These 2 roles get treated as seniority levels of the same job. They are not. They answer different questions, and a business can have an excellent bookkeeper and still close its year 3 weeks late with numbers nobody will sign.
Two different questions
A bookkeeper answers: was it recorded? Every transaction captured, coded to the right account, reconciled to the bank, with AP and AR kept current.
A controller answers: is it right? Not whether the entry exists, but whether it is in the correct period, at the correct value, under a policy applied the same way it was applied last quarter.
The first question has a factual answer. The second requires judgement, and judgement is what you are paying the difference for.
| Accountable for | Bookkeeper | Controller |
|---|---|---|
| Transaction capture and coding | Yes | Reviews |
| Bank and card reconciliation | Yes | Reviews |
| AP and AR runs | Yes | Sets policy |
| Accruals, prepayments, cut-off | No | Yes |
| Revenue recognition policy | No | Yes |
| Chart of accounts design | No | Yes |
| Close calendar and sign-off | Executes | Owns |
| Audit and diligence readiness | Supplies | Defends |
Why the difference only shows at year end
A well-kept ledger with no controller over it looks correct for 11 months. Cash reconciles. The P&L has no obvious holes. Nothing in the monthly management accounts announces a problem.
Then the year closes and 3 things surface at once. Revenue booked on invoice date rather than delivery. A 12-month software contract expensed in the month it was paid. Bonuses accrued in the year they were paid rather than earned. None of these are bookkeeping errors. Each one is a policy decision nobody was accountable for making.
The cost is not the correction. It is that the corrected numbers are different from the ones you reported to a board, a lender or an investor 11 times.
Bookkeeping errors are found by reconciliation. Policy errors are found by someone asking why, and only a controller is paid to ask.
When you need the second seat
Most businesses need a bookkeeper from day one and a controller considerably later than they think. The trigger is not size. It is the arrival of judgement calls that recur.
- Deferred or recurring revenue. Any subscription, retainer or multi-period contract creates a recognition question every single month.
- Inventory or work in progress. Valuation and cut-off are judgement, not capture.
- More than one entity, or more than one currency. Consolidation, intercompany elimination and translation all need a policy someone owns.
- An external reader with consequences. A lender with covenants, an investor with information rights, or a buyer.
- A close that is drifting. If day 8 has become day 20, it is usually because nobody owns the calendar.
What it costs, and the sequence that works
Outsourced bookkeeping starts around $350 a month at our end and typically runs $500 to $2,000 a month in the wider market, scaling with transaction volume rather than revenue. A controller layer is a different order of cost, whether hired or bought as part of a retainer.
The sequence that works is unglamorous: get capture right, then put a close calendar over it, then add the review layer. Buying the review layer first produces a well-reviewed set of unreliable numbers, which is worse than either seat alone because it carries authority it has not earned.
One thing to check this month
Take your last three monthly P&Ls and find your largest recurring expense line. If the amount is materially different in one of those months and nobody can explain why in a sentence, you have found the seam. That is a policy question, not a coding question, and it will still be there at year end.