Research · Published · Updated
Auditing SOP Coverage Against the Work That Actually Arrives
A repeatable protocol for comparing a procedure library with the request log, so gaps are found before a new assistant hits them.
On this page
Headline signal
A coverage audit protocol for local teams, not an industry completeness standard. (https://csrc.nist.gov/pubs/sp/800/53/r5/upd1/final). Treat this as context for planning, not a guarantee about a particular assistant or operating result.
Analysis and operating context
Teams often judge their documentation by how many pages it contains. A more useful question is whether a new assistant could complete the work that actually arrives. This protocol compares the request log with the procedure library and produces a ranked list of gaps. It does not attempt to document everything, and it does not treat a long library as evidence of quality.
Begin with a fixed export of completed and returned tasks from a defined period. Group them by the action a person had to perform, not by the client who sent them. Two clients with different names may need the same calendar procedure, while one client may generate five distinct procedures because of local quirks. Record the grouping rule and keep the raw task identifiers so the grouping can be challenged.
Build a coverage matrix with request groups as rows and procedure attributes as columns: exists, current, complete, findable, and approved. Exists means a document is present. Current means it reflects the tool and policy in force during the observation period. Complete means it covers prerequisites, the steps, the stopping point, and the evidence to keep. Findable means a person can reach it from the place where work begins. Approved means a named owner has accepted it.
Score each cell with evidence rather than opinion. For findability, test by asking someone unfamiliar with the group to locate the procedure within two minutes using only the links available at intake. For completeness, use a real task and check whether following the document would have produced the accepted result. Record the specific missing element. A binary score with a note is more useful than a five-point scale with no explanation.
Select a sample of request groups for deep review. Include high-volume groups, groups that touch money or access, groups with a high return rate, and groups that are newly created. Oversample the risky and unstable groups for learning, but report weighted and unweighted results separately. Do not let a convenient sample of simple tasks stand in for the whole library.
Interview the assistant who does the work, using the task records rather than memory alone. Ask which step required a question, which document was out of date, and which instruction was ignored because it did not match reality. These accounts explain why a procedure that looks complete still fails. Keep the interview notes separate from the audit score and do not attribute a systemic gap to an individual.
Rank the gaps by consequence and frequency. A missing procedure for a monthly report may be inconvenient; a missing stopping point for an access request may be a control failure. Use a short scale that combines how often the group appears with what happens when the procedure is wrong. The ranking should produce a work plan, not a compliance certificate.
Report coverage as counts and rates with the denominator visible. For example, twelve of eighteen groups have a current procedure, and seven of those twelve include a stopping point. Avoid a single percentage that hides which attribute failed. Include the excluded groups and the reason for exclusion so a reader can see what the audit did not cover.
Repeat the audit after a tool change, a client onboarding, or a staffing transition, because those events invalidate procedures faster than normal drift. Keep the prior matrix and the change note. A later comparison is only meaningful if the definitions stayed stable or the change is disclosed.
The limits are familiar. A procedure can be complete and still be wrong. An assistant may succeed despite poor documentation because of experience that is not written down. The request log may miss work that was handled informally. The audit is a structured way to find the most consequential gaps first, not a guarantee that the library is sufficient for every future request.
Apply this to a research support lane
Use a project coordination lane to keep the request log, procedure gaps, and review record together. Your operations owner still decides which procedure changes, approvals, or access rules to adopt.
Review project coordination supportSources
Frequently asked questions
How many procedures should a team have?
There is no universal number. The right count covers the recurring work with enough detail to act safely and stops before the library becomes unmaintainable.
Who should own the audit?
An operations owner who can change the intake form, the procedure, or the approval path, with an assistant participating as the person closest to the work.