Bookkeeper vs Controller vs Fractional CFO: Who to Hire When
A bookkeeper costs $400 to $2,500 a month. A controller, $2,500 to $4,200. A fractional CFO, $8,000 to $15,000.
Three roles. Three price points. Three completely different problems. And most founders I talk to are trying to solve all three with one hire — usually the most expensive one, usually a stage too early.
I wrote about the most expensive version of that mistake — hiring a fractional CFO when what you needed was a controller — in why most founders hire a fractional CFO too early. This post zooms out to the whole stack: all three roles, what each one actually does in a month, and the order to add them so you're never paying senior rates for junior work.
The cost of getting the order wrong isn't theoretical. Hire out of sequence and you'll spend 6 to 12 months and $50,000 to $150,000 before anyone notices the finance function is quietly broken. I see it every week, and it's almost always avoidable.
Three roles, three price points, three problems
Most founders treat "finance hire" as one category. It's three. The cleanest way to keep them straight is by the question each one answers:
| Role | Cost (2026) | The question it answers | Hire when |
|---|---|---|---|
| Bookkeeper | $400–$2,500/mo | What happened? — transactions categorized, accounts reconciled, the month closed | Day one — first outside money or ~$10K/mo in expenses |
| Controller | $2,500–$4,200/mo | Can I trust the numbers? — accuracy, a real close process, statements that tie to the bank | Around $1M ARR, or when bad data starts blocking decisions |
| Fractional CFO | $8,000–$15,000/mo | What should I do? — models, fundraises, unit economics, capital allocation | Series A prep, with the books already clean |
Past, present, future. The bookkeeper tells you what happened. The controller makes that record trustworthy. The CFO uses the trustworthy record to decide what happens next. When a founder says "I need finance help," they're usually feeling one of these three pains specifically — but they reach for the title with the most prestige instead of the one that matches the pain.
What each one actually does in a typical month
Titles are abstract. Here's the concrete work, so you can match the role to what you actually need done.
The bookkeeper's month. Categorizes every transaction. Reconciles the bank, the credit cards, and the Stripe or Mercury feed. Matches invoices to payments. Runs the monthly close and hands you a P&L and a balance sheet. That's the job. A bookkeeper does not tell you what the numbers mean, whether your burn is too high, or how to model a raise — and shouldn't be asked to.
The controller's month. Owns the close calendar and gets it down to five business days. Reviews the bookkeeper's work and fixes the chart of accounts when it drifts. Builds the AR/AP rhythm so cash in and cash out are both predictable. Produces statements that tie to the bank to the dollar and survive a board's questions. Handles audit prep and the tax-package handoff. A controller does not build the fundraise model, sit in the room defending a pivot, or decide whether to cut burn or push growth.
The fractional CFO's month. Builds and maintains the operating model and the 13-week cash forecast. Runs scenario analysis — base, downside, upside. Sits in board prep and owns the numbers in the deck. Diagnoses unit economics by cohort and finds the customer segment quietly destroying your margin. Leads the raise. Makes the call, at nine months of runway, on whether to extend or accelerate. A fractional CFO does not fix your books — and if yours is spending the month in your chart of accounts, you have the wrong person in the seat or hired them a stage too soon.
Notice the handoffs. Each role hands clean output up to the next. Skip a layer and the one above it spends its expensive hours doing the missing layer's cheap work.
The right order, stage by stage
The figure above this post is the whole argument in one image: each hire enters at a different stage and stays as the next one joins. Here's how that plays out in practice.
Pre-revenue and pre-seed: a bookkeeper, full stop
You do not need a controller. You absolutely do not need a CFO. Your "model" is a back-of-the-envelope, and that's the correct level of effort for your stage.
Get a bookkeeper — even the $400-a-month software-led tier is fine here — track your runway monthly, and put every dollar you didn't spend on premature finance hires into product and the first few sales. Doing the books yourself in QuickBooks is the most expensive cheap thing in startup finance: your time is worth more than $1,500 a month, and the categorization errors compound silently until someone has to unwind them later. If you want to sanity-check burn and runway in the meantime, the CFO Toolkit does it free.
Around $1M ARR: add a controller
Now data integrity starts to matter. You have real revenue, real vendors, real payroll, and decisions riding on numbers that need to be right. This is the moment to add a controller — roughly $2,500 to $4,200 a month depending on whether you go part-time domestic or full-time offshore, with around $2,900 the typical effective spend.
A senior controller at that price does roughly 70% of what most early-stage founders think they want from a CFO — clean books, reliable statements, a close you can trust. The other 30%, the strategy and fundraising work, usually isn't urgent yet. One option worth knowing about: a full-time controller in LATAM (Mexico, Colombia, Argentina), timezone-aligned and on U.S. GAAP, often lands at the same ~$2,900 effective spend as a part-time U.S. fractional controller but gives you a dedicated owner of the entire close. For a U.S. startup, that's frequently the highest-leverage hire on the team. If cleaning up the foundation is the whole job right now, our Lean Finance Stack setup is built for exactly this stage.
Still no CFO. A sharp founder plus a good controller produces everything a seed-stage investor actually wants to see.
Approaching $5M ARR and Series A: add the fractional CFO
The books have been clean for a while — the controller has seen to that. Now the model becomes the central document of your next chapter, and the questions get strategic: what's our CAC payback by cohort, what does the raise look like under three scenarios, do we buy the competitor or build. That's fractional CFO work, and on top of a clean foundation a good one earns the fee in about six weeks. If you're inside the 12-to-24-month window before the round, the Raise-Ready Sprint is the project-based version of this.
This is also where getting the order right pays off most visibly: because the controller already built the foundation, the CFO spends month one doing actual model work instead of chart-of-accounts triage.
Series A and beyond: the stack matures
The controller stays — often going full-time or offshore. The CFO's scope deepens, embedding hard during the raise itself. The bookkeeper keeps running underneath both. Eventually, past Series A, you hire a full-time finance leader above the whole stack. But you grow into that, role by role — you don't leap to it.
Four signs you hired the wrong role
Three of these are over-hiring. One is under-hiring. They're equally expensive.
Your fractional CFO spent month one in your chart of accounts. You needed a controller. You're paying $8K-a-month rates for $2,900-a-month work, and getting a worse controller than you'd have hired directly, because cleanup isn't a CFO's training or tooling.
Your bookkeeper is being asked to build the board model. You've up-leveled someone past their role. Bookkeepers record; they don't forecast. The model you get will be confident and wrong, which is worse than no model. You need a controller, then a CFO — not a stretched bookkeeper.
You have a CFO, and your runway still changes by 30% depending on which spreadsheet you open. That's a data-integrity problem, and it lives below the CFO. Per Robert Half's 2026 Salary Guide, full-time controllers run $10K to $14K a month base — so paying CFO rates to patch a controller-shaped hole is the second-most-expensive way to solve it.
Your close is clean and trustworthy, but nobody can answer "what happens to runway if we miss Q3 by 20%?" This is the under-hire. You have accuracy and no strategy. A controller got you a reliable rear-view mirror; now you need someone looking through the windshield. That's the CFO.
The five-question diagnostic
Answer these honestly. They point to the right hire faster than any sales call.
- Can you close the month in under 10 business days? No → you have a bookkeeper or controller gap, not a CFO need.
- Do your runway numbers agree across every spreadsheet you keep? No → controller. Full stop.
- Are you within 12 months of a Series A? No → you almost certainly don't need a fractional CFO yet.
- Could you build a credible three-scenario raise model on your current books? If the books are clean but you can't build the model → that's the CFO. If the books aren't clean → controller first; a model on bad data breaks under diligence.
- Could you write, on a napkin, the three specific things a CFO would do in their first 90 days? No → you don't have a CFO problem yet. You have a "my peers have a CFO" feeling, which is not the same thing.
Fail 1 or 2 and your next hire is in the bookkeeper-to-controller range, regardless of how badly you want the CFO title on the org chart. Pass 1 and 2 but stall on 3 through 5, and now it's a real CFO conversation.
What this looks like in practice
A composite of the founders I talk to most weeks, anonymized.
The call: seed-stage SaaS, around $1.4M ARR, and the founder wants a fractional CFO because a board member told them to get one. First question I ask — can you close last month and hand me a P&L that ties to your bank? Long pause. The books are three weeks behind and two of their four bank feeds were never connected.
They don't need me yet. They need a controller for ninety days to make the close clean and reliable, around $2,900 a month, with a bookkeeper running underneath. Then, twelve months out from a Series A, we have the CFO conversation — the model, the raise, the capital plan — on top of numbers that actually hold. Same destination, right order: roughly $40K a year instead of $96K, and a finance function that works instead of one that's quietly on fire.
A note on AI in 2026
AI has eaten most of the bookkeeper's month and a real slice of the controller's: categorization, reconciliation, variance analysis, a first-draft 13-week forecast off clean ERP data. The "what happened" layer is cheaper and faster than it's ever been.
What it hasn't touched is the "what should I do" layer — deciding whether to cut burn or push growth at nine months of runway, spotting which contract has real collection risk, defending a pivot at the board. So the calculus has actually sharpened: if the cheap layer is now nearly free, paying a $10,000-a-month CFO to do it is more wasteful than it was two years ago, not less. Buy software for the arithmetic. Hire humans for the judgment.
Three things you can do right now
If you're not sure which stage you're in, take the Raise-Ready Scorecard. Four minutes, and it tells you whether you're having a CFO conversation or whether you've got controller-stage cleanup to do first.
If you need to pressure-test burn or runway today, the CFO Toolkit is free — same modeling backbone I use with clients, browser-based, no signup. You'll get the most from it on clean books, but even on rough numbers it'll tell you if your runway picture is roughly right.
And if you want a second opinion on which of the three you actually need, book a 30-minute call. I'll tell you honestly — a bookkeeper, a controller, a CFO, or none of them yet. I've sent founders away with "you need a $3,000 controller, not me," and I'd rather do that than take a CFO retainer to do controller work for a year.
I'm not against any of these hires. I'm against paying for one when you needed another. Hire to the problem in front of you — the late close, the number you can't trust, the model you can't build — not to the title you think a company your size is supposed to have.
Three roles, three price points, one rule: add them in order. Skip a step and you'll pay senior rates for junior work — $50,000 to $150,000 of it — before anyone notices the order was wrong.