The sales demo is always impressive. Buttons get clicked, a radar-like dashboard fills with findings, and someone says the word “visibility” at least four times. Two weeks later the same dashboard is a wallpaper of orange warnings nobody opens, and the org concludes that security is “a work in progress”. If you have lived this, you already understand the question this piece argues: when an aviation academy, a flight school, or any small software-operating team wants security, is the move buying a scanner or building a DevSecOps practice? Short answer: the second, and the tools come along as passengers. Long answer: four rounds below, scored honestly. When you want the receipts behind every claim, they live at the DevSecOps chapter of the Ops Atlas.
What DevSecOps means, so the comparison stays fair
It is security installed inside the delivery pipeline and owned by the team that ships, so each change is examined and leaves evidence behind. Not a product, not a certificate — and this matters in a comparison — no official standard defines the term and nothing certifies it. The scope is concrete, though: threat modeling changes, secret detection, code and dependency analysis, infrastructure and image checks, dynamic and API tests, policy as code, bills of materials, signing and provenance, pipeline hardening, champions, and gates. Runtime monitoring and incident response belong to different rooms down the hall. A scanner covers roughly two items on that list; a practice covers all of them, using scanners where they help.
Round one: the first ninety days
The scanner’s best round. Unbox it, connect the repositories, and by Friday you have findings — hundreds. It feels like progress and photographs beautifully in a steering meeting. The practice path looks slower: write the gate policy first, store results in SARIF, model who may act on the build system, then run everything in report-only mode while the numbers settle. SBOM and provenance switch on from the very first build, because evidence is cheapest to collect early. Ninety days in, the scanner team owns a backlog; the practice team owns a triaged list with owners attached, plus paperwork they can hand to an auditor or a buyer. Speed favors the tool; staying power favors the system. And teams forget the demo long before they forget a skipped calibration.
Round two: the noise problem
Here the tool usually loses the crowd. Every analyzer ships with rules that cry wolf, and an uncalibrated scanner teaches a team to ignore it within a month. The corrective comes from Google’s Tricorder experience: a rule gains the right to refuse a change only after its effective false alarms measure essentially zero, its remarks stay under a tenth of review conversation, it inspects just the lines a change touched, and it survives an audition on thirty sampled findings. The monthly override share carries its own bar — under five percent. In the field’s reference case, a gate that was overridden on 22 of 96 blocks across 1,240 changes — 22.9 percent — was pronounced untrustworthy, which is precisely what happens when someone buys round-one visibility and skips round-two calibration. A practice includes this calibration as a standing job; a scanner purchase assumes it will happen by itself.
Round three: who answers the alert
Findings are only as good as the human who owns them. The tooling story tends to end at “notify the team”; the practice story names that person in advance. The champions pattern gives each squad one member with a protected slice of the week — ten to twenty percent — for carrying security questions, and DORA’s 2022 research explains why this matters beyond tidiness: teams at the weak end of security practice carried a 1.4 times greater chance of burning out, and culture turned out to be the stronger adoption predictor, ahead of tooling. A scanner purchased without an owner is a smoke detector with the battery out — it still beeps, just never at the right time. Round three goes to whoever planned for people, and it is not close.
Round four: the bill
Licenses are the visible cost; the invisible ones decide the comparison. On the tooling-only path they include the bypass culture that over-strict gates create, the exception sprawl of waivers with no expiry, and the incident where nobody can say which builds used a compromised action — the tj-actions episode of March 2025 and the force-pushed Trivy tags of March 2026 both punished teams that lacked inventories and verification. On the practice path the overhead is measured, as one of the field’s proposed metrics — pipeline overhead — right next to escape rate and provenance coverage. Nothing certifies DevSecOps, so neither path buys a badge; one of them buys evidence. Score: tooling, a subscription; practice, an asset.
The DevSecOps verdict, and how to have both
The honest scorecard says the comparison is lopsided on purpose: a DevSecOps practice is the thing you are building, and scanners are the appliances it buys when it needs them. Sequence it the way the field’s own lifecycle suggests:
- day zero: gate policy, SARIF storage, and an identity model in writing;
- day one: report-only pilot, false-positive measurement, SBOM and provenance live;
- day two onward: monthly metric reviews, noisy rules retired, the compromise runbook rehearsed.
Then, and only then, shop — with Gitleaks watching secrets, CodeQL or Semgrep reading code — noting that Semgrep’s –config auto phones home with your project URL — Trivy covering images and dependencies, ZAP driving dynamic tests, and the Sigstore trio handling signatures. Buy to serve the practice and every round above flips in your favor. Buy instead of the practice and you have simply paid list price for the orange wallpaper.
