraint Anyone copying the approach writes the detection content themselves and discovers their own Elastic sizing under load, because the account generalizes the technical detail.
Failed authentication counts, RBAC changes and an unexpected shell in a container do not come out of the same emitters as the application logs a team keeps in Elastic for troubleshooting [4][5]. The write-up says the work drew on the telemetry the company already had "or could enable with minimal friction" [16]. The post leaves that second category unnamed, and the published text lists categories of work instead of rules, sources or ingest volumes [20].
The distinction between operational and security telemetry is the part of the account that generalizes. Health metrics such as CPU, memory, latency, I/O, cache and restart counts tell you a system is up [15]. "They do not automatically tell you whether someone is repeatedly failing authentication, changing privileges, opening an unexpected shell in a container, or creating a risky access pattern," the author wrote [14]. The post frames the gap as an operational question: if something suspicious happens tomorrow, who sees it first, and what happens next [18].
The account is mostly about sequence. The stated model was four steps: reuse what exists, collect what matters, prioritize the signals, make the team operationally ready [10]. Four of the five first-phase deliverables are artifacts you can build in a couple of weeks: a baseline view of high-value signals, a small alert set, a shared dashboard and lightweight response guidance. The fifth is clear ownership for triage and escalation [9].
For the three weeks to transfer, several things have to hold. The existing Elastic deployment has to absorb the new sources without a resize. That depends on audit-event volume. The DevOps team has to accept alert triage as standing work, because the response workflow was built for them to own [1]. And the comparison runs against a proposal. The earlier recommendations were to add a commercial SIEM, introduce more security products, pay for professional services and potentially add headcount [7]. The post reports the avoided cost as a large first-phase licensing and implementation commitment, and gives no figure [11].
The author's framing is direct: "security maturity does not always begin with more tooling. Sometimes the highest-value first step is to make better use of the systems you already operate" [12]. The disclaimer says the engagement was real and completed, with the client's name and identifying details withheld under NDA and some details generalized to preserve confidentiality [13].
The post also argues that the client had plenty of data and lacked a security-oriented operating model around it [17]. That claim is the one most likely to survive contact with a different environment. I would run this order of operations in a shop that already has an ELK cluster and no security budget. I would also expect to write the detection content myself, because the account supplies the sequence and stops there [20].
What to watch
- A published detection set or ingest configuration from this pattern would make the three weeks reproducible by someone else.
- Whether the same client later buys the commercial SIEM, and which parts of the baseline survive that migration.
- Elastic sizing and retention cost once audit-grade security sources land in a cluster provisioned for troubleshooting.