How to handle a spike of support tickets during a SaaS outage

Declare early, name one person to speak, post one update everywhere on a fixed rhythm, and write to everyone who complained once it is fixed.

Thinkiyo team··12 min read

When tickets spike during an outage, stop answering them one by one. Declare an incident, name one person who speaks for it, publish one short update, and paste that same update into every channel on a fixed rhythm. Keep a list of everyone who wrote in, so each of them hears when it is fixed. Support's job in the first ten minutes is to tell engineering what customers are seeing and to stop the team writing twenty different explanations.

What the people who run incidents say

The published guidance is short and it agrees.

Declare early. Google's SRE workbook says to "declare incidents early and often", and that its experience is that managed incidents get resolved faster. Support is often in a position to make that call first, because customers report what monitoring has not caught. Agree in advance that support can raise one.

Split the roles. Google's model separates an incident commander, an operations lead and a communications lead. The communications lead is the public face, and only the operations team changes the system. PagerDuty's public playbook has a customer liaison who drafts customer messages, gets the commander's approval, and passes back how many customers are reporting the problem and what they see.

Write it down in one place. Google keeps a living incident document with the most important facts at the top. Support should read from that and nowhere else.

Prepare the words. Google suggests two or three ready-to-use templates. PagerDuty keeps a default first message for when the scope is still unknown.

Update on a rhythm. Atlassian's Statuspage guidance suggests updates every 30 minutes, or whatever suits the situation, and the same message on every channel.

Own it. The same page points out that customers may not care that another provider caused the problem. To them it is your service.

Don't estimate the fix. PagerDuty advises against giving resolution times, because a missed estimate frustrates people more than none. Give the time of the next update. That is a promise you can keep.

What you may owe afterwards

The principle that holds almost everywhere: an outage is mostly a matter of your contract, and consumer law sits underneath it where the customer is a consumer. A personal data incident is a separate track with its own clock, and it can start in the middle of an outage. Whether a given outage breaches anything depends on the facts. This is general information and not legal advice. Check the rules in your own market and read your own terms.

MarketService failureIf personal data is involved
USWe know of no general federal outage credit rule, so your contract and SLA do the workListed companies file Form 8-K Item 1.05 four business days after deciding a cyber incident is material. Other rules may apply, so take advice
UKServices must be performed with reasonable care and skill for consumersNotify the regulator within 72 hours where feasible, unless risk to people is unlikely
EUDigital services have a minimum 2-year guarantee. If the supplier can't fix a fault in reasonable time, a consumer can ask for a price reduction or end the contract72 hours where notification is required. Availability incidents count as breaches, including ransomware and lost data with no backup
AustraliaConsumer guarantees require due care and skill. Business buyers under A$100,000 can be coveredAssess a suspected breach within 30 days. Includes loss of personal information likely to lead to unauthorised access
IndiaCheck local advice for your contract typeCERT-In directions require listed cyber incidents, including denial of service attacks, to be reported within 6 hours of noticing
CanadaCheck local advice for your contract typeNotify individuals as soon as feasible once you determine a real risk of significant harm, which also triggers a report to the regulator
SingaporeCheck local advice for your contract typeNotify the PDPC within 3 calendar days of determining a breach is notifiable, after an assessment within 30 days

Support does not decide any of this. Support's job is to notice the words "I can see someone else's data" and escalate at once. Decide before the incident who that goes to.

Service credits. Read your own SLA before someone asks. A common shape is that of Google Workspace: 99.9% monthly uptime, credits of 3, 7 or 15 days depending on how far it falls, a support case within 30 days or the credit is lost, and credits as the sole remedy under the SLA. If your contract requires customers to claim, say so in the closing message. Hiding it just produces a second wave of tickets.

Who does what

SituationOwnerFirst moveNever
Several tickets, same symptomSupport leadRaise an incident, give engineering the count and samplesWait for a dashboard to agree
Confirmed outageCommunications leadPost first update, share the macroLet each agent describe it in their own words
Partial, some customers onlyCommunications leadSay who is affected and who isn'tReport "all systems normal" for the rest without checking
Someone sees another customer's dataSecurity and legalEscalate at once, preserve the messageDiscuss it in a public reply or call it a glitch
Enterprise customer with an SLAAccount ownerPersonal note, link to the status pagePromise a credit before checking the contract
Public postSupportShort reply, status linkArgue about whether it was "really down"
Customer's own problem during an outageSupportCheck it against the incident firstPaste the outage macro without reading

Replies you can copy

Fill in the brackets. Use the same wording in every channel.

Status update, first

We're investigating a problem with invoice PDF exports. Exports may come out blank or fail to load. Other features are working. Next update by 2:45pm UTC.

First reply to a customer

Hi Daniel, you're not doing anything wrong. Invoice PDF exports are failing for a number of customers and we're working on it. You don't need to do anything or send more details. I've added you to the list and you'll hear from us as soon as it's fixed.

Public reply

It's on us, and we're working on it. Live updates here: [status link]. Nothing you need to do on your side.

Update when there is no news

Still working on invoice PDF exports. We've ruled out the template service and we're looking at the file store. No fix yet. Next update by 3:15pm UTC.

Closing reply

Hi Daniel, exports are working again as of 4:02pm UTC. Anything you exported between 2:04pm and 4:02pm may be blank, so please export those again. We'll publish what happened by Friday. Sorry for the lost afternoon, particularly with a client waiting.

Hard case: asking for compensation

Hi Daniel, I understand. Under your plan, service credits apply when monthly uptime falls below the level in our SLA, and we'll know that figure at month end. If it applies, I'll put the credit through and tell you the amount. I can't promise one today, but I will come back to you on 3 December either way.

Hard case: the customer's problem isn't the outage

Hi Mei, we do have an export problem today, but your error looks different. It's a permissions message, and the export issue doesn't produce one. Can you check whether your role changed recently? If it's not that, send me the time it happened and I'll look at it myself.

A timeline to adopt on Monday

These times are a suggested starting point. They are not a benchmark, so adjust to your product.

Before anything breaks

  1. Write down who can declare an incident, and include support.
  2. Save a first message, an update and a closing message.
  3. Name a backup for the communications role across every time zone you serve.
  4. Decide who hears about possible data exposure, and how.

During

  1. Within 10 minutes of the third matching report, raise an incident.
  2. Within 15 minutes, post the first update on the status page.
  3. Share one macro with every agent. Set an auto-reply on the busiest channels pointing to the status page.
  4. Every 30 minutes, post an update even if nothing changed.
  5. Keep sending engineering the count, the customer segments and any odd cases.
  6. Tag every related ticket so you can reach them later.

After

  1. Message every tagged customer the same day.
  2. Tell people what to redo, if anything was lost or wrong.
  3. Publish what happened, and link it from the closing message.
  4. Have support review the first hour, since that is where the macros and the handoffs get tested.

Mistakes teams make

  • Waiting for engineering to confirm before saying anything. "We're looking into reports" needs no root cause.
  • The status page says green while the inbox is on fire. Customers screenshot that.
  • Giving a fix time, then missing it twice.
  • Closing tickets when the macro goes out, so nobody is told about the fix.
  • Letting agents guess the cause. One wrong guess gets quoted back for weeks.
  • Replying in public with an apology and no link. The next person then asks the same thing.
  • Offering credits case by case, so the loudest customers get more.
  • Treating the outage macro as the answer to everything that arrives that day.

What to measure

  • Time from first customer report to incident declared.
  • Time from declaration to first public update.
  • The longest gap between updates.
  • Share of affected customers who received a closing message.
  • Tickets reopened within 48 hours of the fix. A high number can mean the fix did not hold or the message missed something.
  • Contacts per incident, tracked over several incidents. It should drop as the status page gets used.

It is working when customers quote your status page back to you and fewer people ask whether it is just them.

A worked example

This is an example scenario. The product, people and messages are made up to show how Relay works. It is not a customer story.

Ledgerly is an invoicing tool. Around 2pm UTC its PDF exports start failing.

Daniel Okafor emails: "Every invoice export gives me a blank PDF since about 2pm. I have a client waiting." Mei Lin posts on X: "@ledgerly exports are broken again? Getting a spinning wheel then nothing." A Reddit thread asks: "Is anyone else's Ledgerly export dead today or is it just me?" Arjun Mehta emails: "Export button does nothing. Is this a known issue?"

Four messages, three channels, one outage. Those messages are grouped as one theme, invoice export failing. Export complaints normally come once or twice a day, so a dozen inside fifteen minutes crosses the alert threshold and the alert goes to the team's Slack channel with the count, the channels and samples. The owner of exports picks it up with a clock running.

The team writes one update. Each customer gets a reply built from it on the channel they used, and an agent approves each one. The X post gets a short public reply. When the fix ships, every message in the theme is still linked, so the closing note goes to the whole list.

What runs in Relay today

The shared inbox, tagging, theme grouping, crisis alerts with learned baselines, Slack and signed webhook alerts, routing, SLA clocks, the war room view, drafted replies that a person approves and reopen rate reporting run in Relay today. Gmail and Outlook connect once you add an access token, public replies on X and Reddit need each channel's token, and email alerts work once an email sending key is added.

Status page updates and Jira, Linear and GitHub sync are on the way. Until then, the team posts the status update and files the incident in its own tools. If you hear about outages from angry customers first, contact us. For the reply side of a public pile on, see how to do online reputation management. For bug reports that arrive without an outage, see duplicate bug reports.

Common questions

How often should we update during an outage?

Atlassian's guidance suggests every 30 minutes or whatever suits the situation. The promise that matters is the time of the next update, so make it one you can keep. Post at that time even if the news is that nothing has changed.

Should we give an estimate for the fix?

PagerDuty's playbook advises against it, because unmet estimates frustrate customers. Say what you are doing and when you will next update. Give a time only when the fix is already rolling out.

Who should write the updates?

One named person, separate from the people fixing the problem. Google's model calls this the communications lead, and PagerDuty calls it the customer liaison. Have the incident lead approve each message before it goes out, as things can change while you draft.

Do we owe customers money for downtime?

It depends on your contract, your SLA and the local law where the customer is a consumer. Read the SLA before an incident, including how a customer must claim. Take advice on larger losses.

How does Relay tell an outage from a normal busy day?

Relay learns the usual volume for each theme and raises an alert only when a theme climbs well above it and stays there. A busy Monday lifts every theme a little. An outage lifts one theme a lot.

Can Relay send the update to every affected customer on its own?

No. Relay drafts each reply from the update your team wrote, and a person approves it before it goes out. On a large outage, plan for that approval time.

Sources

See what Relay can do for your customer service

Get a personalised 30 minute demo on your own channels and policies. We show you Relay handling your real conversations, then set it up with you.

  • Built and connected for you
  • Can run in your own cloud
  • A person on every refund

Brands we work with

  • Bolt
  • Dave & Buster's
  • ETV Win
  • Malachite Cars
  • Space Angel
  • Infinity Quest
Made by Thinkiyo© 2026 Thinkiyo. Relay is a Thinkiyo product.Examples use Lumen Parcel, a fictional brand.