On this page
- Evidence snapshot
- Decide what counts as an outage
- Write the first five minutes
- Data and decision boundary
- Move the record to a controlled fallback
- Tell customers what is true
- Protect promises made before the outage
- Reconcile the fallback when systems return
- Copy-ready call and handoff lines
- Review the gap with the client
- Keep the fallback small enough to test
- Questions managers ask
- Keep planning
- Sources
An outage plan earns its place when the queue can still account for every customer promise while the client system is down.
Evidence snapshot
Decide what counts as an outage
Write a definition that separates a slow system from an outage the queue must escalate. Include the systems covered, the users affected, and the evidence that starts the clock. A single agent's login failure is different from a client-wide record system failure. Publish the definition with the escalation contacts. Share the definition with the client's service desk so both teams declare the same event.
The definition should name who declares the outage and who can end it. A declaration made by several people at once produces conflicting customer messages. Give one role the authority and a backup for off hours. Record the declaration time and the first affected case. The backup should be reachable by phone when the primary contact is offline.
Write the first five minutes
The first five minutes decide whether the queue keeps a usable record. Tell agents what to stop, what to write down, and where to send the case. Move to the approved fallback only if the primary system is confirmed unavailable. Keep one person on the status channel so the floor hears the same update. The checklist should fit on one page and use the names the floor actually uses.
A short checklist beats a long procedure during an incident. Put the checklist where agents can reach it without the failed system. Include the fallback form, the customer message, and the supervisor contact. Test the checklist once a quarter so it does not describe a tool the team no longer uses. Review the checklist after every test and remove any step that no longer matches the tools.
Data and decision boundary
Use this table as a starting point, then match each row to the client's tools and call guide. The manager column stays outside the team member's normal authority.
| Data or request | Team member can | Manager keeps |
|---|---|---|
| Outage declaration | Report the failed function and the first case | Declare the outage and start the timeline |
| Fallback entry | Record the required fields and the owner | Approve edits and protect the original |
| Customer message | Use the approved sentence | Update the wording when the status changes |
| Reconciliation | Match each fallback row to the main system | Sign off before normal work resumes |
Move the record to a controlled fallback
The fallback record needs a case number, a timestamp, the customer request, the verification state, and the next action. It also needs an owner and a place to reconcile later. Use a form that limits free text so the fields can be merged back into the main system. A shared spreadsheet with no owner becomes a second source of truth. The fallback form should reject a row without an owner or a next action.
Protect the fallback from casual editing during the outage. Limit who can change a row and record the change. Keep a copy of the original entry before reconciliation. The fallback is evidence, and evidence that can be edited without a trace is weak. Keep the original entries in a read-only copy until the reconciliation is signed off.
Tell customers what is true
Give agents one approved sentence that says what is unavailable and what the customer should expect next. Do not promise a time the queue cannot control. Offer the channel that still works and explain how the case will be followed up. Update the sentence when the status changes. The customer sentence should avoid internal system names and blame.
Customers should not hear a different story from the queue, the website, and the account manager. Share the wording with the client before the next outage, not during it. Keep a record of when each version was used. A truthful message reduces repeat contacts even when the system is down. Store each approved version with the date it took effect.
Protect promises made before the outage
List the promises that were open when the outage began and give each one an owner outside the failed system. Contact the customers whose windows will be missed before they call to ask. Record the new commitment and the reason for the change. Do not let the outage erase an obligation the queue already accepted. The promise list should include the customer contact route, not only the case number.
The list should include callbacks, escalations, and refunds that were waiting on a system action. Some may be completed manually, and some may need the client's authority. Decide which is which before the shift ends. The customer should hear from the queue rather than discover the delay alone. Contact the customer before the window closes whenever the queue can still reach them.
Reconcile the fallback when systems return
Reconcile every fallback row against the main system before the team returns to normal work. Check for duplicates, missing cases, and actions that were taken twice. Give the reconciliation a named owner and a completion time. A queue that resumes without reconciling carries two versions of the same case. The reconciliation owner should be different from the person who entered most fallback rows.
Sample the reconciled cases for record quality, not only for count. Look for entries that were merged into the wrong case or lost their next action. Ask the agents who used the fallback what was hard to capture. Their answers should change the form before the next incident. Check the reconciled cases against the phone log before closing the incident.
Copy-ready call and handoff lines
Declaring the outage
The client case system is unavailable for the whole queue, and the first failed case is at ten past two. We are moving to the fallback form, keeping the promised callbacks with the named owners, and posting updates on the status channel.
Telling a customer
Our case system is unavailable right now, so I cannot confirm the refund status. I have recorded your request and the promise date, and the queue will call you before five with an update.
Closing the incident
The system returned at four, and all eighteen fallback entries are reconciled with no duplicates. Two promises moved to tomorrow, and both customers have been told.
Review the gap with the client
Bring the client a timeline, the affected cases, and the customer messages that were used. Separate the vendor's response from the client's system failure without assigning blame. Show what the fallback protected and what it could not. The review should end with specific changes to the plan. The client review should name the system change that would reduce the next outage.
Include the promises that were missed and the ones that were saved. A complete review shows both, because a plan that only lists failures hides the parts that worked. Agree on the next test date and the owner. Keep the review short enough to be read by the people who will act on it. Ask the client to confirm the timeline before it is shared more widely.
Keep the fallback small enough to test
A fallback that requires a new system, a long form, or a special team will not be ready when it is needed. Keep it small enough that the queue can test it in an hour. Test the form, the customer message, and the reconciliation step together. Fix the part that fails before the next test. The shared record should show which facts are confirmed and which are still estimates.
Store the test result with the outage plan and note who reviewed it. A plan that is tested once and never read again drifts from the tools in use. Put the next test date on the operations calendar. The goal is a fallback the floor can use under pressure, not a document that satisfies an audit. Put the next test date on the same calendar as the queue's other readiness checks.
Questions managers ask
Should agents keep taking calls during an outage?
Only for work the fallback can record safely. If the queue cannot capture the request and the promise, route the customer to a channel that can.
Who reconciles the fallback?
A named owner should reconcile before normal work resumes. The supervisor signs off on the match and on any promise that changed.
Sources
- NIST SP 800-34 Rev. 1: Contingency Planning Guide for Federal Information SystemsNational Institute of Standards and Technology, 2010. Contingency planning, testing, and recovery roles for information systems.Source 1
- NIST SP 800-61 Rev. 3: Incident Response RecommendationsNational Institute of Standards and Technology, 2025. Incident response lifecycle, communications, and evidence handling.Source 2
- ISO 18295-1:2017 Customer contact centresInternational Organization for Standardization, 2017. Requirements for customer contact centre processes, people, and performance.Source 3
- GAO-20-283G: Assessing Data ReliabilityU.S. Government Accountability Office, 2020. A practical framework for documenting data sources, testing reliability, and stating limitations.Source 4
