What share of borrower document rejections repeat the same reason?

A borrower document rejection is supposed to correct a problem. The borrower receives the message, fixes the issue, and uploads a new file. Sometimes the next review rejects the document for the same reason, and the borrower concludes that nobody read the first message. The repeat consumes time and erodes trust in the process.

This protocol measures how often a rejection repeats the same reason. It defines a repeat, offers a reason taxonomy, and explains how to sample without over-weighting the files with the most uploads. The result helps an operations team decide whether to fix the borrower request, the review rule, or the feedback message. It is not a measure of individual reviewer performance.

Question and scope

The question is what share of rejected borrower documents are rejected again for the same reason. The scope names the team, the document types, the channels, and the period. A team might study all borrower-uploaded documents during an eight-week window. The result applies to that setting and should be reported with it.

Define a rejection as a reviewer decision that a submitted document does not meet the condition requirement. A request for a different document is not a rejection. A document that is accepted with a note is not a rejection. The definition should be documented before the sample begins.

Define a repeat as a rejection of a later submission for the same condition with the same root reason. The later submission may be a new file or a corrected version of the same file. The root reason must match, not just the message text. A repeat can occur with a different reviewer and a different channel.

Why repeat rejections matter

A repeat rejection is a failure of the feedback loop. The borrower did not understand what to fix, or the reviewer applied a different standard. Either way, the file spends more time in the document stage and the borrower loses confidence. The cost is visible in cycle time and in the number of borrower questions.

Repeat rejections also distort capacity planning. A team that counts documents received may underestimate the review time because the same condition is reviewed more than once. The extra review is real work even if it does not appear as a new condition. A capacity plan that ignores repeats will be too optimistic.

Not every repeat is preventable. A borrower may send a new version that introduces a different defect, or a lender may change the requirement between submissions. The protocol should separate a true repeat from a new rejection. A credible rate does not blame the borrower for a moving target.

Defining the reason categories

Use a small taxonomy that covers the common causes. A practical set includes missing page, unreadable image, wrong date range, wrong account or property, expired document, name mismatch, and insufficient information. Add a category for other only when the reason does not fit. Too many categories produce inconsistent coding.

Write one sentence for each category. Missing page means the submission does not include a page required by the condition. Unreadable image means the required information cannot be read. Wrong date range means the document covers a different period than requested. The definitions should be visible during coding.

Pilot the taxonomy on ten rejections before the main sample. Ask two reviewers to code the same rejections and compare. If they disagree, rewrite the definition or add an example. A category that requires a debate will not be applied consistently during a busy week.

Sampling rejections rather than files

Sample rejections, not files. A file with five uploads and three rejections would dominate a file-based sample. Use the rejection log as the sampling frame and select events across reviewers, document types, and weeks. If the log is incomplete, note the gap before interpreting the result.

Use consecutive weeks to avoid seasonal bias. A closing week may produce more rejections than a quiet week. If the team cannot cover every day, use a rotating schedule that includes each weekday. Record the period and the sample size in the report.

Include rejections that were resolved without a second upload, such as a rejection followed by a phone call and a waiver. These events still count in the reason distribution, but they should not be counted as repeats unless a later submission was rejected again. The distinction keeps the numerator and denominator clear. A clear rule also prevents a waived condition from being counted as a repeat.

Coding the first and second rejection

For each rejection, record the condition number, the document type, the submission date, the reviewer, and the reason. For a repeat, record the earlier rejection and the later rejection side by side. The record should show whether the reason is the same or different. A note that says "rejected again" is not enough evidence.

Use the source document and the review note to confirm the reason. A reviewer's memory is weaker than the file history. If the review note is vague, code the reason as unknown rather than inferring from the outcome. A high unknown rate is a signal that the review notes need improvement.

Do not count a rejection as a repeat if the borrower submitted a different document for the same condition. The reason may be the same, but the event is a new attempt with new evidence. The protocol should distinguish a corrected file from a different file. This distinction affects the fix.

Measuring agreement

Have a second reviewer code a subsample without seeing the first coder's choice. Compare the results and report agreement by category. A low agreement on missing page may mean the condition request was ambiguous. A low agreement on unreadable image may mean the standard is unclear.

Resolve disagreements with a written rule. For example, decide whether a document with one unreadable page but all required information visible is a rejection or an acceptance with a note. Write the answer in the protocol and apply it to the rest of the sample. Apply the rule to the next ten events and check agreement again.

Do not change the definitions mid-sample without marking the change. A mid-study change can make the repeat rate move for reasons unrelated to the work. If the team must change a rule, start a new segment and report both segments separately. Report the first segment as the baseline.

Reporting the rate

Report the repeat rate as a range with the sample size, the period, and the scope. Show the median across weeks and the spread. A rate that moves from five percent to twenty percent may reflect a change in the review standard rather than a change in borrower behavior. A range is more honest than a single decimal.

Break the rate down by reason and by document type. A high repeat rate for missing pages may point to an incomplete request. A high repeat rate for name mismatch may point to an intake form or a review rule. The breakdown is more actionable than the total.

Pair the rate with the time between the first and second rejection. A repeat within one day may mean the borrower did not receive the message. A repeat after a week may mean the correction was misunderstood. The timing helps the team choose between a notification fix and a clarity fix.

What the result can change

A high repeat rate does not automatically justify more training. It may justify a better request template, a clearer rejection note, or a review checklist. The study should list the changes that could reduce repeats before it recommends additional capacity. The team should separate a communication problem from a knowledge problem.

The borrower communication support team can test one change on a small group and rerun the protocol. A rejection note with an example is easy to test because the team can compare repeats before and after. The test should run long enough to cover a normal document cycle. Keep the test group small enough that the rest of the team can continue the normal work.

If the repeat rate stays high after the change, the team may have a review consistency problem or a source document problem. The loan processing support lead can review the acceptance rules and the review sequence. The result should distinguish a communication defect from a workload problem. A measure that leads to the wrong fix is worse than no measure at all.

Limits and sources of error

The study measures recorded rejections, not every rejection. A reviewer who tells a borrower by phone and never logs the event is invisible. A borrower who resends without a review may not create a second rejection. The result is a lower bound on the true repeat rate in many teams.

Recall and attribution can bias the coding. A reviewer who remembers the file may assign a reason that fits the final outcome. Use the source document and the review note to check the reason. When the evidence is insufficient, code the event as unknown rather than guessing.

The sample may miss documents that were rejected by a different team or channel. A document sent to a loan officer instead of the processing queue may not appear in the log. Add a short end-of-day prompt that asks for unlogged rejections. The prompt improves coverage without turning the study into surveillance.

Review cadence

Repeat the study after a change to the request template, the rejection note, or the review checklist. Use the same categories and the same sampling rule so the comparison is meaningful. A rate that falls while borrower questions rise may mean the team is rejecting fewer documents but explaining less. A comparison across different definitions is not valid, so freeze the codebook before the second round.

Share the result with the people who recorded the events. Ask which reasons were hardest to code and which rejections felt avoidable. The discussion can reveal a request that nobody understands or a review step that applies two standards. The next study should test one of those findings.

End with a single sentence that states the repeat rate, the period, and the main reason the team can fix. A precise finding with a named owner is more useful than a long report. The goal is a borrower who corrects the document once and moves on. The sentence should name the change the team can make, not a cause outside its control.