Product1 distinct publisher3 min readUpdated
A devops.com column argues teams bolt security scanning onto CI/CD and rarely go back to measure the throughput they spent. Six cost terms, and only one of them shows up on a vendor quote.
The Product Desk · Product desk
Compiled by The Product DeskSomething wrong?How this is made
A column published by devops.com makes an argument that security tooling pitches tend to skip: many pipeline teams add scanning to CI/CD, and relatively few go back afterward to measure what it cost the delivery process [1]. The consequence is not a rounding error. It is a change in delivery economics that nobody wrote down, made by teams who thought they were accepting a free upgrade.
The column's framing is precise about where the free-upgrade story holds. Shift-left is treated as pure gain: catch problems earlier, at lower cost, with no real downside. That is true for the cost of fixing a vulnerability, and it is not automatically true for the cost of running your pipeline [2].
The mechanics are unglamorous. SAST, SCA, container scanning, secret scanning and dependency analysis all do real work, and each is a pipeline stage with its own execution time. Many of them scale with codebase size or dependency count rather than staying flat, so a scan that is quick on a small service can take meaningfully longer on a large monorepo [3]. Image scanning grows with the number and size of container layers, not just the size of the application code [4]. That means the stage you benchmarked on a sample service is not the stage you will be paying for in eighteen months, because the input keeps growing.
The second-order cost is worse than the stage duration. New findings create triage, false-positive review, and in some cases blocked builds while a team decides whether a flagged issue is real [5]. Blocked builds mean retries, and retries mean re-running earlier stages that had already passed, which is easy to overlook if you are only watching how long the new scanning stage takes rather than its effect on the whole run [6]. Running stages in parallel can offset some of this where the pipeline architecture allows, but parallelization has its own compute cost and does not remove the triage and retry work downstream [7].
On magnitude, the author is careful, and it is worth repeating the caveat rather than the headline. The author writes that across pipelines they have worked on, adding security tooling increased overall pipeline workload, delayed artifact and binary production, and produced a noticeable increase in CI/CD infrastructure cost [8]. They also state plainly that this is a practitioner observation, not a benchmark, not a percentage, and not a claim that every team sees the same magnitude [9]. Treat it as a hypothesis to test on your own pipeline, not a number to quote.
The cheapest recovery the column offers is a question about replacement. A new SCA tool does not automatically make an existing manual dependency review obsolete; the same applies to container scanning layered over a manual image-review checklist, or secret scanning added alongside a code review that was already supposed to catch hardcoded credentials [10]. Overlap without a corresponding removal is duplicated engineering effort dressed up as added security, accumulating one well-intentioned addition at a time, because it is rarely anyone's job to retire the process a tool made redundant [11].
The proposed model is: tool or service cost, plus additional CI/CD compute, plus additional execution time, plus retry and rebuild cost, plus engineering triage effort, plus delivery and release impact. Most cost conversations stop at the first term [12]. Six terms, and only one of them arrives as a quote from a vendor [14].
What to watch is whether your own reporting can answer the question at all. For high-frequency pipelines, the column argues the useful comparison is not scan duration but the cumulative effect across builds: execution minutes, compute consumption, reruns and time-to-artifact over a representative period [13]. If you cannot pull those four numbers for the quarter before and the quarter after your last security control landed, you did not decide the trade-off. You inherited it.
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
SAST, SCA, container scanning, secret scanning and dependency analysis are each a pipeline stage with its own execution time, and many scale with codebase size or dependency count rather than staying flat; a scan that runs quickly on a small service can take meaningfully longer on a large monorepo.
Image scanning in particular can grow with the number and size of layers in a container, not just the size of the application code itself.
Findings from new scanners generate new work: triage, false-positive review, and in some cases blocked builds while a team decides whether a flagged issue is real.
Blocked builds mean retries, and retries mean re-running earlier pipeline stages that had already passed, which is easy to overlook if you are only looking at how long the new scanning stage itself takes rather than its effect on the whole pipeline run.
Running new stages in parallel where the pipeline architecture allows it can offset some of the added cost, but parallelization has its own compute cost and does not eliminate the added triage and retry work downstream.
The author says the observation is a practitioner observation from real pipelines, not a benchmark, not a percentage, and not a claim that every team will see the same magnitude of effect.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
One opinion column, mechanism reasoning only
The cluster is a single practitioner column from one publisher. Its mechanical claims (scanners as stages that scale with code, dependency count or image layers; blocked builds causing retries of passing stages; parallelization trading wall-clock for compute) are internally coherent and verifiable in principle, but the central empirical claim of increased workload, delayed artifacts and higher CI/CD spend rests entirely on the author's unquantified recollection, with no benchmark, telemetry, pipeline identifier, tool name or third-party corroboration anywhere in the supplied material.
No adoption signal in supplied sources
The column reports no release, deployment, benchmark run, pricing or licence change, incident, or disclosed usage figure. It does not name a tool, vendor, team, or organisation, and the practitioner observation is described without pipeline counts or scale. There is nothing in the supplied material to measure adoption of either the described practice or any product.
Rhetoric slightly below its own evidence
Unusually for this genre, the column caps its own claims: it says explicitly that the finding is not a benchmark, not a percentage and not generalizable in magnitude, declines to argue for fewer security controls, and states the terms of its cost model do not need numbers attached to belong in the model. The claim inventory therefore sits at or a little below what the reasoning would license, so there is a small negative gap rather than overstatement. It is not further negative because the unquantified cost-increase assertion still travels under a strong headline framing about a cost nobody puts on the invoice.
Trade opinion column, no product on sale
The piece names no vendor, tool brand, platform or service and makes no purchase recommendation; it argues for a measurement practice and closes by stating this is not an argument for fewer controls or that security tooling is a poor investment. The visible incentive is the ordinary one of a trade-publication opinion column built on the author's practitioner credibility, which rewards a memorable framing and rests the key claim on unverifiable personal experience. No commercial or funding interest is disclosed or inferable from the supplied material.
Clear single-source read, thin verification
What the column says is unambiguous and the ledger maps cleanly onto it, so confidence in the reading is high; confidence in the world-state is limited by one publisher, zero adoption signal, and an empirical core the author deliberately leaves unquantified. The mechanism claims are safe enough to act on as a measurement checklist; the magnitude claim is not.
product
OpenTelemetry is free; the collector fleet, the retention policy and the on-call rota are not1 distinct publisher
product
A green rerun is not a repair: self-healing tests need a merge gate outside the healer1 distinct publisher
product
Cloudsmith's cooldown policies make delay a control, and that makes it your decision1 distinct publisher
product
The cheapest model scored 10 out of 100: assistant choice is now a code-security decision1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 17, 2026