Documenting a Close So It Does Not Depend on One Person

A close that only one person can run looks identical to a close anyone can run, right up to the week that person is unavailable. Then it either stops or restarts as a series of guesses, and nothing in the file tells the next preparer which of the two is happening.

That is the risk documentation addresses, and it is worth being precise about what the risk actually is: not that the work is hard, but that the knowledge needed to run it sits in somebody’s memory rather than in the engagement. This is part of CAS production support.

The useful content is the client-specific part

Documentation that explains how to reconcile a bank account is wasted on a competent preparer. Documentation that records which of this client’s three bank accounts funds payroll saves a question every period.

That distinction is the whole discipline. The goal is to remove decisions, not to teach accounting, and the things worth writing down are the ones that cannot be inferred from competence:

• The step sequence, in the order it is actually done
• Conventions specific to this client, and who decided them
• The source of each recurring entry, and when it was last re-derived
• Where each input comes from and who supplies it
• Open items carried forward from prior periods

The test is whether a preparer who has never seen the client could run the period from it and produce something the reviewer recognizes.

Written during the work, not afterwards

Documentation reconstructed from memory omits exactly the decisions that were made without noticing — and those are the ones a new preparer gets wrong. Someone recalling a close remembers its structure and forgets the forty small judgments inside it.

Writing it as the close runs has a second effect worth as much as the first: it exposes the conventions that were never actually decided. A preparer who cannot write down why a transaction is coded the way it is has found a question for the firm, which is the same mechanism the exception list uses for items inside a single period.

Unread documentation is worse than none

It has to live where a preparer is already looking when they run the close. Held somewhere separate, it is not read — and unread documentation is actively worse than none, because it is counted as a control that is not operating.

Which system it lives in is the firm’s decision about its own tooling. That it sits with the work rather than beside it is not really optional.

Update it when the decision changes, not on a schedule

A document reviewed quarterly drifts for a quarter at a time, and a preparer who finds one stale entry reasonably stops trusting the rest of it. Credibility is all-or-nothing here.

Tying the update to the decision — a convention changes, the document changes in the same pass — is what keeps it worth reading at all.

It converts re-performance back into review

A reviewer who can see the method applied is checking whether it was applied correctly. A reviewer with no method in front of them has to reconstruct one before forming any view, which is re-performance wearing the word review.

Since licensed review capacity is the constraint a firm cannot quickly expand — the argument in quality review before release — anything that spends it unnecessarily is expensive in the scarcest resource the practice has.

Recording decisions is not making them

The conventions belong to the CPA firm. Where writing them down reveals that one was never settled — an ambiguous coding, a treatment that has simply been repeated for years — that is a question for the firm, not something preparation resolves on its own authority.

The cost of all this is front-loaded into the first few cycles, which is precisely why it tends to be abandoned a month or two before it starts paying.