Key takeaways
- Measure from approval, not from the new hire's first question.
- Split waiting time by stage and owner.
- Do not treat a created account as verified access.
Question and scope
This study asks how long it takes for an approved new-hire access request to become verified access. It measures the interval between the approval event and the first successful sign-in, then splits that interval into request, provisioning, and verification stages. The result is intended to help an onboarding team find where the wait accumulates. The scope is a defined set of new hires and systems during a fixed period. The study does not measure the security quality of the access or the productivity of the new hire. A short provisioning time does not prove that the right permissions were granted, and a long time does not prove that the request was mishandled.
| Stage | Start | End | What to record |
|---|---|---|---|
| Request | Manager submits | Approval recorded | Approver and scope |
| Provisioning | Approval recorded | Account created | System and owner |
| Verification | Account created | First successful sign-in | Time and method |
Population and unit
The population is the set of access requests tied to new hires who started during the window. Include requests for the systems required in the first week, and record any request that was withdrawn or replaced. Exclude shared mailboxes and service accounts unless the protocol treats them separately. The unit is one request for one new hire and one system. If a request bundles several systems, either split it into separate units or define a bundle rule and report the bundle size. A bundle that mixes a simple account with a complex role can hide the stage that actually caused the delay.
Timestamps and definitions
Define each timestamp from a system record rather than a recollection. The approval time should come from the approved request, the provisioning time from the identity system, and the verification time from the first successful sign-in or the documented test. If a timestamp is unavailable, mark it missing rather than estimating it from an email. Distinguish calendar time from business time. A request submitted on Friday may look slow on Monday even when the team responded within the service window. Report the rule used, the time zone, and any holiday or outage that affected the interval.
Sampling
Draw the sample across systems, roles, locations, and start dates. Include requests approved before the start date and requests approved after the start date. The second group often reveals a different failure path, because the new hire may already be waiting. Record the selection method and the frame version. If the sample is small, report the individual intervals rather than only an average. A handful of very long requests can make the mean misleading, so show the median and the range as well.
Measures
Report the median and range for total lag, then the median and range for each stage. Show the number of requests with a missing timestamp and the number that could not be verified. A request without a verification event is not evidence of fast access. Segment by system only when the subgroup has enough observations. A single slow system may be a configuration problem, while a pattern across systems may point to an approval bottleneck. Report absolute counts beside every rate so the reader can judge the size of the pattern.
Interpretation
A long approval stage points to a decision owner who is unavailable, a request that lacks information, or a rule that requires unnecessary review. A long provisioning stage points to a queue, an integration, or an account setup that needs manual work. A long verification stage points to a communication gap or a sign-in problem. Do not attribute the delay to the new hire without checking the instruction and the recovery path. A verification step that depends on a message the new hire never received is a process failure. The study should locate the stage and the owner before assigning a cause.
Limitations
The study observes requests that reached the tracking system and may miss access arranged informally. It cannot prove that the granted permissions matched the task, only that the account was created and used. A later access review is needed for the permission question. System outages, start-date changes, and role changes can affect the intervals. Note those events and avoid comparing periods with different operating conditions. A small sample may show a useful pattern without supporting a precise benchmark.
Privacy and governance
Use study keys and avoid including unnecessary employee information in the analysis. Keep the crosswalk restricted and follow the retention rule. If the analysis includes system names or role details, treat them as internal operational information rather than public content. The onboarding team may measure the process, but the access owner decides whether a permission is appropriate. The study should not be used to rank individual technicians or new hires. Its purpose is to improve the request, approval, and verification path.
Operational follow-up
Choose one stage and one change to test. A pre-approved role template may shorten approval, a daily provisioning queue may shorten setup, and a pre-start verification message may shorten the final stage. Name the owner and the date of the next measurement. Recheck the same intervals after the change and report both the total and the stage result. If the total improves but one stage gets worse, the change may be shifting work rather than removing it. Keep the decision log with the access standard so the next reviewer can see what was tested.
Worked example: a request split across two systems
A manager requests a case queue and a document repository for a new hire. The queue is approved on Monday, but the repository request is still waiting for a data-owner approval. The new hire can sign in to the queue on Tuesday and cannot open the repository until Thursday. The study records two units because the systems have different owners and approval paths. If the request had been treated as one bundle, the total lag would have hidden the repository approval as the real constraint. The example supports a simple rule: split by system when the approval path differs, and report the bundle only when the systems share the same owner and rule.
Stage ownership
Name the owner for each stage before measuring it. The manager or hiring owner owns the request, the access approver owns the decision, the identity or IT team owns provisioning, and the new hire or coordinator owns the verification. A stage without an owner will look like a delay with no fix. The owner of the whole process should be able to see the stage result and ask the right person for a change. If the process owner is also the person doing every stage, the study may reveal a capacity problem rather than a handoff problem. Report that possibility.
Exception and outage handling
Record system outages, approval holds, and start-date changes as separate events. An outage that adds two days to provisioning is different from a request that sat in the wrong queue. The study should be able to exclude an outage from a process comparison or report it as a limitation. A request that was cancelled or replaced should remain in the frame with its outcome. Removing it from the denominator can make the process look faster than it was. A short reason code for cancellation or replacement is enough.
Reporting template
Report the frame size, the number of requests, the number of systems, and the number of new hires. Show the median and range for total lag and for each stage, with the count behind each result. Include the number of requests with missing timestamps and the number that could not be verified. Add a short note on the systems, locations, and start-date range in the sample. If a subgroup is too small, present the individual intervals rather than a rate. End with one change, one owner, and a date for the next measurement.
Review cadence
Measure after a process change, before a peak hiring period, and at a regular interval that matches the team's capacity. A monthly review may be useful for a high-volume team, while a quarterly review may be enough for a small one. The cadence should be written down and tied to a named owner. Keep the raw request list with the study so a later reviewer can repeat the calculation. If the identity system changes, record the date and avoid comparing the new intervals directly with the old ones without a note.
Sources and methodology
A bounded operational measurement of access requests with a frozen request set, defined timestamps, and stage-level reporting. The method describes waiting time and does not estimate the employer's security posture.
- NIST Cybersecurity Framework 2.0 current.
- NIST SP 800-207 — Zero Trust Architecture current.
- Cybersecurity and Infrastructure Security Agency current.
- U.S. Department of Labor — New Hire Reporting current.
- U.S. Small Business Administration current.
- U.S. Equal Employment Opportunity Commission current.
Source count: 6. Last verification date: November 3, 2026.
Questions about the findings
When does the clock start?
At the recorded approval event, not at the new hire's first question or the manager's informal request.
Is a created account the same as access?
No. The study treats first verified sign-in as the endpoint.
Put the finding to work
Review the full research library, compare cluster coverage inside IT Onboarding, and pair these findings with our VA candidate screening support.
