Key takeaways
- Measure submission to approval, approval to account creation, and account creation to a successful test as separate stages.
- Count reminders and the final unblocker so a process gap is not mistaken for a technical fault.
- Test the account on the real task screen, not only at the login page.
- Change one request-form input and compare the stage medians after two weeks.
Question and scope
This study asks where waiting time accumulates between a new virtual assistant's access request and a working account. It separates the approval stage from the provisioning stage and the verification stage. The scope is one role, one access request log, and a fixed thirty-day window. The result describes that role and should not be treated as a universal service level.
A request is eligible when it names an account, a role, and a business reason. Requests for personal devices or unrelated software are excluded. A request is complete when the owner records a decision and the account works for the assigned task. The study follows each request until that point or until the window closes.
The study also records the number of reminders sent and the person who finally unblocked the request. A request that needed three reminders is a different process problem from one that needed a technical fix. The reminder count is part of the result, not a complaint. Report the counts by system and by approval owner.
A request that is submitted before the hire starts is recorded with the planned start date as context. The clock still begins at submission, but the report separates pre-start waiting from in-role waiting. A request that waited for a start date is not an approval defect. The separation keeps the result fair to the approver.
| Stage | Evidence | Repair |
|---|---|---|
| Request created | Complete form timestamp | Start the clock here |
| Approval | Decision date and owner | Name one approver |
| Provisioning | Account creation record | Fix the group template |
| Verification | Successful task test | Test before live work |
Definitions and evidence standard
Define the request start as the time the requester submitted a complete form, not the time someone first mentioned the need. A verbal request that was never written down has no reliable start. Define the approval end as the time the owner recorded a decision. Define the provisioning end as the time the account existed.
Use the request log, the identity system, and the first successful sign-in as evidence. A message that says the account should work is not proof that it does. The study should record the test action and the screen it reached. A login that cannot open the needed screen is still a provisioning failure.
Separate a missing approval from a slow approval. A missing approval has no owner or no decision, while a slow approval has an owner and a date that arrived late. Separate those conditions from a provisioning error caused by a wrong group or a duplicate account. Each condition points to a different repair.
Use one time source for every timestamp so the stages can be compared. A request log in one system and an identity record in another may use different clocks or zones. State which source is authoritative for each stage. A mixed clock can create a delay that never happened.
Sampling and measurement
Collect every eligible request in the window and freeze the list before coding. Include more than one system and more than one approver so the result is not one person's habit. If a system migration occurred, report it separately. Record the number of requests, systems, and approvers.
Measure the time from submission to approval, approval to account creation, and account creation to a successful test. Add the number of reminders and the number of handoffs between people. Report the median and the longest case rather than only an average. The longest case often reveals the failure mode that matters most.
If a request was delayed for a reason unrelated to access, such as a hiring date change, record it separately. A request that arrived before the start date is not a delay. Report those cases beside the access results. The separation keeps the waiting time attributable.
Have two reviewers code the first ten requests and compare the stage boundaries. Discuss every disagreement and write the agreed rule in one sentence. Repeat the comparison at the midpoint of the window. A short calibration note makes the final medians easier to defend.
Findings and interpretation
Present the stage totals before naming a cause. A long approval stage may mean the owner is unclear or the form asks for evidence nobody collects. A long provisioning stage may mean the request reaches the identity team without the right group. Read the actual requests behind the largest stage before suggesting a repair.
Compare the waiting time with the number of systems requested in the same week. A new hire who needs six accounts may wait longer than one who needs two because each approval is separate. Report the system count with the result. The comparison keeps a volume effect from being read as an owner problem.
Check whether the assistant could do any useful work while waiting. A role with a reading task and a shadowing schedule may be productive before every account is live. A role with no safe work will sit idle. Report the available pre-access work with the delay. The combination shows the real cost.
Compare the stage totals by system before combining them into one average. One system may require a manual group change while another provisions automatically. Report the systems separately when the difference is large. The split shows where a technical fix would have the most effect.
What the record shows
The record should show the request as the requester experienced it, including the reminders and the silence. A log that only stores the final approval date hides the effort spent chasing it. Report the reminder count and the longest silence beside the stage time. The pair shows whether the delay was waiting or neglect.
A missing owner is a different defect from a slow owner. A missing owner needs a name in the form, while a slow owner may need a backup or a shorter review window. Code the two conditions separately before recommending a change. The repair should match the condition.
The record also shows whether the assistant had safe work while waiting. A role with a reading task and a shadowing schedule can use the time, while a role with no access cannot. Report the available work with the delay. That context turns a raw number into a decision about the start date.
Limitations
The study measures recorded requests, not every access need. An assistant may borrow a shared login and never create a request. The result therefore describes the formal path, not all access in use. It also cannot show whether the borrowed access created a security risk.
The request log may be updated after the fact, which can make a slow approval look faster. Freeze the log at the end of the window and compare it with the identity system. The sample may also be small in a role with few systems. State both limits beside the counts.
Deciding whether a delay belongs to approval or provisioning requires judgment when the two overlap. Two reviewers may code the same request differently. Record the disagreements and resolve them with examples before reporting. A calibration set of ten requests improves consistency.
Practical next step
Choose the stage with the largest correctable delay and change one input to that stage. If approval is slow, add the evidence the owner needs to the request form. If provisioning is slow, add the exact group and license to the request. Test the change on the next ten requests.
Add a field that records the reminder count and the final unblocker. The field can be a short list such as owner, identity team, manager, or vendor. A small list is faster than free text and easier to count. The note should take less than a minute to complete.
Review the change after two weeks and compare the stage medians with the baseline. Keep the revision if it reduces waiting time without weakening the approval standard. If the pattern does not move, the cause may be the start date rather than the request process. Read the test with the people who approve and provision access.
A short follow-up record should note whether the same approval owner was slow in the next request. A repeat delay suggests that the role is overloaded rather than careless. Record the date and the request beside the new entry. The note turns a single fix into a test of whether it held.
Sources and methodology
The study follows eligible access requests through one role and one request log for thirty days, coding the approval date, the account creation date, the first successful task test, the reminder count, and the person who unblocked each request.
- NIST Special Publication 800-63B - Digital identity guidelines current.
- NIST Special Publication 800-46 Rev. 2 - Enterprise telework and remote access current.
- U.S. Cybersecurity and Infrastructure Security Agency - Require multifactor authentication current.
- U.S. Small Business Administration - Hire and manage employees current.
Source count: 4. Last verification date: November 16, 2026.
Questions about the findings
Should the clock start when the need is first mentioned?
Use the complete request as the start. A mention without the required evidence cannot be approved and would make the waiting time look longer than the process allows.
Is a login enough to prove access?
No. The assistant should reach the screen or action the role needs. A login that opens an empty portal may still be a provisioning failure.
How should shared logins be counted?
Record them as an exception rather than a normal request. A shared login has no named owner and cannot be revoked cleanly for one person.
What if the delay is caused by the start date?
Report it separately. A request that arrives before the start date is a scheduling choice, not an approval delay.
Which measure is most useful?
The median time for each stage plus the longest case. The medians show the common path while the longest case shows where the process breaks.
Put the finding to work
Review the full research library, compare cluster coverage inside Onboarding And Retention Research, and pair these findings with our VA candidate screening support.
