Research question and scope

This protocol measures how often a password reset request reaches a delegated service desk without a recorded identity check. The unit is one reset request tied to one account. A missing check means the record does not show which identifiers were verified or who confirmed them. The measure describes the control around the request, not the intent of the specialist.

Set the scope to one client, one identity system, and a fixed period. Record the verification methods the client approved and the queue that received the requests. If the method changes during the period, split the sample at the effective date. A different verification rule creates a different denominator.

Keep the measure separate from an incident count. A request can be handled correctly without any account compromise. The purpose is to find where the record is silent before an incident occurs. A neutral framing helps the client share the admin export.

Population and unit of analysis

Define the population as every reset request that entered the queue during the period. Include requests that were completed, refused, and abandoned. The unit is the request, not the account or the person. A person who requests three resets in a week is three units.

Include requests received by phone, chat, and email, and record the channel on each one. Exclude self-service resets that never reached the desk. Report the excluded volume so the reader can see the desk's share. A rate built only from completed resets will miss the refused cases.

Exclude a request that was a duplicate of an earlier request within the same session. Link it to the original and count it once. Duplicate handling is a separate efficiency measure. State the exclusion count beside the denominator. Record the reason for each exclusion.

Definitions that make the measure reproducible

Define an identity check as at least two account-based identifiers or one approved callback that is recorded on the request. Define a partial check as one identifier or an unrecorded callback. Define a missing check as no verification evidence at all. Publish one example of each outcome before sampling.

Define a privileged request as any reset for an account on the client's privileged list. Define a shared request as one for an account without a single holder. These categories need separate reporting because the approval path differs. Do not merge them into the main rate.

Decide how to treat a request verified in a chat and later pasted into the ticket. Count it as recorded only when the ticket shows the identifiers and the confirming person. A pasted note that says verified without detail is partial. Record the rule so two reviewers agree.

Sampling and data collection

Draw the sample from the service queue and the identity admin log for the full period. Record the export date, the filters, and the row count. Do not build the sample from incident reports. A sample built from incidents cannot produce a request rate.

Capture the request identifier, the account, the channel, the arrival time, the verification method, the confirming person, and the close time. Add the admin tool reference where the platform supports it. Store the export with the report. The verification field and the timestamp are the core evidence.

Check a small overlap sample with a second reviewer. If the reviewers disagree about whether a check is partial, the definition is too vague. Rewrite the rule and retest before the full sample. Agreement matters more than sample size for a measure that will be reused.

Record the time spent matching queue rows to the admin log. A report that requires manual matching for every request will not be repeated. Note the fields the platform does not expose. The collection cost belongs beside the result.

Measuring missing identity checks

Report the share of requests with a complete check, the share with a partial check, and the share with no check. Show the counts behind each share. The three groups describe different levels of control and should not be collapsed. A single missing-check rate hides the difference between a weak record and no record.

Break the result down by channel, system, and shift. A phone channel may show more partial checks than a written form. A night shift may show a different pattern because the approver is unavailable. Compare only groups with similar volume and method.

Report the share of privileged requests that received a second approver. Include the approval time and the approver role. A privileged reset with a complete identity check but no second approval is still a control failure. Keep it separate from the main rate.

Report the age of each missing check at the time of review. A gap found the same day is different from one found at day thirty. Include the review date and the queue's stated review cadence. The delay distribution shows whether the problem sits in the request or the audit.

Limitations and next action

The measure cannot show whether a reset was used to take over an account. It also cannot show a check performed outside the recorded channel. Treat the result as a measure of the request record, not of security outcomes. Note the systems and channels that were not included.

Give the result to the service owner and the identity owner with one change to test. The change might be a required verification field, an approved callback script, or a second approver for privileged accounts. Record the effective date and repeat the measure with the same definitions. A controlled change is easier to judge.

Review the boundary cases before widening the rule. A few shared accounts with approved owner callbacks may explain a large share of partial checks. Fix the classification first if that is the pattern. The goal is a record that proves the check, not a lower number by definition.