Evidence desk / Pipeline quality

What share of loan files have a complete income calculation before review?

A file completeness protocol for income calculations that keeps the worksheet and the conflict cases visible.

Research brief cover for “What share of loan files have a complete income calculation before review?”

Key finding

Completeness must be judged against a written worksheet, and the short-notice files should be reported separately rather than blended into the rate.

Question and scope

This review examines income calculation completeness as a records question rather than a sales or credit question. It asks whether a second person could follow the file from the source document to the recorded decision. The unit, population, and window are fixed below so the result cannot drift while the sample is collected. The closing section states the limits before anyone repeats the number in a meeting.

The scope here is deliberately small: one mortgage workflow, one population, and one window of observation. The Mortgage Assistant wants to know whether the income calculation completeness file shows its evidence, its reviewer, and its decision date. The brief will not score loan quality, predict fallout, or compare one lender against another. It offers a repeatable way to inspect the record and a clear statement of what the record cannot show.

This brief defines one mortgage measurement question: how income calculation completeness behaves inside a single lending team. The answer is a local reading of evidence, not a benchmark, a compliance opinion, or a prediction for another lender. Everything below covers one named population and one fixed window, and the report repeats both beside every number. An operations lead with the same export can rerun the procedure without asking the author for help.

Population and unit

The unit is one loan file with a self-employed or variable income source, counted once under a written identity rule. Name the opening event, the closing proof, and the condition that pauses a file before the first record is pulled. If the pipeline changes size mid-window, a bare total stops being a measurement. List every removal with its reason so another analyst can rebuild the same list.

Count one one loan file with a self-employed or variable income source once, and write the identifier rule beside the extract. The window is wide enough to catch a normal month and at least one heavy week. Note system downtime, staff turnover, and investor changes as context instead of dropping those days. When the rule changes, keep both versions and mark which files belong to each.

Definitions and measures

The Mortgage Assistant keeps a short field guide that lists each field, its permitted values, and the document that proves it. Blank, unavailable, and confirmed-negative answers go into separate columns. The headline number is the share of files with a complete calculation worksheet before the reviewer opens them, and it never appears without the denominator and the window. A percentage alone cannot tell a team what to fix.

Define a finished income calculation completeness file before the export is opened. The definition names the required documents, the accepted source, and the treatment of a blank field. Report the share of files with a complete calculation worksheet before the reviewer opens them with the eligible count and the exact dates of the window. Beside it, show the missing-document count and the age of the oldest open file.

For this review, a file counts as complete when it supports the share of files with a complete calculation worksheet before the reviewer opens them with a dated source and a named reviewer. The guide also settles how a reopened file and a duplicate file are counted. The Mortgage Assistant prints the sample count, the unresolved total, and the oldest open item under each result. Those three figures keep volume, evidence, and delay from hiding one another.

Evidence standard

A file passes only when and sit in the approved system under the same loan reference. If two records conflict, keep both and write down which one drove the decision. The Mortgage Assistant tags every estimated figure as an estimate so a later reader cannot mistake it for a verified value. The goal is a trail another person can follow, not a flawless folder.

Settle the evidence test for income calculation completeness before the first file is chosen. The test names the document that counts, the date attached to it, and the rule for a blank value. The Mortgage Assistant attaches a one-page guide to the extract so the standard does not live only in one reviewer's head. The guide is part of the method and is published with the finding.

Sampling and collection

Pick a random or systematic start, then take consecutive files inside each stratum. Stratify by stage, owner, and exception type so the easy files do not dominate the result. Include one hard case: a missing document, a conflict, an absent owner, or a deadline that forced a quick call. The Mortgage Assistant logs every access failure instead of quietly swapping in a different file.

The extract begins with the pipeline export for income calculation completeness, not a set chosen after the result is known. The Mortgage Assistant keeps the selection rule, the seed, and the exclusion log with the sample. No file may be added once the result is visible, and no hard file may be dropped to improve the rate. A repeatable draw lets the team rerun the review after the workflow changes.

Pull files from the whole window, including the week when coverage was short and volume was high. A list drawn only from calm days will make a weak control look steady. Write down the observation time and the initials of the person who opened the record. If the eligible list is short, show the raw count and say the estimate is unstable.

Analysis and reporting

Publish the spread, not only the average, because a median can conceal a long tail of untouched files. Show the open count and the oldest file by owner, then point to where delay collects. A finding without an owner or a next step is not ready for the pipeline meeting. The Mortgage Assistant writes the result in ordinary language and puts the technical detail in an appendix.

Separate what was observed from what it means, and label every table with its denominator and dates. Show the oldest open file and the missing documents, not just the completion percentage. The Mortgage Assistant names the decision that comes next and the operations owner who must make it. Place the sample count beside each rate so a reader can weigh the evidence.

Limits and sources of error

Three limits bound the answer: sampling error, coder judgment, and records that never reached the system. The Mortgage Assistant publishes the overlap agreement and any disputed code rather than declaring the coding clean. One team and one window produced this number, so it cannot stand in for an industry benchmark. The closing paragraph states what the review still cannot see.

A finished file does not prove the lending call was right, because this review scores documentation rather than credit quality. The design cannot read intent and cannot describe a file that left no trace in any system. The Mortgage Assistant prints the caveats directly under the finding so a reader cannot lift the number without them. If the rule changes midstream, retain the earlier and later versions and tag each file with the version applied.

The central limit is that a income calculation completeness record can be late, partial, or kept outside the team's systems. This method cannot recover work that left no trace or prove that a policy was followed in a silent file. The Mortgage Assistant prints those caveats next to the finding instead of hiding them in a footnote. Changes to systems, investor rules, product mix, or volume can date the result, so list those triggers.

Review cadence

Put the next review on the calendar and name the person who owns both the measure and the follow-up. Close with a single process change, its owner, and a date to recheck, not a wish list. If the same defect returns, repair the form, the source, or the routing rule before retraining anyone. The Mortgage Assistant reuses the same sampling rule and stores the note beside the change record.

The review for income calculation completeness is finished when the sample closes and the field guide is attached. The Mortgage Assistant files the extract, the guide, and the finding together so a later reviewer can rerun the work. The closing note names what moved, what did not, and what the review still cannot see. That note is worth more than a neat rate built on a denominator nobody can explain.

Put the brief to work

Use the finding as a review prompt, not as a substitute for policy or accountable judgment. Compare the cited source material with your current record, decision owner, and escalation path.

For implementation context, review the scope and handoffs in CRM and pipeline support.

Compare this evidence with How often do investor delivery files have a complete audit trail? to test a related workflow boundary.

Sources and further reading

These are the public sources associated with this brief. Open each source to review its scope and current guidance directly.

  1. Consumer Financial Protection Bureau
  2. U.S. Department of Housing and Urban Development
  3. Federal Housing Finance Agency
  4. U.S. Government Accountability Office

Evidence into operations

Review the workflow behind the evidence

Bring one recurring mortgage workflow and its current handoff. We will help identify the preparation, review, and escalation boundaries.

Discuss the workflow in a free call