Summary

  • A Canadian telecom provider may have to notify authorities while cause, impact and restoration estimates are still moving; “if known” preserves honest uncertainty but does not stop the clock.
  • The workable control is a field-level incident ledger linking the reportability threshold, awareness time, recipients, material updates, restoration and the 30-day report.

At minute 93 of a serious network failure, the incident commander may know which services have disappeared and how many customers are exposed, but not which change, supplier or device caused the break. That is not an exceptional state. It is the normal distance between detection and explanation. Under the Canadian Radio-television and Telecommunications Commission’s major-outage framework, it can also be the moment when the reporting clock is already running.

Telecom Decision CRTC 2025-225 makes the distinction explicit. The required initial notification asks for the causes of the outage “if known”. The post-outage report uses the same qualification for root cause. Uncertainty is therefore a valid fact state. It is not permission to wait until uncertainty disappears, because the framework asks for many other facts that can be known earlier: the service class, affected area, start time, reporting threshold, estimated restoration, user impact, affected networks, other providers and obstacles to repair.

The first operating task is not root-cause analysis. It is classifying the event. Primary services include telephone, Internet, cellphone and data services used by individuals and businesses. For this class, an outage means complete loss, not degradation. A major primary outage lasts at least 30 minutes and either reaches 600,000 affected user-minutes or isolates a remote, rural or isolated community. The Commission says that a provider close to the numerical threshold and unsure whether it has crossed it should report.

That calculation turns telemetry into a regulatory object. User-minutes are duration multiplied by affected end-users. Neither input is necessarily stable during an incident. An outage may be detected by alarms before the customer denominator is reconciled; restoration may return in segments; two faults with different causes should not automatically be pooled. A dashboard that shows one live customer-impact number but cannot reproduce the value used when the decision was made is not an evidence system.

The awareness timestamp is equally important. For a major primary outage, the CRTC, Innovation, Science and Economic Development Canada and the relevant emergency-management organizations must be notified within two hours after the provider becomes aware that it is a major outage. That is not simply the first device alarm and not necessarily the point when a senior executive joins the bridge. The operator needs a defensible record of what observation established reportability, who accepted it and when.

Other service classes make a single generic clock dangerous. Every 9-1-1 outage is treated as major for notification to the relevant public safety answering point. Wireless public-alerting and specialized services such as TTY relay, IP relay and 9-8-8 have their own conditions, recipients and 30-minute or two-hour timings. The major-primary path is a useful worked example, not a universal recipient matrix.

The first notice is an evidence snapshot, not a miniature postmortem. Decision 2025-225 requires the type and criterion that triggered reporting, the affected services and areas, outage start, expected restoration, end-user impact, networks and providers affected, cause if known, restoration steps and factors preventing repair. A provider that owns these fields only as prose in an email forces its incident team to reconstruct the same state repeatedly for regulators, emergency authorities, customer communications and internal leaders.

That reconstruction becomes harder after the first message. Providers must update the required authorities as soon as possible after a material change. Updates continue until all required information has been given, the outage has ended and every service is restored; the final update confirms the end. “Material” is not a synonym for “anything changed”. It needs an operating rule: a new affected service, a revised denominator, a geographic expansion, a missed restoration estimate, a new repair constraint or a transition from suspected to confirmed cause.

The clean control is a field ledger. Each required fact carries a value, an unknown state where appropriate, an observation time, an evidence source, an owner, the recipients who have seen it and the update that superseded it. The notification itself becomes a versioned view of the ledger. Engineers can continue restoring service while the reporting owner changes only the fields whose evidence has moved.

Restoration does not close the evidence chain. A post-outage report is due within 30 days after the major outage ends and services are restored. It covers root cause if known, affected networks and services, other providers, geography, end-user impact, restoration work, prevention measures with timelines and lessons learned. If information is claimed confidential, the provider must also submit a meaningful abridged version for publication.

That public abridgement changes the value of early discipline. A provisional estimate that was properly labelled, timestamped and later corrected can be explained. A number copied between emails without provenance becomes difficult to defend. The post-outage team should not discover that the restoration bridge, regulatory notice and public channel used three different denominators with no record of why.

The framework largely took effect on 4 November 2025, but its perimeter is still an object of procedure. Several providers sought review and variance of the decision, Telesat filed a separate application, and the CRTC combined the proceedings in February 2026 before issuing further requests for information. Those filings do not by themselves erase the decision. They do mean that the live form, the decision page and any later order should be checked before a company freezes its playbook.

An executive should see five timestamps for every reportable incident: first technical signal, first reliable impact estimate, reportability determination, initial notification and restoration. Beside them should sit the specific threshold evidence and the latest version of every required field. The root cause can arrive later. The evidence clock cannot be rebuilt later with the same authority.

Sources