These 3 roles are usually presented as a ladder, which makes it sound like a budget decision. It is not. Each one answers a different question, and buying them in the wrong order is the most expensive mistake in small-company finance.
3 questions, not 3 levels
- Bookkeeper: was it recorded? Every transaction captured, coded and reconciled. A factual question with a factual answer.
- Controller: is it right? Correct period, correct value, under a policy applied consistently. A judgement question.
- CFO: is it going to be a problem? What the numbers imply, what to do about it, and what to tell the board. A forward question.
Each one depends on the answer below it being reliable. That dependency is the whole argument for the sequence.
| Accountable for | Bookkeeper | Controller | CFO |
|---|---|---|---|
| Capture, coding, reconciliation | Yes | Reviews | No |
| Accruals, cut-off, policy | No | Yes | Sets direction |
| Close calendar and sign-off | Executes | Owns | No |
| Forecast and scenarios | No | Supports | Yes |
| Board and investor reporting | No | Prepares | Owns |
| Fundraising and transactions | No | Supplies | Yes |
| Typical monthly cost | $350 to $2,000 | Part of a retainer, or a hire | $1,750 to $12,000 |
The order that works
Capture first, always. Then a close calendar with someone accountable for the date. Then policy and review. Then the forward layer.
Skipping ahead does not accelerate anything. A CFO whose first 6 months go on cleaning up bookkeeping is an extremely expensive bookkeeper, and they usually leave, because that is not the job they took.
Every seat above bookkeeping is only as useful as the ledger underneath it. Buying authority over unreliable numbers just makes the wrong answer more convincing.
When each seat earns its fee
- Bookkeeper: from the first month of trading. There is no threshold. Doing it yourself in a spreadsheet is a decision to pay for it later at catch-up rates.
- Controller: when judgement calls recur. Deferred or recurring revenue, inventory, more than one entity or currency, or an external reader with consequences. Also when a day-8 close has quietly become day 20.
- CFO: when decisions outpace the data. A board asking harder questions, a raise, a lender, or a plan that needs modelling rather than reporting.
What each costs, honestly
Bookkeeping starts at $350 a month with us and runs $500 to $2,000 in the wider market, priced on transaction volume rather than revenue. A fractional CFO retainer starts at $1,750 a month here and runs $3,000 to $12,000 across the market. The controller layer is rarely priced on its own; it is either bundled into a retainer or hired permanently.
A full-time CFO is roughly $260,000 loaded plus equity, which is why the fractional version exists. But note the shape of the numbers: the cheapest seat is the one everything else depends on, and it is the one most often bought last.
The two failure modes
Buying too high, too early. A CFO on an unreliable ledger. 6 months of remediation, an expensive departure, and a board that has now been given 3 quarters of numbers that later moved.
Buying too low, too long. Perfect bookkeeping and nobody asking what it means. Nothing visibly breaks. The plan simply stops being ambitious, because no one has the bandwidth to model the ambitious version.
Where you actually are
Answer 3 questions. What day do your accounts close? Can you explain last month’s largest variance in one sentence? Does anyone model a decision before you take it? A no to the first is a bookkeeping problem, a no to the second is a controller problem, and a no to the third is the CFO seat.