Protocol premise

The study would test whether a remittance batch can be applied from its own evidence, without an analyst guessing the payer or the invoice.

Protocol status and question

This is a proposed research protocol, not a report of completed findings. It asks how often a lockbox remittance supplies enough information to identify the payer, the invoice, and the amount without an analyst inference. The answer would describe the evidence a cash application team actually receives, not the evidence its policy assumes. Record who approved the question and when. A later reviewer should be able to see what the study was allowed to conclude.

The sponsor should state in advance which decisions the result could inform. A finding about missing payer fields might support an intake change, while a finding about missing invoice detail might support a customer communication change. The protocol cannot decide whether a customer is creditworthy or whether a payment should be written off. Write the intended use in one sentence so the scope cannot drift. The sponsor should record that decision before collection starts.

Population and observation unit

Define the population as all lockbox receipts processed in a stated window, with the bank, the currency, and the account included in the frame. The observation unit is one receipt record, not one deposit batch, because a batch can mix several levels of evidence. Receipts that arrive outside the lockbox should be excluded and counted separately. Write the frame as a dated list so a later reviewer can rebuild it. Describe the unit in one sentence that a new reviewer can apply.

Freeze the inclusion and exclusion rules before looking at outcomes. A receipt with no remittance page is still part of the population, because the absence is the phenomenon under study. A duplicate scan of the same receipt should be linked to the original rather than counted twice. Missing records should remain visible in the denominator. Keep duplicate and missing records visible in the denominator.

Definitions and measures

Define payer completeness as the presence of a legal name and a customer identifier that match an approved master record. Define invoice completeness as the presence of at least one invoice number or a remittance reference that the ledger can resolve. Define amount completeness as a stated payment amount that reconciles to the bank record. Each definition needs an example and a counterexample. A definition that two reviewers read differently is not ready.

Report the measures separately and together. A receipt can identify the payer but not the invoice, and that combination is more useful than a single percentage. Also record the source of each field: remittance page, bank reference, customer portal, or analyst inference. The source distinguishes evidence from recovery. Write each definition beside the field it governs.

Evidence standard

Every classification should link to the receipt image, the bank record, and the customer master entry used to verify it. A field counts as present only when the reviewer can see it in a permitted source. An analyst's memory or a later chat message is not sufficient evidence, even when it turns out to be correct. A summary without the source is not evidence. Record the date the evidence was checked.

Use a versioned data dictionary with field names, permitted sources, and missing-value codes. Unknown, not applicable, and unavailable should be distinct. A receipt that has no remittance page is not the same as one whose page is illegible. The distinction affects the recommendation that follows. Record the date the evidence was checked.

Sampling and collection

A census of one month is preferable to a small sample when the volume is manageable. If sampling is necessary, stratify by bank, customer type, and receipt amount so the common and uncommon cases are both represented. Select the sample before reviewing outcomes and keep a log of replacements. Do not replace a hard case with an easier one. Log every replacement and the reason for it.

Train reviewers on a common set of ten receipts and compare their classifications before the main collection starts. Resolve disagreements about what counts as present. During collection, preserve the first classification and record any later correction separately. The initial judgment is part of the evidence about how difficult the classification is. Do not replace a hard case with an easier one.

Analysis and reporting

Report counts and rates for each field, with the denominator and the observation window beside every number. Show the combination of fields needed for a clean application, not only the average completeness. A high payer rate can hide a low invoice rate, and the combination is what determines whether the receipt can be applied. Break the result down by the groups named in the protocol. Name the counterexamples as well as the main pattern.

Present the results by source and by bank so the reader can see where the gap originates. Avoid ranking customers or analysts unless the study was designed to compare them. The useful output is a description of the evidence supply and a short list of process questions, not a verdict about individual performance. Report counts with rates and keep the denominator visible. Break the result down by the groups named in the protocol.

Limits and sources of error

The protocol does not measure whether a receipt was applied correctly, only whether the evidence was sufficient to apply it. It may miss receipts that were resolved before the study window or through a channel outside the lockbox. Reviewer learning during collection can make later classifications more consistent than earlier ones. Explain how selection may have shaped the sample. Record any policy or system change during the window.

A change in bank formatting or customer behavior during the window can bias the result. Record any known change and shorten the window if necessary. The finding should state its population and date range clearly, because remittance quality can shift with a single customer migration. Do not generalize beyond the documented frame. Explain how selection may have shaped the sample.

Review cadence

Run the review quarterly if the lockbox mix is stable, or after any bank or portal change. Compare the same measures across periods and explain any change in the frame. A sudden improvement may reflect a process fix or a change in which receipts are included. Version the protocol when the method changes. Keep the same window length so the comparison means something.

Keep the protocol and the data dictionary with the results so the next review can repeat the method. Assign an owner for the follow-up questions and a date for the next review. The protocol is useful only if it produces a decision about what to change and a way to check whether the change worked. Compare the result with the prior period and explain the change. Store the data and the protocol together for the next pass.

Sources

  1. U.S. Small Business Administration
  2. IRS Publication 583, Starting a Business and Keeping Records
  3. Financial Accounting Standards Board
How often do partial short-pays carry an unverified deduction code?What share of suspense balances clear without an assigned owner?How often do customer overpayments remain unapplied at statement date?

Put the protocol into operational context.

Discuss the queue, evidence, ownership, and safeguards that apply to your workflow.

Discuss your workflow