Research question and scope
This protocol measures how many contractor accounts remain active after the project or assignment that justified them has ended. The unit is one account in one client system. An active-after-end account has a working login or permission after the recorded end date. The measure describes access hygiene, not whether the person did anything wrong.
Set the scope to one provider, one client, and a fixed period. Record the systems in scope and the source of the end date. If the scope or the end-date source changes during the period, split the sample at the effective date. A different source creates a different denominator.
Keep the measure separate from an incident count. An active account may never be misused, and an incident may involve an account that was properly closed. The purpose is to test the offboarding control. Report incidents and access findings separately.
Name the decision the result will support before collection starts. The client may use it to add an offboarding trigger, shorten the removal window, or require a weekly reconciliation. A report that only lists stale accounts leaves the process unchanged. Write the intended action beside the scope.
Record the account types included, such as interactive, shared, service, and vendor admin. A service account can remain active for a legitimate integration after a person leaves. Note the type and the reason. Mixing types hides where the control actually fails.
Population and unit of analysis
Define the population as every contractor account that was active at any point during the period. Include accounts created by the provider and accounts created by the client for provider staff. The unit is the account, not the person or the project. One person with three accounts is three units.
Include accounts that were disabled during the period and accounts that remain active. Mark the removal date and the method. Exclude accounts that never had access to client data, such as a training sandbox. Report the excluded count and the reason.
Exclude a shared account from the person-level count and analyze it separately. A shared queue login needs an owner and a credential rotation rather than a departure date. State the rule before sampling so a second reviewer applies it the same way. The shared-account group often hides the oldest access in the environment.
Decide how to treat an account with a planned extension. Include it and mark the extension with its approval reference. A planned extension is not a stale account, but an extension without a date is. The approval is the evidence that separates the two.
Decide whether a disabled account that can be re-enabled by a group membership counts as active. Count it as active when the person can regain access without a new approval. Record the group and the path. A soft disable is not a removal.
Definitions that make the measure reproducible
Define the project end date as the date in the statement of work, the assignment record, or the client's written notice. Define the removal deadline as the number of business days the agreement allows after that date. Define an active-after-end account as one that still works after the deadline. Publish one example of each.
Define active access by a successful login or a successful permission check after the end date. A disabled flag in one system is not proof when a group or a token still grants access. Record the test and the result. The definition should describe the effect, not the label.
Define a planned extension as a written approval with a new end date and a named approver. A verbal extension or an open ticket is not enough. Record the reference. The extension rule protects legitimate work without weakening the control.
Define the removal evidence as a system confirmation plus a user-side test. A ticket that says removed is weaker than a login attempt that fails. Record both where available. The evidence standard should be the same for every system.
Agree how to classify an account whose owner has left the provider but whose client contact still uses it. Count it as active and mark the ownership gap. Record the client contact and the last review. The account is a shared risk even when the original owner is gone.
Sampling and data collection
Build the sample from the account inventory, the project roster, and the client's system export. Record the export date, the systems, and the row count. Do not build the sample from a list of known departures. A departure-based sample cannot show the accounts nobody remembered.
Capture the account, the person or owner, the system, the role, the project, the end date, the removal date, the removal method, and the last login. Add the extension reference where one exists. Store the exports with the report. The end date and the access test are the core evidence.
Check a small overlap sample with a second reviewer. If the reviewers disagree about whether a soft disable counts as removal, the definition is too vague. Rewrite the rule and retest before the full sample. Agreement matters more than sample size.
Record the time spent matching the roster to the inventory. A report that requires opening every account will not be repeated. Note the systems that cannot export a last-login date. The collection cost belongs beside the result.
Test a small number of accounts from the user side with the client's permission. A failed login is stronger evidence than a status field. Record the test time and the result. Do not test in a way that locks out an active worker.
Measuring active-after-end accounts
Report the share of accounts removed by the deadline, removed late, still active, and covered by an approved extension. Show the counts behind each share. The four groups describe different levels of control and should not be collapsed. A single stale rate hides the late removal that still created exposure.
Break the result down by system, account type, and removal method. A system without an offboarding trigger may show more stale accounts. A shared account may need a different control. Compare only groups with a similar access level.
Report the median days between the project end and the removal, including late removals. A median within the deadline is different from a median that relies on a few fast cases. Include the longest active account and its age. A mean would let one very old account distort the result.
Report the share of stale accounts that had a login after the end date. A stale account with no use is a latent risk, while one with activity is a live finding. Record the last login and the action taken. The activity changes the urgency.
Compare the stale rate for accounts with a named offboarding owner against accounts without one. A large gap suggests the owner field matters more than the removal tool. Record both counts. The comparison points to a fix that does not require new software.
Limitations and next action
The measure cannot show whether an account was used improperly or whether the person intended to keep access. It also cannot show an account created outside the inventory. Treat the result as a measure of recorded access state. Note the systems and account types that were not included.
Give the result to the provider manager and the client security owner with one change to test. The change might be an automatic end-date trigger, a weekly exception report, or a required removal test. Record the effective date and repeat the measure with the same definitions. A controlled change is easier to judge.
Review the boundary cases before shortening every deadline. A few approved extensions may explain a large share of active accounts. Fix the extension record first if that is the pattern. The goal is access that ends with the work, not a lower number by definition.
Repeat the measure after the change with the same systems and account types. Report the before and after counts together, including the inventory size. Record any new system or project in the period. A comparison with a changed scope is not a comparison.
