Three Ways to Run DetectionOps, Compared Side by Side

Health Technology

Hospitals and clinics run on technology that most visitors never see: scheduling tools, records systems, billing platforms, door controls and the monitoring that watches over all of them. When administrators ask their technology partners how that monitoring is managed, they often hear a blend of jargon and reassurance. This comparison piece, written for non-specialists and strictly about organisational technology rather than anything clinical, sets three common approaches side by side so you can ask sharper questions. The framework is DetectionOps, the practice of managing alarm rules with software-delivery discipline.

Quick background on the framework

One long online guide, the Atlas from the xOps School site, devotes a full entry to the discipline: the DetectionOps entry in the Atlas. Its account runs like this: write the requirement, keep the rule as code, have somebody else read it, test it, ship it by automation, watch how it performs, adjust it, and one day remove it. The Atlas is upfront that the word itself overlaps with older terms and is not a formal standard, so a supplier may describe the same thing differently. What matters is whether the habits exist.

A note on scope. Nothing here is medical, legal or investment advice, and no organisation is being rated. The point is to understand the questions behind the technology.

Approach A: Rules managed by hand in a console

This is the oldest and most common pattern. Someone logs in, creates a rule, adjusts a threshold when it becomes noisy, and moves on. It is quick and requires no extra tooling.

  • Strengths: low setup cost, fast edits, suits a very small estate.
  • Weaknesses: no history of who changed what, no review, no tests, and changes can disappear when an automated update runs. The Atlas names this habit, edits made live and then wiped by the next release, as a classic anti-pattern.
  • Ask: can we see the change history for the last three months?

Approach B: Rules as code in a neutral format

Here every rule is a text file in a shared repository, written in an open, tool-neutral layout such as Sigma, which describes every rule by what it is called, which records it reads and what condition makes it fire. A pipeline converts the files for whatever monitoring platforms the organisation uses.

  • Strengths: portable across tools, easy to review, friendly to automated tests and to community-shared rules.
  • Weaknesses: an extra conversion step and a toolchain to maintain. The Atlas points to a recent episode where a fresh major release of the converter appeared before its sibling tools had caught up, which is why freezing versions is essential.
  • Ask: which versions are locked, and who upgrades them?

Approach C: Native queries kept as code

The third path keeps the same discipline but skips translation. Rules stay in the monitoring platform’s own query language, yet live in version control with the same review and testing steps. The Atlas is explicit that this can be the right answer for an organisation committed to one platform, provided the written requirements and the tests stay independent of the platform so a future move remains possible.

  • Strengths: fewer moving parts, no conversion layer, direct use of platform features.
  • Weaknesses: tied to one vendor’s syntax and roadmap. The Atlas mentions, for instance, that one major cloud platform’s older portal experience is scheduled to lose support after March 2027.
  • Ask: if we changed platform, what would we keep?

A short illustration

Consider a hypothetical regional network of clinics, invented purely for this example. It has twelve sites, one monitoring platform, and a team of two people looking after the rules. Under approach A, the two of them share a console login and keep rules in their heads. When one person goes on leave, the other cannot say why three rules exist. Under approach C, the same team stores each rule in a repository with a note of its purpose. The second person can read the history, reproduce the logic and, if needed, hand the work to a new colleague without a long briefing. The costs are a few hours of setup and a habit of writing notes. The gain is continuity, which administrators value in every other area of operations.

Comparing them on the questions that matter

Traceability

B and C win clearly. Each change carries a name, a rationale and a second reader. In a hand-managed console, you rely on memory.

Speed for emergencies

A is fastest on paper. Yet the Atlas proposes a dedicated urgent lane for B and C, usable during a live campaign: the silent trial can be cut short, while checks and sign-off stay, and a follow-up review lands within the week. That keeps most of the speed and adds a record.

Cost over time

A is cheapest on day one and tends to be dearest later, when nobody can explain a rule or trust an alarm. B and C cost more to start and less to explain. This mirrors what administrators know about record-keeping in general: tidy files cost effort today and save it later. If you are preparing for a technology review meeting, you might find the checklist-style approach in our article on preparing for a heart check-up appointment a useful model for organising questions in advance, even though the subjects differ.

Evidence of effectiveness

This is where the Atlas draws its sharpest line, and it applies to every approach. A rule labelled with a technique from a public attack catalogue is not proof that the technique would be caught. Its worked example contrasts a tag heatmap claiming 100 percent with a rehearsed figure of 60 percent, from 18 of 30 priority implementations. When a vendor quotes “90 percent coverage”, the Atlas calls that a mapping, not an outcome.

Whichever route is chosen, ask for a validated number, its date and its method.

A simple scorecard for the conversation

  1. Are rules stored in one place with owners and purposes?
  2. Does every rule have at least two tests, one that fires and one that should not?
  3. Does a second person review each change?
  4. Do changes go out through an automated pipeline rather than by hand?
  5. Are tool versions pinned?
  6. Is there a measured precision figure for the noisiest rules?
  7. Are exceptions recorded with an owner and an end date?
  8. When was coverage last validated by rehearsal?

Teams that can answer six or more with confidence are doing well. Fewer than four suggests that approach A, or a poorly kept version of B or C, is in play.

Which approach suits which situation?

A small organisation with one monitoring platform and a handful of rules may reasonably begin with C. A group of facilities using several different tools may prefer B for portability. Approach A should be a starting position at most. The Atlas also supplies a maturity ladder with five levels, from improvised to fully managed, and cautions that skipping steps leads to burnout.

If you are curious about the people side, our guide to interview preparation with a structured DevOps practice plan shows the skills such teams look for, and the Atlas reminds readers that a small lab can start with surprisingly little equipment.

Bottom line

None of the three approaches is a villain. What separates them is evidence: records, tests, measurements and dates. Use the scorecard, ask the questions politely, and expect clear answers.