Case Study: A Safety Team Learns to See Its Own Risk

Safety management system metrics are easy to preach and hard to run, so instead of preaching, here is a story. The team below is fictional — a composite of patterns we see across training operations — but every problem in it is real, and so is every fix. Call the school Meridian Flight Academy: four trainers, nine instructors, a part-145 shop next door, and a safety officer named Dana who inherited an SMS binder nobody read.

Month one: a binder full of hazards, a wall with no gauges

Dana’s first audit of her own program was blunt. Hazard reports trickled in — maybe three a month, always from the same two instructors. The risk register was eighteen months stale. Occurrence reviews happened “when needed,” which meant after something scared somebody. Meanwhile the chief flight instructor swore the operation was safe, and his evidence was that nothing terrible had happened lately. Dana wrote one line in her notebook that drove everything after: we have opinions where other industries have instruments.

Her fix didn’t start in aviation. Her brother ran platform engineering at a logistics firm, and over dinner he showed her the screen his executives watched: a single view that fused how fast teams shipped changes, how often things broke, and how much margin remained on each service promise. The product behind it, DevOpsIQ from BestDevOps, had sold his leadership on one idea — that delivery pace, breakdown history and promise-risk belong in the same frame, assembled from read-only records the systems already produced. “We stopped arguing about whether we were healthy,” he told her. “The screen ended the argument.” Dana decided Meridian’s safety program needed the same thing: fewer adjectives, more gauges.

Month two: choosing instruments instead of collecting everything

Her first draft listed twenty-three possible metrics. Her second draft listed five, because the no-nonsense rule she adopted was: track only what the safety committee will act on. The survivors:

  • Report flow — hazard and occurrence reports per hundred flight hours, as a health sign of the reporting culture itself, not a scold.
  • Closure time — days from report to closed corrective action, with the slowest open item named at every meeting.
  • Repeat findings — the share of new reports that matched a hazard already in the register. High repeats meant fixes weren’t sticking.
  • Dispatch integrity — scheduled sorties that launched as planned, because schedule chaos is a stress multiplier that shows up later as safety events.
  • Promise margin — her favourite borrowing. The committee set explicit tolerances (no open high-risk item older than 30 days, no month with fewer than four reports) and tracked how close each tolerance was to being breached. Engineering teams would call the remaining headroom an error budget; Dana called it “runway left,” and the name stuck.

Month three: the board goes up, the arguments change

One whiteboard in the crew room, five rows, updated every Friday, reviewed every Monday for fifteen minutes. The early numbers embarrassed everyone — closure time averaged 47 days, and repeat findings ran high — but Dana enforced a single rule copied straight from the mature reporting cultures: the board judges the system, never a person. No name ever appeared on it, and any instructor could challenge a number by producing a record — challenges welcomed, feelings not admissible.

The arguments changed within three weeks. Instead of “are we safe?” (unanswerable, endless), the committee argued about why closure time wasn’t moving (answerable: corrective actions had no owners) and whether the dispatch dips on Thursdays were a maintenance pattern or a rostering one (answerable: rostering). The chief instructor, initially the loudest skeptic, converted the day the board predicted a problem before he felt it — report flow from one trainer’s students dropped to zero two weeks before a formal complaint revealed its instructor had been waving off “minor” squawks. The silence had been visible on the board; nobody had known how to read silence before.

Month five: runway left, spent on purpose

The deepest change was how Meridian used its tolerances. When the “no high-risk item older than 30 days” margin got thin, the committee didn’t write a memo — it cancelled discretionary projects that week and swarmed the backlog. When reporting volume stayed healthy and repeats fell, they allowed themselves changes they’d postponed, like onboarding a new practice area. Spending safety margin consciously, instead of discovering breaches after the fact, is the whole trick — and it’s precisely the operating rhythm reliability engineers run when a service’s failure allowance thins. Dana’s brother recognised it instantly on a visit, laughed, and pointed at the board: “That’s my dashboard with propellers.”

Did it cost much? One whiteboard, ninety minutes of Dana’s Friday, and the committee’s standing Monday quarter-hour. The expensive part was cultural: surviving the embarrassing first month without anyone killing the messenger.

There was one near-death moment worth recording. In month four an insurance renewal questionnaire asked for “evidence of proactive safety monitoring,” and the owner wanted to mail photos of the board. Dana vetoed sending the ugly rows — then changed her mind and sent all of them, trend lines included. The underwriter’s response reshaped the owner’s view of the whole project: operations that disclose measured weaknesses with visible improvement read as lower risk than operations that claim perfection. The board had become an asset outside the crew room. After that, nobody asked Dana what the Friday updates were for.

Safety management system metrics Meridian would recommend

  1. Five gauges beat twenty-three. Every metric must have a committee action attached, or it’s decoration.
  2. Measure the reporting culture, not just the reports. Volume dropping is a warning light, not a victory.
  3. Name the slowest item out loud, weekly. Aging corrective actions die of attention, not of process.
  4. Set tolerances and watch the margin. A limit you only notice after breaching it is a tombstone metric.
  5. Keep the board blame-free or lose the data. The first name on the board is the last honest report you’ll get.
  6. Steal shamelessly from other industries. The one-view discipline Meridian copied was built for CTOs and SRE leads watching software estates; the knowledge base BestDevOps maintains around delivery metrics, incident handling and service-level thinking reads surprisingly well with “aircraft” substituted for “service.” Good instruments don’t care what vehicle they’re bolted to.

The ending that isn’t an ending

A year in, Meridian’s board shows closure time in single digits and repeat findings at a third of their starting share — numbers we’ve kept deliberately round because Meridian is a composite, not a customer testimonial for anyone. The honest conclusion isn’t the numbers anyway. It’s that Dana’s safety meetings got shorter and calmer while catching more, because the fifteen minutes started from shared evidence instead of competing memories. Safety management system metrics, done plainly, don’t add bureaucracy; they delete argument. Put up one board, pick five gauges, guard the margins, and let the instruments end the debates your binder never could.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top