Research question and scope

This protocol asks whether a shift-swap record can support a payroll decision without a reviewer reconstructing the story from chat. The unit is one approved swap that affects a scheduled shift inside a defined observation window. The study examines the record, the roster, and the payroll input; it does not evaluate the worker, the supervisor, or the fairness of the underlying schedule. A complete record may still accompany a poor decision, and an incomplete record may accompany a sound one.

Set the scope to one site, one client, or one scheduling team and record the roster system, payroll system. Approval workflow version in force at the start. If the team changes its swap template during the window, keep the earlier definition for the full sample or analyze the change as a separate period. Mixing definitions makes a process change look like a change in record quality.

Population and unit of analysis

Define the population as all swaps that were approved, rejected, or left pending at the payroll cutoff during the window. The unit is the swap event, identified by the original shift, the requested replacement, and the approval timestamp. A swap that is reversed and reapproved remains one event with a revision history. A genuinely new request for a different shift is a separate unit.

Record the reason category, the hours affected, whether the swap crossed a site or cost center. Whether it created overtime or a rest-day conflict. These strata matter because a swap between two qualified workers on the same team carries a different risk from a swap that moves hours into a premium category. Test the unit definition on two ordinary cases before collection begins.

Definitions that make the measure reproducible

Define an approved swap as one where a named supervisor with authority confirms the replacement and the effective roster version. Define a payroll match as the condition where the paid hours, cost center, and premium treatment agree with that approved roster. Define a complete packet as one containing the request, the approval, the roster comparison, and the consequence check. Each definition should include one positive example and one boundary example.

Do not treat a message saying "approved" as equivalent to an approval if it does not identify the shift and the roster version. Do not treat a payroll match as proof that the approval was proper. It only shows that the two systems agree. Separate evidence completeness from decision quality so the analysis does not overclaim what a record can prove.

Evidence standard and data dictionary

For each swap, capture the request source, the original assignment, the replacement, the affected hours, the approver, the approval time, and the roster version. Capture the consequence checks for overtime, rest-period conflict, credential validity, and coverage elsewhere. Code each field as present and usable, present but empty, missing, or not applicable. Record where the evidence lives.

Add a field for the payroll disposition: matched, corrected before cutoff, corrected after cutoff, or unresolved. Also record whether the reviewer had to contact a supervisor to interpret the packet. That contact is a useful signal because it shows where the record failed to carry the decision on its own. A second coder should be able to classify the same packet without asking the author what was meant.

Sampling and collection

Draw the sample from the scheduling and payroll exports rather than from a list of memorable swaps. Record the export date, filters, row count, and the random or systematic start rule. Include approved, rejected, and pending swaps, because rejected requests can reveal whether the reason was recorded or simply disappeared. If the population is ordered by shift date, a fixed interval spreads the sample across the window.

Collect the packet from the systems that existed at the time of the decision. Do not rebuild missing evidence later, because reconstruction measures the investigator rather than the original process. If an attachment expired, record the expiry and preserve the link metadata. Note the collection date separately from the shift date so the reader can see how much of the record has aged.

Analysis and reporting

Report the number of swaps by status, the share with a usable approval, the share with a complete consequence check, and the share that matched payroll on the first attempt. Present missing, corrected, and unresolved cases separately. Show the numerator and denominator for every rate. If a headline is needed, choose the field most likely to change a payroll decision and report it beside overall completeness.

Compare swaps within the same risk stratum rather than ranking supervisors by speed. Inspect the cases that required a follow-up contact and the cases corrected after the cutoff, because those are the events that consume review capacity. Show the age distribution of pending swaps rather than only an average. A small tail of old, unresolved cases can matter more than a moderate average.

Limitations and sources of error

The record cannot show whether the supervisor considered the worker's preference, the fairness of the schedule, or the client's real need. It can show only what was documented and how the payroll input compares. A complete packet is not proof of a lawful or fair decision, and the report should say so plainly. Local rules and contracts may impose requirements this protocol does not interpret.

Exports can omit deleted messages, earlier roster versions, and approvals given in person. A swap that was approved before the window but paid inside it may belong to a different workflow. Record those cases and their treatment instead of dropping them silently. If the reviewer and approver are the same person, note the independence limitation and analyze that subgroup separately.

Review cadence and next action

Give the result to the scheduling and payroll owners and ask them to choose one observable change: add the roster version to the approval, require a consequence check before approval, or move the approval deadline earlier. Record the owner, the effective date. The measure that will show whether the packet improved. Avoid changing three things at once if the team needs to learn which change worked.

Repeat the protocol with the same definitions after the change and compare the same fields. If the result did not move, inspect whether the template was ignored, the approval path was bypassed, or the underlying roster made the evidence hard to capture. Close the review when the owner accepts the finding and the next check is scheduled. The same review should preserve the original finding so a later reader can see why the protocol was repeated.

Sources

- [U.S. Department of Labor, Wage and Hour Division](https://www.dol.gov/agencies/whd) — primary guidance on hours, overtime, and recordkeeping.

- [Occupational Safety and Health Administration, recordkeeping](https://www.osha.gov/recordkeeping) — incident and exposure record requirements.

- [U.S. Small Business Administration, manage your business](https://www.sba.gov/business-guide/manage-your-business) — operating and workforce planning guidance.

- [NIST Cybersecurity Framework 2.0](https://www.nist.gov/cyberframework) — governance, access, and recovery language used in the access checks.

What the result can change

The result can change the template, the approval deadline, or the point at which a reviewer checks the roster version. It cannot decide whether a particular swap was fair, whether the worker was pressured, or whether the schedule complied with every local requirement. Keep those questions with the appropriate owner and use the protocol only for the evidence trail.

If the finding shows that most errors are corrected after payroll, the first change should be earlier verification rather than a longer form. If the finding shows that approvals are missing, the fix may be a clearer authority list. Match the recommendation to the failure mode the data actually shows.

A worked reporting example

Suppose the sample contains forty swaps, of which thirty have a usable approval and twenty-eight match payroll on the first attempt. Report those counts with the observation window and the strata, then describe the twelve cases that needed a correction. Do not turn the counts into a claim about all swaps at the site or about the quality of the supervisor.

The useful next question is why the twelve cases differed. Were they concentrated in one shift type, one approver, or one roster version? A short cross-tab gives the owner a place to look without pretending that the sample proves a cause. The report should end with one change, one owner, and one date for the next review.

Reporting the limitation beside the result

Every report should state that the sample describes the selected shifts and does not prove that all swaps are handled the same way. It should also name the systems that could not be checked and the cases that were excluded. A reader who understands the boundary is less likely to turn a local finding into a broad claim about the provider or the client.

If the sample is too small to support a rate, report counts and describe the cases instead. The owner can still act on a clear pattern, such as repeated missing roster versions, without pretending that the pattern is universal. The next review can expand the sample if the decision requires more precision.