Table of contents
Outsourced data entry is easy to measure by volume and hard to measure by quality. A vendor can return a hundred files in a day while ten of them contain a transposed loan amount or a wrong property address. The error may not surface until underwriting reviews the file or a closing document is generated. A completion check catches the mistake while the source document is still easy to compare.
The check is not a full re-entry of every field. It is a short sample of the fields that carry the most risk and the highest cost when wrong. It should take minutes per file and produce a record the team can review. This guide explains how to build the check and how to use the results without turning it into a blame exercise.
Choose the fields that carry the most risk
Start with the fields that change a decision, a disclosure, or a payment. Common examples include the borrower's legal name, the loan amount, the property address, the occupancy type, the income figure, and the asset total. The exact list depends on the loan type and the lender's requirements. Ask the processor and the underwriter which errors cause the most rework.
Separate a critical field from a convenience field. A wrong middle initial may create a minor correction, while a wrong loan amount can change a disclosure and delay closing. The check should focus on the critical fields first. A short list that is actually checked is better than a long list that is skimmed.
Compare the system value with the source
Read the source value first, then the system value. Reading the system value first creates an anchor that makes a wrong source value look correct. This small habit improves accuracy and reduces confirmation bias. The checker should record both values when they differ.
Check the format as well as the content. A date entered in the wrong format can sort incorrectly even if the day and month are right. A phone number with a missing digit can break a borrower contact. The check should confirm that the format matches the system's expectation. Format errors are easy to fix and easy to miss.
Record the result as pass, corrected, or escalated. A pass means the values match. A corrected means the checker fixed the field and noted the change. An escalated means the difference needs a decision from the processor or underwriter. The three outcomes keep the record simple and actionable.
Record corrections and near misses
Every correction should include the field, the system value, the source value, and the person who corrected it. The record should also include the source document and the date. This detail makes it possible to see patterns and to answer a later question. A note that says "fixed data" is not evidence.
A near miss is a field that was wrong but did not affect a decision because the checker caught it. Near misses are valuable because they show where the process is fragile without the cost of a bad outcome. Include near misses in the weekly review. A team that only tracks failures misses the chance to prevent them.
Do not delete the original entry after a correction. The system audit trail should show the change, and the check record should explain why. A clean history helps a reviewer understand what happened. It also protects the checker if the correction is later questioned.
Sample a small batch each week
Sample a fixed number of files each week, such as five or ten, rather than checking everything. Choose the sample across different loan types, processors, and source documents. A sample that always comes from the same queue will not reveal the range of errors. Random selection is better than choosing the files that look easy.
Use a consistent sample size so the results are comparable over time. If the volume changes, keep the sample size and report the sampling rate. A week with a small sample has a wider margin of uncertainty, and the report should say so. Do not present a five-file sample as a precise error rate.
Feed errors back into the checklist
When the same error appears twice, update the checklist or the source instructions. A repeated wrong loan amount may mean the field is easy to confuse with the purchase price. A repeated wrong address may mean the source document has two addresses. The fix should address the cause, not just the symptom.
Add a short example to the checklist when a field is ambiguous. The example should show the correct value and the common mistake. Keep it brief so the checklist stays usable. A checklist that grows into a manual will not be read during a busy week.
Review the checklist with the vendor's team lead. The vendor should understand the standard and the examples. A conversation about the check is more useful than a scorecard sent after the fact. The goal is a shared definition of correct, not a surprise penalty.
Separate training from system fixes
Some errors come from a person who does not know the rule, while others come from a system that invites the mistake. A field that accepts a date in any format may need a validation rule. A source document that places two addresses side by side may need a clearer label. The response should match the cause.
Before assigning training, ask whether the system could prevent the error. A drop-down list, a required field, or a format mask can remove a whole category of mistakes. A training session may still be useful, but it should not be the only fix. The team should choose the cheapest reliable control.
Track the error rate before and after the fix. If the rate does not move, the fix may not address the real cause. Reopen the review with the checker and the vendor. A small test can show whether the change works before it is rolled out to the whole team.
Keep the check short enough to survive
A completion check competes with the daily queue. If it takes too long, it will be postponed during busy weeks and abandoned during a rush. Keep the field list short, the sample small, and the record simple. The check should fit into a regular part of the day, not an extra project.
Use the same form each week so the team does not rebuild the process. Store the results in one place and review them on a fixed cadence. A consistent small check produces more learning than an occasional exhaustive audit. The goal is a habit, not a perfect measurement.
Share one finding and one fix with the team each month. Recognize the checker who caught a near miss and the vendor who corrected a pattern. Do not use the check to rank individuals on a small sample. The purpose is a more accurate file, not a scoreboard.
The Consumer Financial Protection Bureau's mortgage resources describe how accurate borrower information supports fair treatment and clear disclosures. A completion check supports that principle with a focused comparison. The loan processing support team can run the sample while the vendor keeps the entry workflow. Pair it with the monthly borrower data audit so the check and the audit use the same field definitions.