# A Bookkeeping Quality Control Process for Firms

- Author: AutoBankConvert (https://www.autobankconvert.com)
- Published: 2026-03-13
- Updated: 2026-03-13
- Category: Client work
- Canonical: https://www.autobankconvert.com/blog/bookkeeping-quality-control-process

> A bookkeeping quality control process for firms: sample closed months, score errors by type, feed fixes back into rules and training across all clients.
A bookkeeping quality control process checks months that are already closed, not months in progress. Each quarter, a person outside the client's team samples closed months, re-checks a set of lines, scores every error by type and severity, and turns repeated errors into rule changes or training. The scores show whether the firm's books are getting better.

## What is a bookkeeping quality control process?

Quality control is a second look at finished work, run on a schedule across the whole client list. Monthly review catches errors before a month is released. Quality control asks a different question: of the months we released, how many were right? It measures the review itself, not just the preparer.

The two are easy to confuse because both involve checking books. The difference is timing and scope. Review happens on every client, every month, before sign-off, and is done by the client's reviewer. Quality control happens on a sample, after sign-off, and is done by someone who did not work on that client.

### Why a firm needs both

Review alone cannot tell you how good it is. A reviewer who misses a type of error will keep missing it, and nothing in the monthly routine will reveal that. A periodic sample, scored the same way each time, shows the error rate the client actually receives.

Quality control also produces evidence. If you ever sell the firm, a record of scored samples over several years is one of the clearest answers to a buyer's question about the state of the books; the guide to [selling a bookkeeping practice](/blog/selling-a-bookkeeping-practice) covers what buyers ask for. If you are the buyer, the guide on [buying a bookkeeping practice](/blog/buying-a-bookkeeping-practice) explains why a sample of the seller's closed months is worth running before you sign.

## Sampling closed months for a quality control check

A quality control sample pulls a few closed client months each period and re-checks a fixed set of lines in each. Choose months across every preparer and every client tier, so the sample says something about the whole firm. A practical size for a small firm is one month per preparer per quarter, plus any client with a known problem.

### Choosing which months

Spread the sample on purpose rather than picking at random from the full list:

- **Every preparer appears.** Otherwise one person's work dominates the result.
- **Every client tier appears.** Simple clients will score well; complex ones show where the risk is.
- **Recent and older months.** A month closed last week and one closed three months ago test different things: the current work, and whether errors were found later.
- **Clients with a change.** A new bank, a new entity or a new preparer are all moments when errors rise.

### Choosing which lines

Within each sampled month, check a fixed set rather than everything:

1. The balance chain: opening balance matches the prior month's closing, and each statement balanced.
2. The ten largest transactions, in and out.
3. Every line in the uncategorized or holding account at month end.
4. Every transfer between the client's own accounts.
5. A random ten lines from the rest.

That is around thirty to forty lines per month. It takes a reviewer well under an hour on a typical small business client and still finds the errors that matter.

### Who does the checking

The checker should not have touched the client in the sampled month. In a firm of two, each person checks the other's clients. A solo bookkeeper can still run a version of this by checking months at least a few weeks after closing them, when memory of the reasoning has faded and the work is read as it stands.

## How should quality control errors be scored by type?

Score each error by what it does to the books, not by how much effort it takes to fix. An error that changes a total or a tax-relevant figure weighs far more than an untidy memo. A small, fixed set of types and weights keeps scores comparable from quarter to quarter and between checkers.

### An error scoring table

The types and weights below are an example. Choose your own, then do not change them often, or the trend loses meaning.

| Type | What it means | Example | Weight |
| --- | --- | --- | --- |
| Missing or doubled line | The books do not contain exactly what the statement shows | A statement imported twice | 10 |
| Wrong side | Income as expense, a transfer as income, a draw as an expense | Owner contribution booked as sales | 8 |
| Tax-relevant misclass | Wrong account where the account changes the tax result | Equipment expensed when it should be an asset | 6 |
| Ordinary misclass | Wrong account, same tax result | Software booked as office supplies | 3 |
| Unexplained holding item | A line left in a holding account with no note or question | Unknown deposit with no follow-up | 3 |
| Documentation | Right entry, missing support or memo | Large purchase with no description | 1 |

### A worked score for one sampled month

Here is an example. The client, the counts and the score are illustrative. Suppose a checker samples one month for a therapy practice, 36 lines checked.

| Error found | Type | Weight |
| --- | --- | --- |
| Insurance payer deposit booked as client fees from a person | Ordinary misclass | 3 |
| Transfer to the owner's savings booked as a supply expense | Wrong side | 8 |
| Two unknown deposits left with no question sent | Unexplained holding item, twice | 6 |
| Equipment purchase over the practice's threshold booked as expense | Tax-relevant misclass | 6 |

The month scores 23 points across 36 lines. On its own the number says little. Against the firm's other sampled months, and against this preparer's last quarter, it shows whether things are improving. The insurance deposit is a known trap for that kind of client, and the guide to [bookkeeping for therapists](/blog/bookkeeping-for-therapists) explains how those deposits differ from client payments.

### Score per line, not per month

Divide the points by the number of lines checked. A long month with many lines will collect more points by size alone. Points per line, tracked over time, is the figure to watch.

## Turning quality control findings into rules and training

Every scored error leads to one of three fixes: a change to a client's rules, a change to the firm's procedure or review list, or a training note for a person. Decide which fix applies for each error type before the next sample. A finding that leads to no change will appear again next quarter.

### Errors that point at rules

Errors that repeat on the same vendor or the same kind of line usually come from a categorization rule that is missing or wrong. Fix the rule for that client, and check whether the same rule exists on other clients. In AutoBankConvert, approving a line in the review queue teaches that client's rules, so the same vendor is categorized correctly the next month; a rule fixed after quality control works the same way going forward.

### Errors that point at the process

Errors that appear across clients and preparers, such as transfers booked as income, point to a gap in the procedure or the review list. Add a line to the review list for that error type. If a step can be checked automatically, such as confirming each statement balanced before a month is released, make it automatic. The guide to [accounting automation for bookkeepers](/blog/accounting-automation-for-bookkeepers) covers which checks are worth moving out of human hands.

### Errors that point at a person

Errors concentrated in one preparer's work are a training matter. Share the specific lines with the person, explain the right treatment, and look at their work again in the next sample. Keep this private and factual. Quality control that feels like a hunt for blame makes people hide errors instead of reporting them.

### Close the loop in writing

Keep a short log after each round: the errors found, the fix chosen for each, and the date it was made. At the next round, check first whether the previous fixes worked.

## A quality control calendar across the whole client list

Quality control works when it runs on a calendar, not when someone finds time. Quarterly is a sensible pace for a small firm: frequent enough to catch a trend, rare enough to fit around closes. Avoid the busiest weeks of the year and schedule the round for the same weeks each quarter.

### An example calendar

- **Week 1 of the quarter.** Choose the sample: months, clients and checkers.
- **Weeks 2 and 3.** Check and score each sampled month.
- **Week 4.** Meet for half an hour to agree the fixes, then make them.
- **Next quarter.** Start by confirming the previous fixes worked.

### Who checks whom

Rotate checkers so no one always checks the same colleague. In a larger firm, a senior bookkeeper or the owner runs the round and checks a few months personally. In a firm of two, swap clients each quarter. In a firm of one, the delayed self-check described above is the honest option, and an occasional outside review from a peer you trust adds a view that self-checking cannot.

## Where quality control sampling misses errors

A sample finds errors that are common. It will not reliably find rare ones, such as a single misbooked loan or one month a whole account was left out. Quality control lowers the routine error rate; it does not prove that any one client's books are right. Pair it with checks that cover every month in full.

### Gaps a sample leaves

- **A whole missing account.** If the client opened a card you never heard about, no sampled line will show it.
- **Errors spread over many months.** A recurring charge booked to the wrong account looks right line by line.
- **Problems only the client knows.** A deposit that was really a loan looks like income to the checker too.

### Checks that cover what sampling cannot

Run these on every client, every month, rather than on a sample: the balance check on each statement, the balance chain from month to month, a list of every account you expect statements for, and a year-over-year look at each expense account at year end. Sampling then measures quality on top of a floor that every month already meets.

If you want every statement balance-checked before it is marked done, so quality control can concentrate on judgment, [see how AutoBankConvert runs each month](/).