The technical fix is rarely what users remember about an outage. They remember whether they were told what was happening, when they would hear next, and whether the next update actually arrived. Incident communication fails in predictable ways: silence while engineers investigate, a vague "we are aware" post that never gets followed up, conflicting messages from support and engineering, and a premature "resolved" that has to be walked back. Each of those erodes trust more than the outage itself.
Good communication does not slow down incident response. It runs in parallel with it, owned by a different person, following a structure decided before the incident. The incident commander fixes the problem. The communications lead owns the status page, stakeholder updates, and internal coordination. When one person tries to do both, communication is always what gets dropped, usually at exactly the moment users are most anxious.
This article is a practical playbook for the first 60 minutes of a major incident and beyond: who does what, what to post and when, how to handle stakeholders who need more than a public status page, and how to close the loop afterwards. It assumes you already have monitoring and an on-call rotation; its focus is the communication layer on top of them.
#Quick Answer: Name One Communicator, Post Within 15 Minutes, Update Every 30
Within 15 minutes of declaring a major incident, publish a status page post that states what is affected, what users will experience, and when the next update will come, even if the cause is still unknown. Assign a dedicated communications lead who does nothing else during the incident, update at least every 30 minutes even when there is no new technical progress, and never declare resolution until the fix is verified in production. Saying "we don't know the cause yet, next update at 14:30" is far better than silence, because silence is interpreted as either ignorance or indifference, and both damage trust.
The first hour has a simple rhythm: declare and post, establish an update cadence, brief internal stakeholders before they learn about the incident from customers, share a workaround if one exists, and only then move to root cause messaging. Technical accuracy matters, but timeliness and predictability matter more in the early phase. Users can tolerate an outage they understand; they cannot tolerate being left in the dark.
#The First 15 Minutes: Declare, Assign, Post
Declaration is a decision, not a feeling. Teams that wait for certainty before declaring an incident lose the most valuable communication window. Define declaration criteria in advance: for example, a customer-facing service down for more than five minutes, error rates above an agreed threshold, or data integrity suspected. When the criteria are met, declare. You can always downgrade later; downgrading a declared incident costs little, while a late declaration costs credibility.
Once declared, assign roles immediately. The incident commander owns technical response and the decision to declare, escalate, and resolve. The communications lead owns all external and internal messaging. A scribe or deputy keeps the timeline, because reconstructing one afterwards from chat messages is unreliable and the timeline is essential for the post-incident review. If your team is small, one person may hold two roles for a minor incident, but for a major incident, communications must be a dedicated role, even if that means pulling someone off engineering work.
The first status post should be short and structured. State the affected service, the user-visible symptom, the scope if known, and the next update time. Avoid technical jargon and avoid blame, speculation, or promises about resolution time you cannot keep:
Investigating - Checkout is returning errors for some customers.
We are investigating elevated checkout errors affecting a subset
of orders. Placing an order may fail and show an error page.
Next update: 14:30 UTC.Notice what this post does not do. It does not say "a small number of users" when the scope is unknown, it does not name a suspected cause, and it does not promise a fix by a specific time. Each of those creates a second incident when it turns out to be wrong: the correction. If scope is genuinely unknown, say so and say what you are doing to determine it. Honesty about uncertainty reads as competence; false precision reads as spin once it is corrected.
#Running the Update Cadence
An update cadence is a promise, and keeping it is more important than having news. Commit to updates every 30 minutes for a major incident, or every 60 minutes for a lower-severity one, and publish on schedule even if the update is "still investigating, no change." Users who know the next update is coming at a fixed time stop refreshing the page compulsively and stop opening support tickets asking for news, which reduces load on the team that is already under pressure.
Each update should move the story forward in at least one dimension: scope refined, workaround identified, cause confirmed, fix in progress, fix deployed and being monitored, or resolution verified. A status lifecycle helps here. Most status page tools support states such as Investigating, Identified, Monitoring, and Resolved. Use them consistently, and do not skip from Investigating straight to Resolved without a Monitoring phase. The Monitoring phase exists because fixes sometimes fail under real load, and declaring resolution too early forces an embarrassing re-opening.
Workarounds deserve special attention. If users can achieve their goal another way, even imperfectly, say so early and prominently. A workaround that reaches 80 percent of users is worth more than a perfect fix that arrives two hours later. Be specific: name the alternative path, its limitations, and whether data entered via the workaround will be preserved. Vague advice such as "please try again later" is not a workaround; it is a polite way of saying nothing.
Internal updates need the same discipline as external ones, with more detail. Support teams should never learn about an incident from a customer ticket, and sales or account teams should never learn about it from social media. Send an internal briefing within 30 minutes of declaration: impact, affected customers if known, workaround, next update time, and a single point of contact for questions. Give support a holding response they can paste, so customers get a consistent message regardless of which channel they use. Inconsistent answers across channels are a common complaint in post-incident reviews, and they are entirely preventable.
#Stakeholders, Executives, and Customers Who Need More
A public status page is necessary but not sufficient. Enterprise customers, regulators, and executives often need direct communication, and they need it before they ask for it. Segment stakeholders in advance: who gets a phone call for a severity-1 incident, who gets an email, and who is served adequately by the status page. Waiting until an incident to decide this guarantees delay and argument about who should have been told.
For enterprise customers, assign a named contact per key account during the incident, and proactively email their technical contact with impact, workaround, and next update time. Do not make them discover the outage themselves and then chase your support team. If you have contractual notification obligations, such as service-level agreement credits or incident notification windows, the communications lead should know them before the incident, not discover them during it. A missed contractual notification can turn a technical incident into a commercial dispute.
Executives need a different kind of update: short, business-focused, and free of technical speculation. A useful executive briefing covers business impact in plain terms, customer impact, estimated timeline in ranges rather than false precision, what is being done, and what decisions are needed from them, if any. Executives who are kept informed do not interrupt the response team for updates; executives who are kept in the dark will, usually at the worst moment. One short briefing per hour, plus a message at declaration and resolution, is usually enough.
Social media is a separate channel with its own risks. Do not ignore it, but do not let it become the primary incident channel either. Post a short acknowledgement linking to the status page, and direct replies there, so there is one source of truth. Never debate technical details or individual customer cases in public replies. If misinformation spreads, correct it once on the status page and link to that correction, rather than arguing thread by thread. The status page is the record; social media is a pointer to it.
#Closing the Incident Without Creating a Second One
Resolution messaging is where many teams undo good work. The temptation is to declare victory the moment a fix deploys. Resist it. Verify the fix in production with real traffic and error rates, confirm with support that customer reports have stopped, and only then move to Resolved. Include a brief summary of what happened, in plain language, and state whether a fuller post-incident review will follow and when. Users do not need a root cause in the resolution post, but they do need to know the incident is genuinely over and what happens next.
The post-incident review, or postmortem, is a communication artifact as much as a technical one. Write it promptly, while the timeline is fresh, and share an appropriate version externally for major incidents. A good public postmortem covers what happened, impact and duration, timeline, root cause in understandable terms, and concrete preventive actions with owners. It does not assign individual blame, and it does not hide behind passive language. Phrases such as "mistakes were made" damage trust; naming the systemic cause and the fix builds it. Customers who read a candid, specific postmortem often report higher trust afterwards than before the incident, because they have seen how the organisation responds under pressure.
For teams building operational maturity, incident communication sits alongside alerting and on-call health. Alert fatigue, for example, delays declaration because responders stop believing pages, and a related guide on reducing SOC alert fatigue covers the tuning discipline that keeps signals trustworthy, even though its context is security operations. The principle transfers directly: if responders cannot trust the alert, they cannot declare promptly, and communication starts late through no fault of the communicator.
Finally, practise. Run a tabletop exercise focused purely on communication: hand the team a scenario, a ticking clock, and a blank status page, and see what they post in the first 15 minutes. Most teams discover gaps in role assignment, stakeholder lists, or status page access that are trivial to fix in advance and painful to discover during a real outage. A communications playbook that has never been exercised is a draft, not a plan.
#Frequently Asked Questions
What should the first status page post say if we do not know the cause yet?
Say what is affected, what users will experience, that you are investigating, and when the next update will come. Cause is not required for a useful first post. "We don't know yet" is an honest, complete answer in the first 15 minutes; a guessed cause that turns out wrong creates a second credibility problem when you correct it.
How often should we update during a major incident?
At least every 30 minutes, on schedule, even with no new technical progress. A "still investigating" update on time is better than a detailed update two hours late, because predictability reduces user anxiety and support ticket volume. For lower-severity incidents, hourly updates are usually sufficient. State the cadence in the first post so users know when to expect news.
Who should write incident updates, the engineer fixing the problem?
No. Assign a dedicated communications lead who does nothing else during a major incident. The engineer fixing the problem should brief the communicator in short, regular bursts, and the communicator turns that into clear external and internal messages. When one person does both, communication is always what gets dropped under pressure.
Should we post a public postmortem after every incident?
For major, customer-facing incidents, yes, in proportion to impact. A short public summary with timeline, cause in plain language, and preventive actions builds trust. For minor incidents, an internal review is usually enough. The decision should be made by severity criteria defined in advance, not by how embarrassing the incident feels in the moment.
What if executives or customers pressure us to promise a resolution time?
Give a range and state the uncertainty honestly, or give the next update time instead of a fix time. For example: "We expect a fix within one to three hours, and we will update at 15:00 either way." Never promise a specific time to relieve pressure in the moment; a missed promise becomes a second incident and undermines every later update.
#Sources and Further Reading
- Google Site Reliability Engineering book, chapters on incident response and postmortems: https://sre.google/sre-book/table-of-contents/
- Atlassian incident handbook, communication templates and severity guidance: https://www.atlassian.com/incident-management/handbook
- PagerDuty incident response guide, roles and stakeholder communication: https://www.pagerduty.com/resources/learn/incident-response/
- NIST Special Publication 800-61, Computer Security Incident Handling Guide: https://csrc.nist.gov/pubs/sp/800/61/r2/final
- "The Field Guide to Understanding Human Error" by Sidney Dekker, on blameless review culture: https://www.routledge.com/The-Field-Guide-to-Understanding-Human-Error/Dekker/p/book/9781472439055