This study asks why work is returned after review and which causes are actionable. Define a return as a reviewer sending an item back before acceptance, then sample ordinary and high-consequence work from one lane. Do not treat every return as an error: a new fact or prudent escalation may correctly reopen an item.
For each case, retain the instruction version, submitted evidence, reviewer comment, return category, correction, second disposition, and owner decision. Have a second reviewer code a subset without seeing the first category. Disagreement can reveal an ambiguous rubric rather than poor contributor performance.
GAO and NIST offer control and governance concepts; the ILO provides context on decent conditions in home-based work. They help frame responsibility and fair review but do not prove productivity, worker quality, cost savings, or causation in a particular outsourced team.
Limitations include inconsistent reviewer notes, silent corrections that never appear as returns, and differences in task difficulty. A small convenience sample may miss rare failures. Report category definitions and missing records, and keep personnel judgments outside a process study not designed to support them.