What buyers need from bulk processing
- Rows should not disappear just because one data source fails or one property has weak coverage.
- Partial results should still be useful when enough evidence exists to write the row safely.
- Refund logic should be explicit when a property does not meet the platform's minimum quality threshold.
How the workflow should feel
- Good rows export cleanly with all populated fields that were available.
- Weak rows are visibly flagged so the acquisition or servicing team can decide whether to ignore, rerun, or escalate them.
- Refund-related behavior is deterministic and explainable rather than buried in a generic error.
Why this matters
- Budget trust: users can reason about what was charged and why.
- Operations trust: downstream files stay usable even when some properties are imperfect.
- Support trust: when a user asks “what happened?”, the answer is visible in the exported outcome.