Every SaaS product goes down eventually. A database migration takes longer than planned, a cloud provider has a bad hour, a deploy breaks login for one region, or an integration partner changes something without warning. Customers rarely leave over one incident. They leave over how it felt: finding out from their own users, refreshing a status page that says "All systems operational" while nothing works, or getting a vague apology a week later that explains nothing.
Small teams struggle with incident communication for an obvious reason. The same three people who are fixing the problem are the ones who would write the update, and in the moment, fixing wins. The answer is not more people. It is deciding in advance who writes, what they say, when they say it and where it goes, so that during an incident nobody has to think about wording.
This guide gives you a severity scale, a clear owner model, and copy-ready templates for each stage: investigating, identified, monitoring, resolved, a direct email to affected customers, a scheduled maintenance notice, an in-app banner, a support macro and a post-incident summary. It is written for founders, engineering leads, support and customer success leads, and PMs at small B2B SaaS companies in the US and Europe.
No product is promoted here. Everything below works with whatever status page, email tool and help desk you already use, or with a shared document and your inbox.
Decide severity before anything breaks
Your team should be able to pick a severity level in under a minute. Keep it to three or four levels and write them down where everyone can find them.
- SEV 1: What it means: The product is unusable for most customers, or data is at risk; Examples: Login fails for everyone; data not saving; security incident; Who is told: Status page, in-app banner, direct email to all affected accounts, account owners call top accounts; Update every: 30 minutes
- SEV 2: What it means: A core feature is broken or badly degraded for many customers; Examples: Exports failing; one region slow; a major integration down; Who is told: Status page, in-app banner, direct email to affected accounts; Update every: 60 minutes
- SEV 3: What it means: A minor feature is broken, or a small group is affected; Examples: One report type wrong; a rare edge case; Who is told: Status page (optional), support macro, direct reply to those who wrote in; Update every: When there is news
- Maintenance: What it means: Planned downtime or degradation; Examples: Database upgrade; data center move; Who is told: Notice 3 to 7 days ahead, reminder the day before, status page during; Update every: Start and end
When in doubt, choose the higher level. Downgrading an incident later is easy and nobody minds. Discovering that customers were affected for two hours with no update is the thing they remember.
Security incidents involving customer data follow your legal and contractual obligations first. Bring in whoever handles legal and privacy before you publish anything, and do not use the general templates below for those without their review.
Who owns what
Name these roles in advance. In a team of five, one person may hold two of them, but never the incident lead and the communications lead at once during a SEV 1.
- Incident lead: runs the fix, decides severity, tells the communications lead what is true right now.
- Communications lead: writes and posts every update, keeps the update clock, drafts the customer email. Often a support lead, founder or PM.
- Support lead: applies the support macro, tags incoming tickets to the incident, tells the communications lead what customers are reporting.
- Account owners: call or message the largest affected accounts personally during a SEV 1.
- Approver for the post-incident summary: usually the engineering lead or founder.
The single most useful rule: the communications lead posts nothing the incident lead has not confirmed, and the incident lead does not have to write anything.
Five rules for every update
- Say what customers see, not what broke inside. "You may not be able to log in" is useful. "Our auth service is returning errors" is not, to most readers.
- Never guess a fix time. Say when the next update will come instead. If you do give an estimate, it must come from the incident lead and be stated as an estimate.
- Post on the clock, even with no news. "Still investigating, next update at 14:30 UTC" tells customers you have not forgotten them.
- Use one time zone and say which. UTC is the safest choice for a customer base across the US and Europe. Add a local time only if most customers share one zone.
- Same words everywhere. The status page, banner, email and support replies should match. Customers compare them.
The templates
Replace everything in curly braces. Keep the structure; trim the words to what is true.
1. Investigating (first update, within 15 minutes of confirming the problem)
Title: {Short description of the symptom}
We are investigating reports that {what customers see, e.g. some customers cannot log in / exports are failing / the dashboard is loading slowly}.
Affected: {who, e.g. customers in the EU region / customers using the {integration name} integration / all customers}.
We will post an update by {time} UTC, or sooner if we learn more.
Post this even if you are not yet sure how big the problem is. A first update that says "some customers" is better than a perfect update an hour later.
2. Identified
We have identified the cause of {symptom}. {One plain sentence on the cause, e.g. "A configuration change made at 09:40 UTC affected how sessions are created."}
We are {what you are doing, e.g. rolling back the change / restoring the affected service}.
{If customers can do anything: "In the meantime, {workaround}."}
Next update by {time} UTC.
Only share the cause when the incident lead is confident. If you are not sure, skip the second sentence and say you are working on a fix.
3. Monitoring
A fix has been applied and {symptom} should be resolved for most customers. We are monitoring to make sure it stays that way.
If you are still seeing {symptom}, please {action, e.g. log out and back in / refresh the page}, and contact support if that does not help.
Next update by {time} UTC.
4. Resolved
This incident is resolved. Between {start time} and {end time} UTC, {who} {what they experienced}.
{Data statement, only if confirmed: "No data was lost." / "Data submitted during this window was saved and is now visible." / "Some {items} created during this window may need to be resent; details below."}
We are sorry for the disruption. We will publish a summary of what happened and what we are changing by {date}.
Do not write "no data was lost" unless the incident lead has checked. If you do not know yet, say you will confirm in the summary.
5. Direct email to affected customers (SEV 1 and SEV 2)
Send this once the incident is resolved, or during a long SEV 1 if the status page is not enough. Send it from a named person, not a no-reply address.
Subject: {Product} {symptom} on {date}: resolved
Hi {first name},
On {date}, between {start} and {end} UTC, {what customers experienced, in one sentence}. {Account-specific line if you have it: "Your workspace was affected for about {duration}."}
It is now resolved. {Data statement, if confirmed.}
What happened, briefly: {one or two sentences in plain language}.
What we are doing about it: {one or two concrete changes, e.g. "We have added a check that blocks this kind of configuration change from reaching production without a second review."}
{If customers need to act: "One thing to check: {action}."}
I am sorry this got in the way of your work. If anything still looks wrong, reply to this email and it will come straight to me.
{Name}
{Role}
6. Scheduled maintenance notice
Send 3 to 7 days ahead, with a reminder the day before.
Subject: Scheduled maintenance on {date}, {start} to {end} UTC
Hi {first name},
We are {what you are doing, in customer terms, e.g. upgrading our database to make reports faster} on {date} between {start} and {end} UTC.
During this window: {what customers will notice, e.g. the app will be read-only / you may be logged out once / imports will be paused}.
What still works: {e.g. your public pages and forms keep working; submissions are queued and processed afterwards}.
You do not need to do anything. {Or: "If you have an import planned in that window, please run it before {time}."}
We will post updates on our status page during the work.
{Name}
Pick a window when your customers are least active, and check it against their time zones, not yours.
7. In-app banner
Keep it to one line plus a link to the status page.
{Symptom} for some customers. We are working on it. Next update by {time} UTC. {Link text: "Status updates"}
Remove the banner as soon as the incident is resolved, and replace it for a day with: "Earlier today, {symptom}. This is resolved. {Link text: "What happened"}."
8. Support macro
For replies to customers who write in during the incident.
Hi {first name},
Thanks for letting us know. We are aware that {symptom} and the team is working on it now. You can follow updates on our status page, and we will post the next one by {time} UTC.
I have linked your message to the incident, so you will hear from me when it is resolved. {If there is a workaround: "In the meantime, {workaround}."}
{Name}
Then actually reply to every linked ticket when the incident is resolved. Customers notice when that second message never comes.
9. Post-incident summary
Publish within 5 working days for SEV 1, and within 10 for SEV 2. Blameless, plain language, specific about changes.
Title: Summary: {symptom} on {date}
What happened. On {date}, between {start} and {end} UTC ({duration}), {who} {experienced what}.
Impact. {Number or share of accounts affected, if known and confirmed. Which features. Whether any data was affected, and what customers need to do, if anything.}
Timeline (UTC).
{time}: {first sign of the problem}
{time}: {team alerted / incident declared}
{time}: {first public update}
{time}: {cause identified}
{time}: {fix applied}
{time}: {resolved}Root cause. {Two to four sentences, plain language. Describe the system and the gap, not the person.}
Why it took as long as it did. {Honest note on detection or recovery time, e.g. "Our alerts watched error rates, not failed logins, so the first signal was a customer report."}
What we are changing.
- {Change}, {owner}, {target date}
- {Change}, {owner}, {target date}
- {Change}, {owner}, {target date}
We are sorry for the disruption. If you have questions about how this affected your account, reply to {contact} and we will answer directly.
The "what we are changing" list is the part customers read most closely. Two or three concrete changes with owners and dates beat a long list of vague intentions. Follow up when they ship.
The incident log
Keep one row per incident in a shared sheet. It turns incidents into patterns you can act on. Use these columns:
- Date
- Severity
- Symptom
- Start (UTC)
- End (UTC)
- Duration
- Accounts affected
- First update after (minutes)
- Summary published
- Changes promised
- Changes shipped
Review it monthly. Two numbers matter most: how long it took to post the first update, and how many promised changes actually shipped.
Timing at a glance
- Problem confirmed: What goes out: Investigating update; banner for SEV 1 and 2; Owner: Communications lead
- Every 30 or 60 minutes: What goes out: Status update, even with no news; Owner: Communications lead
- Large accounts affected (SEV 1): What goes out: Personal call or message; Owner: Account owners
- Cause found: What goes out: Identified update; Owner: Communications lead
- Fix applied: What goes out: Monitoring update; Owner: Communications lead
- Resolved: What goes out: Resolved update; banner changed; support tickets answered; Owner: Communications and support leads
- Same or next day: What goes out: Direct email to affected customers; Owner: Communications lead, signed by a named person
- Within 5 to 10 working days: What goes out: Post-incident summary; Owner: Approver
- When changes ship: What goes out: Short follow-up to the same audience; Owner: Communications lead
Incident communication mistakes to avoid
- Waiting until you know the cause. Customers need to know you know. The cause can come later.
- "All systems operational" during an outage. If your status page is automated, check it during every incident and override it when needed.
- Fix-time guesses. "Should be fixed in 10 minutes" that turns into three hours does more damage than "next update in 30 minutes."
- Jargon. Customers do not need service names. They need to know what they can and cannot do.
- Minimizing. "A small number of customers" when it was half your base will be noticed.
- Blaming a vendor in public. Name a provider outage if it is a fact and relevant, but own the impact on your customers. They chose you, not your provider.
- No follow-through. A summary that promises changes nobody ships makes the next incident worse.
FAQ
How quickly should we post the first incident update?
Within 15 minutes of confirming customers are affected. It can be short and say you are still investigating.
How often should we update during an outage?
Every 30 minutes for a severe incident and every 60 for a major one, even if the update is "still working on it, next update at {time}."
Should small SaaS companies publish post-incident summaries?
For severe and major incidents, yes. A short, honest summary with two or three concrete changes builds more trust than silence, and writing it forces the team to agree on what will change.
Should we email customers about every incident?
No. Email accounts that were affected by a severe or major incident. For minor issues, the status page and direct replies to customers who wrote in are enough.
Who should sign the customer email?
A named person customers can reply to, usually a founder, the head of support or the account owner. Not a no-reply address.
The short version
Agree severity levels, owners and the update clock before anything breaks. During an incident, post what customers see within 15 minutes, update on schedule even with no news, never guess a fix time, and use the same words on the status page, banner, email and support replies. Afterwards, email affected accounts from a named person, publish a plain summary with two or three changes that have owners and dates, and tell customers when those changes ship.
RG
Top comments (0)