IncidentOps FAQ for Aviation Safety Management Support Teams

Safety management systems in aviation give organisations a structured way to identify hazards, report events and learn from them. Support teams that run the digital tools behind those systems, such as reporting portals, audit trackers and training databases, face a parallel question: what should happen in the first hour after one of those tools stops working? Technology teams use a body of practice informally called IncidentOps to answer it. The questions below are the ones a safety or quality manager is likely to raise, and the answers rest on the Ops Atlas, a learning-oriented reference to operations disciplines. It is operational background for information systems, not guidance on aviation safety regulation, and your approved manuals and authority requirements always take precedence.

What is IncidentOps, in one paragraph?

According to the Atlas, IncidentOps is an informal, uncommon name for incident management: the coordinated response to unplanned disruption of a service, covering declaration, command, communication and learning. The entry is candid that it is not an established discipline with its own standards body, and that very few sources use the term. The practices underneath, though, are long established and borrow from emergency-management structures. Readers can consult the Ops Atlas entry covering IncidentOps for the full treatment.

What counts as an incident, and what does not?

An incident is an unplanned event that interrupts or threatens a service. Inside the scope, the Atlas lists the rules for declaring and grading incidents, the on-call and escalation set-up, roles, the cadence of messages, public status pages, mitigation steps, post-incident reviews, follow-up work and the measurement of incidents. It places other things outside: attacker-driven events belong to security response; reliability targets belong to reliability engineering; alert design belongs to observability practice; invoking a disaster-recovery plan belongs to recovery planning; and removing underlying causes over weeks belongs to problem management. A maintenance window agreed in advance is not an incident, and a noisy alert does not automatically become one.

When should an incident be declared?

As early as written criteria allow. The Atlas cites criteria used by a large technology company: another team’s help is needed, the effect is visible to customers, or the problem remains unresolved after roughly an hour. Declaring an incident that proves small consumes a few minutes. Declaring late consumes hours. Organisations that treat the number of declared incidents as a performance score tend to see declarations fall, and this is the opposite of what a safety culture wants.

Who leads the response?

A designated coordinator, often called the incident commander. The role is a hat worn for the duration, not a job title. The coordinator decides and organises but does not perform the repair; a separate hands-on group alone is allowed to alter live systems. A communications lead issues updates, a scribe records events, and in larger responses there are a deputy and liaisons to other teams. Emergency-service guidance points to about five reports per supervisor, which means larger efforts should be broken into sub-teams.

How often should stakeholders be updated?

The Atlas describes a rhythm used in technology operations: an opening message roughly five minutes after the response call begins, a scoping message soon after, and thereafter further notes spaced no more than twenty minutes apart during the opening two hours. A closing message should state whether data was lost. Hosted status tools usually offer four stages for the public view, named Investigating, Identified, Monitoring and Resolved. Short, regular messages build more confidence than rare, detailed ones; the guidance in the Atlas puts it as favouring clarity over brevity.

Should responders find the cause before acting?

No. The Atlas draws a distinction between mitigating and explaining. All-purpose mitigations such as rolling back a change, draining traffic, disabling a feature, switching to a standby or shedding load require only an idea of where the fault is. Root-cause analysis follows once service is restored. One illustration in the Atlas is a failure to create clusters at a major cloud provider that dragged on for more than six hours; once the faulty component had been pinpointed, reverting to earlier images would have brought service back well before the eventual fix.

How should the paging path and status page be designed?

Both should sit outside the blast radius of the systems they describe. The Atlas offers a well-known example: in February 2017 a large storage service in one region suffered a prolonged disruption after a mistyped input to a playbook removed more capacity than intended, and the provider’s service health dashboard was itself unavailable for a period because it depended on the failing service. The provider then communicated through other channels and later spread the console across regions. For a safety-reporting portal, the practical rule is that the page announcing an outage should be hosted independently of the portal.

How should security-related incidents be handled?

Here the Atlas offers a proposal rather than a settled standard. Under it, the coordinator continues to steer the overall response, and the security workstream shifts to a closed channel run by the lead for security response. Evidence-preservation requirements may constrain how systems are restored, and legal advisers control what is said publicly. The Atlas notes a recent US federal publication on security incident response, dated April 2025, as the relevant guidance on the security side. Organisations subject to breach-notification regimes should seek qualified advice; one example the Atlas gives is the EU NIS2 regime, which sets staged reporting deadlines running from a day to three days to a month.

What should be measured?

The Atlas advises care. The abbreviation MTTR has at least four meanings, so any figure must state its phase and its population. Because incident durations are skewed, a handful of long events can move an average sharply, and the Atlas reports a simulation in which the mean shifted by more than twenty percent in most simulated quarters even though nothing had really changed. It suggests reporting distributions of detection, acknowledgement, mitigation and restoration times; the impact on reliability objectives in error-budget minutes; the load on people from paging; and the proportion of follow-up actions completed on time.

How should communication fail-safes be tested?

A heartbeat check that proves the alert path still works is a sensible early step. The Atlas lists a watchdog heartbeat and a first paging drill in its sixty-day plan. A drill should send a real test page to the real rota and measure how long acknowledgement takes. Where an event-ingestion interface is used, teams should confirm what success looks like; the Atlas notes that one popular API returns an HTTP 202 on acceptance, and a 429 response signals the sender to retry with backoff.

What does a good post-incident review look like?

Its owner is chosen before the response is over, and a facilitator who was not personally involved leads it. It stops at system conditions and does not end with an individual’s name. The output is a handful of actions, every one carrying a single owner and a date. Published examples differ: one vendor’s process asks for a written review within three calendar days at its top severity and five business days at the next, and a large software company allots either a month or two to priority actions, according to how critical the service is. Among the anti-patterns the Atlas lists are reviews that only tally tickets and teams that dispose of an action by lowering its priority.

How can an organisation tell how mature its IncidentOps practice is?

The Atlas proposes five stages. Stage one is improvised effort. Stage two brings in a severity policy plus a rota. Stage three requires that the communication rhythm be kept in a large majority of serious incidents, with ninety percent as the proposed marker. Stage four expects 80 percent of priority actions to land on schedule across two quarters in a row. Stage five is adaptive practice. These thresholds are the Atlas’s own proposal, so organisations should treat them as a starting reference and not as an external benchmark.

What is a sensible first step?

Write one page: a definition of incident, a severity scale, four role cards, and a message template. Assign a rota that is fair, rehearse once with a table-top scenario, and arrange for a hosted status page that sits outside your main systems. Review the arrangement after the first real event, and keep the action list short. Everything beyond that can be added as the team grows.

Leave a Comment

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

Scroll to Top