Key takeaways
- Measure routing from recognition to acknowledgement, not from send to resolution.
- Report speed and misroute rate together so a fast wrong route is visible.
- Test the written rule with tabletop exceptions before relying on it.
Question and scope
The review asks whether a written escalation rule changes how quickly a virtual assistant routes an exception to the right owner. Routing time starts when the assistant recognizes the exception and ends when the correct owner receives it. It does not measure how long the owner takes to resolve the issue. Include the exception type and the owner in every row so the route can be reconstructed.
Scope the review to exceptions that require a decision above the assistant's authority. Exclude routine questions that the task guide answers. Use a fixed window before and after the rule is introduced, or compare two similar queues. Write down the exception types included so the comparison is not distorted by a new category. A written rule should name the channel where the escalation is sent and the information it must carry.
| Rule element | Observable test | Common failure | Fix |
|---|---|---|---|
| Trigger | Event can be named | Vague urgency | Add examples |
| Severity | Consequence stated | Two levels sound alike | Define the outcome |
| Owner | Role and person named | Shared inbox only | Name a backup |
| Window | Response time stated | No coverage after hours | Add a fallback |
What a written rule includes
A written rule needs an observable trigger, a severity level, an owner, a backup, and a response window. It also needs a channel for the escalation and a list of information that must travel with it. Without those elements, the assistant is still guessing even when a document exists. The acknowledgement should come from the owner, not from an inbox auto-reply.
The rule should define what counts as resolved for routing purposes. A message sent to the owner is not the same as an acknowledgement from the owner. Record both times if the process allows it. The difference shows whether the delay sits with the assistant or with the receiving side. Record the time zone for every timestamp so the elapsed time is not distorted.
Method and evidence standard
Use the exception log and the message timestamps to build the sample. For each exception, record the recognition time, the send time, the acknowledgement time, and the owner. Code whether the rule was used and whether the assistant chose the correct severity. Keep the original messages for review. A severity disagreement should be discussed with the owner before the next sample.
Have a second person code a sample of the exceptions. Compare the severity choice and the recognition time. A disagreement about severity is itself a finding, because it shows where the rule is ambiguous. Revise the definition before continuing if the disagreement is systematic. Report the median because a single slow exception can make the average misleading. A third coder is unnecessary unless the disagreement rate is high.
Measures and findings
Report the median and the range of routing time, not only the average. A few slow exceptions can pull an average away from the typical experience. Report the share routed within the target window for each severity. Then report the share sent to the wrong owner. Separate the misroute rate from the late-route rate so the repairs do not blur.
Separate the time spent deciding from the time spent waiting for an acknowledgement. A rule can speed up the decision while the owner remains slow to respond. Report both because they lead to different repairs. Describe the result as local to the reviewed queue and window. A rule that is hard to find should be moved before its wording is rewritten.
Where routing time is lost
Routing time is often lost before the message is sent. The assistant may search for the rule, compare the exception with several examples, or wait for a second opinion. A rule that is hard to find adds time even when its content is correct. Put the rule where the work happens, not in a distant folder. A role name without a current person is a common cause of a stalled route.
Time is also lost when the owner is unknown or unavailable. A role name without a current person, or a primary owner without a backup, creates a pause. The exception may be routed correctly and still sit unread. Record who received it and whether they acknowledged. Compare the number of steps in the old and new routes, not only the elapsed time.
Interpreting the pattern
Faster routing after the rule is introduced is not proof that the rule caused the change. The team may have been paying more attention because the rule was new. Compare the pattern across several weeks and across more than one exception type. Look for a change that persists after the novelty fades. The novelty of a new rule can make the first two weeks look better than the steady state.
A rule can also make routing slower if it adds a mandatory approval step that the old process did not have. Check the number of steps, not only the elapsed time. A slightly slower route to the correct owner may be better than a fast route to the wrong one. Judge the rule on both speed and accuracy. A fast route to the wrong owner should not be reported as an improvement.
Limitations
The sample may be small, and the exception types may change during the window. Timestamps can be missing or recorded in different time zones. The person coding the severity may know the outcome, which can bias the judgment. State these limits plainly. Missing timestamps should be reported as a data limitation, not filled with an estimate. Report the number of exceptions with complete timestamps.
The review cannot show whether the rule improved the client outcome. Faster routing may still lead to a slow resolution. A separate measure of resolution time is needed. Keep the routing finding separate from the outcome finding. The review should keep the resolution time separate from the routing time. Put the outcome measure in a separate section with its own sample.
Practical next step
Publish the rule in the same place as the task instructions and link it from the exception log. Give the assistant one example for each severity and one example of a misrouted exception. Name the owner and backup in the rule, not only in a directory. Review the names every month. Publish the rule beside the exception log so the assistant can reach it during the task.
Add a short acknowledgement requirement for the owner. The assistant should know whether the exception was received. If the owner cannot respond, the backup should be visible without another search. Test the path with a real exception before the next busy week. An acknowledgement requirement should be short enough to use during a busy shift. The owner should be able to acknowledge from the same channel.
Testing the rule before relying on it
Run three tabletop exceptions with the assistant before the rule goes live. Ask them to name the trigger, severity, owner, and next message. Time the decision, not the conversation. Compare the result with the written rule and fix the ambiguous parts. A tabletop test should include one exception where the primary owner is away. Use one exception from the past month so the scenario feels concrete.
Then run the rule for two weeks and review the exception log. Check whether the same exception type is routed the same way. Ask the owners whether the information they receive is sufficient. A rule that is never tested is only a hope about behavior. The two-week review should compare the same exception types before and after the rule. Record any change to the rule before the second sample begins.
What would change the conclusion
The conclusion would change if routing time stayed the same after the rule and the misroute rate also stayed the same. That would suggest the problem is training, staffing, or the exception mix rather than the written rule. It would also change if the rule improved speed but increased misroutes. A rule that improves routing but not resolution may still be useful, but the claim should be limited.
A larger sample across several queues would show whether the pattern is local. A comparison with a queue that has no written rule would strengthen the claim. Until then, report the association and keep the rule under review. A comparison queue without a written rule would strengthen the finding. Report the local result even when the sample is small and keep the comparison window the same length.
Sources and methodology
A bounded review of exceptions handled by one virtual assistant role before and after a written escalation rule. Each sampled exception is coded for recognition time, send time, acknowledgement time, severity choice, and owner. The review reports local routing patterns and does not claim that the rule caused the change.
- NIST Cybersecurity Framework 2.0 current.
- U.S. Small Business Administration — Hire and manage employees current.
- U.S. Office of Personnel Management — Training and Development current.
- FTC — Data Breach Response: A Guide for Business current.
Source count: 4. Last verification date: November 7, 2026.
Questions about the findings
What is routing time?
The time from when the assistant recognizes an exception to when the correct owner acknowledges it. It is not the time to resolve the issue.
Does a written rule always speed up routing?
No. It can add steps or fail if it is hard to find. Measure both speed and accuracy before concluding that the rule helped.
How many exceptions are needed?
Enough to see the pattern across more than one exception type and several weeks. A small sample is a starting point, not a definitive result.
Put the finding to work
Review the full research library, compare cluster coverage inside Manager Enablement, and pair these findings with our VA candidate screening support.
