Call Center Outsourced evidence brief · Desk review · Published
How should a contact center audit a vendor's disaster recovery call tree?
A drill method that tests reachability, authority, timing, and evidence instead of reading a contact list.

On this page
- Key stats
- Key takeaways
- Question and scope
- What the evidence supports
- Map the call tree as a dependency
- Test the first hop, not just the list
- Check authority to invoke the plan
- Time-stamp the test end to end
- Write findings the client can act on
- Limits of a single drill
- Put this into a support lane
- Related operating guides
- FAQs
- Sources
Key stats
- 2.67 million U.S. customer service representative jobs in 2025 (BLS)
- $21.53 median hourly wage in May 2025 (BLS)
- 289,500 projected annual openings, 2025 to 2035 (BLS)
Key takeaways
- Test the first alert hop at the hour it can actually fire.
- Record who can authorize a failover, not only who answers the phone.
- Time each step so the report shows where recovery stalled.
- Schedule the next drill before filing the current report.
Question and scope
This brief asks how a contact center should audit the disaster recovery call tree a vendor proposes for an outsourced queue. The population is a client with a primary site, a backup site, and a list of people to call when the primary fails. The unit is one drill, not the vendor's entire business continuity program. The audit covers reachability, authority, timing, and evidence. It does not certify that the plan will work in a real outage; record the scenario name in the drill plan so a later reviewer can tell exactly which failure was tested and which assumptions framed it.
Define the failure the tree is meant to handle before you test it.
A power outage, a network cut, and a building closure produce different first calls. Write the scenario and the expected first action in one paragraph. If the vendor cannot state the scenario, the tree is a contact list rather than a plan. Keep the scope to the scenarios the client actually depends on, and say which ones are out of scope for this drill; keep the out-of-scope list short and visible so the drill does not drift into a full continuity audit that nobody scheduled.
What the evidence supports
Federal continuity guidance and contact-center standards treat a call tree as a dependency, not a document. NIST and CISA materials emphasize tested recovery, named roles, and recorded results. ISO 18295 describes service continuity as part of the contact-center management system. None of these sources says how many names a tree should contain. The operating inference is to test reachability and authority, then record what failed; attach the source and date to each claim so the client can see which guidance shaped the test and when it was current.
The evidence is stronger on process than on outcomes.
A vendor can show a list and still fail to reach anyone at 2 a.m. A drill produces evidence only if it records the time of each attempt and the response. Treat the drill as a sample of the tree, not proof of resilience. State that limitation in the report so the client does not overread a pass, and keep the raw attempt log with the report. Label the attempt log as the primary evidence and treat the narrative as interpretation whenever the two disagree on timing.
Map the call tree as a dependency
Draw the tree from the first alert to the person who can authorize a failover. Each node should have a role, a primary contact, a backup, and a maximum response time. Mark the nodes that depend on the same phone system or the same building. A tree with ten names on one floor is not resilient. The map should show where a single failure removes several branches; ask each node to name the backup person during the drill, not only on the printed chart that may be out of date.
Include the client-side nodes, not only the vendor's staff.
The vendor may need a client approval to switch sites or to change a routing rule. If that approval sits with one person who is unreachable, the tree stops. Add the client's backup approver and the hours they are available. The client should confirm those names before the drill, and the vendor should note the confirmation date; confirm the client approver's time zone as well as the phone number, because a correct number at the wrong hour still fails.
Test the first hop, not just the list
The first hop is the alert to the on-call person who must act. Test it with a real message at a real hour. Record the time sent, the time acknowledged, and the channel used. A list that looks complete can fail at the first hop because the number is stale. Do not accept a test that starts with a person already expecting the call unless the scenario requires it; use a channel the on-call person actually monitors, not the one the vendor prefers in a plan written for the client.
Run the first hop twice with different scenarios and different people.
One test during business hours and one outside them shows whether the tree depends on office presence. If the outside-hours test fails, the fix may be a rotation rather than a new name. Record the failed number and the reason without exposing personal details in the client report. The pattern matters more than the individual; keep the failed attempt in the record even when a later attempt succeeds inside the window, because the delay still counts.
Check authority to invoke the plan
Reachability is not authority. The person who answers may not be allowed to declare a disaster or spend money on a backup site. Ask each node what they can authorize in writing. Test the approval path with a request that requires a decision, not a notification. If the approver is on leave, the tree needs a named delegate; authority that exists only in a job description is not tested authority; ask for the delegation in writing before the drill, not after a failed approval stalls the test and the client loses confidence.
Record the financial and contractual limits of each approver.
A supervisor may reroute calls but cannot sign for overtime. A client manager may approve a site switch but not a customer credit. The drill should surface these limits before an outage does. Update the tree when a role changes, not only when a phone number changes, and confirm the new holder has accepted the role; test one approval that requires spending money and one that requires a customer-facing decision, since the paths often differ.
Time-stamp the test end to end
Start the clock when the simulated failure is declared and stop it when the backup queue answers a test call. Record each step in between with a timestamp and an owner. A total time without the intermediate steps cannot show where the delay occurred. Use the same clock for every participant and state the time zone. The client should see the sequence, not just the duration; assign one timekeeper so every step time comes from the same clock and the same stated time zone.
Compare the drill time with the recovery objective in the contract.
If no objective exists, the drill produces a number without a standard. Write the objective before the next drill. A queue that can recover in two hours may still fail a promise of thirty minutes. The gap is the finding, not the raw time, and the report should say which step consumed the largest share; show the recovery objective and the actual time side by side in the client report, including the largest contributing step.
Write findings the client can act on
A finding should name the failed step, the evidence, the effect, and the owner. 'The tree is inadequate' is not actionable. 'The second approver did not answer after three attempts over twenty minutes, so the site switch waited' is. Keep personal phone numbers out of the client copy. Use roles and attempt times instead; end each finding with the change requested; put the finding owner's role next to the due date so the item survives a staffing change or a vendor account move.
Separate findings the vendor can fix from findings the client must fix.
A stale number in the vendor's rotation is a vendor fix. A missing client approver is a client fix. Mixing them lets each side wait for the other. Give each finding a due date and a retest method; the next drill should close the list rather than repeat it, so carry the open items into the next test plan; keep a single open-items list that both the client and the vendor can see and update after every planning call.
Limits of a single drill
One drill samples one scenario on one day with the people who happened to be available. It cannot prove the tree works during a regional event when roads, power, and networks fail together. It also cannot test the vendor's willingness to invoke the plan under commercial pressure. State these limits in the report. A passing drill is evidence of readiness, not a guarantee; note the participants' roles and whether any named backup was absent, because an untested backup is only a name.
Schedule the next drill before the report is filed.
Vary the scenario, the hour, and the first contact. Keep a small set of core checks stable so results can be compared. Review the tree after any role change, acquisition, or site move. The audit is a recurring control, not an annual document review, and the client should see the trend across drills rather than one result; compare this drill with the previous one before setting the objective for the next test and the next scenario.
Put this into a support lane
Bring the current call tree, the last drill log, and the contract's recovery objective so the review can test a real first hop.
Schedule a recovery drill reviewRelated operating guides
FAQs
Does a successful drill prove the vendor can recover?
No. A drill samples one scenario with the people who were available. It shows that the tested path can work, not that every path will work during a regional event.
How often should the call tree be tested?
Test at least twice a year and after any role, site, or vendor change. Vary the scenario and the hour so the test does not depend on the same small group.
Sources
- Occupational Outlook Handbook: Customer Service Representatives
- ISO 18295-1:2017 Customer contact centres
- Fact Sheet #64: Call Centers under the Fair Labor Standards Act
- GAO-20-283G: Assessing Data Reliability
- NIST Special Publication 800-92: Guide to Computer Security Log Management
- Cyber Essentials Toolkit
- ISO 9001:2015 Quality management systems
- ISO 10002:2018 Quality management - Customer satisfaction - Guidelines for complaints handling