What a fractional CTO actually does

Three jobs, one title

Before agreeing a rate, decide which of these you are buying, because they need different people and produce different things. Most disappointing engagements are a mismatch between the version the company needed and the version the person is good at.

The leader of a team that already exists. Three to fifteen engineers, no head of engineering, decisions piling up. The work is direction and standards: sequencing, who is accountable, what quality bar applies, whether the rewrite is justified, how to review, how to hire. This version is largely about people and is best done by someone who has managed engineers rather than only built systems.

The decider for a founder who has no technical partner. A non-technical founder with a vendor or a couple of contractors and no way to tell whether what they are hearing is reasonable. The work is architecture judgement, vendor management and translation: is this estimate sane, is this technology choice going to be regretted, is what we have built worth keeping. This version is largely about judgement and needs deep hands-on background.

The assessor for an investor or an acquirer. An operating partner or a board that has a portfolio company with an unclear technology position. The work is diagnosis and a plan: what is the state of the system, what is the team capable of, what does it cost to fix, what must happen in the next two quarters. This version is short, intense, and produces a written document that other people act on.

They overlap, and one person can do all three well. But a person who is excellent at the third is not automatically good at the first, which is a people job in a company they visit once a week.

What the hours actually buy

This is the part most commonly misunderstood, and it is arithmetic rather than opinion. The role is not compressed — a full-time technology leader's job does not fit into eight hours a month at higher intensity. What happens instead is that some of the job is done and the rest is not, and the useful conversation is about which parts.

Commitment	What you get	What you do not
8–16 hours / month	An advisor. Architecture review, sanity checks on plans and estimates, interview support, a monthly written view	Anything requiring presence. They will not know your team, and problems that appear on a Tuesday wait
1 day / week	Real direction. Roadmap sequencing, hiring bar and interviewing, vendor management, decisions made and recorded	Day-to-day management, incident response, or being the person anyone escalates to
2–3 days / week	Effectively an interim head of engineering. Performance conversations, process, on-call ownership	Long-horizon institutional work — culture, deep customer knowledge, multi-year platform bets
Any commitment	Judgement, written plans, hiring quality, a firmer hand with vendors	Meaningful volumes of code. If they are coding, you bought a senior engineer at a leadership rate

The last row is the one to be blunt about. A part-time person writing production code creates a dependency on someone who is not there most of the week, which is the opposite of the point. There is one honest exception: a first prototype at a company with no engineers, explicitly labelled as throwaway and scheduled to be replaced. Anything past that, and the arrangement quietly becomes a bottleneck named after your most expensive contractor.

The deliverables to require

Ask for these in the agreement. All of them are normal, none of them is onerous, and their absence is the single best predictor of an engagement that ends without anyone being able to say what it produced.

A written technical plan within the first month. Current state, the three or four things that matter most, sequence, and what each will cost in people and months. Ten pages is plenty. This is the artifact that makes everything afterward checkable.

A decision log. One or two paragraphs per significant decision: what was decided, what the alternatives were, why. This costs the writer ten minutes and is worth an enormous amount later, particularly to the permanent hire who arrives in a year and needs to know why the database was chosen.

A risk register that is actually maintained. The things that could hurt you, in rough order, with an owner and a status against each. The value is not the list; it is that someone senior looks at it every month and moves things.

A monthly written update to you. What happened, what did not, what is worrying, what they need from you. Written, not verbal. Verbal updates are where difficult news goes to soften.

Hiring materials. Job descriptions, an interview loop, a scorecard, and their participation in interviews. If part of the mandate is building a team, this is the deliverable that outlasts them.

A named succession plan. What has to be true for this role to end, and who takes each piece. Written at the start, not at the end.

What it costs

Ranges seen in the US market, and they vary with seniority, city, and whether you are hiring an individual or a firm. Individuals typically sit at the lower end; firms cost more and cover absence and holiday.

Shape	Typical monthly range	Usually right for
Advisory, 8–16 hours	$3,000–$9,000	A founder who mostly needs a second opinion and an occasional review
One day a week	$8,000–$18,000	A small team without a leader, or an active vendor relationship to manage
Two to three days a week	$18,000–$35,000	An interim need — a departure, a turnaround, a rebuild decision
Fixed-scope assessment	$15,000–$50,000 once	An investor or board that needs a written diagnosis and a costed plan

On equity: it comes up constantly at early-stage companies and it is worth being careful. A small equity component alongside cash is reasonable and aligns interests. Equity instead of cash tends to produce an arrangement where the person's attention follows their paying clients, which is rational and not what you wanted. If cash is genuinely tight, buy fewer hours at the proper rate rather than the same hours at a discount, and be honest with yourself about which parts of the job are therefore not being done.

Compare against the alternative before deciding. A full-time engineering leader in most US markets is a substantial salary plus equity plus the cost of a search that takes months and sometimes fails. Two or three days a week of experienced leadership, starting in a fortnight, is frequently the better purchase for a year — and it also improves the odds that the eventual permanent hire is the right one, because a competent interim writes the job description and sits in the interviews.

The conflict nobody raises

Many fractional technology leaders also sell engineering capacity, either their own firm's or a partner's. That is not disqualifying and it is often practical — a person who can advise and also field a team is genuinely useful, and they carry real accountability for what gets built.

But the incentive is real and it should be named out loud rather than hoped away. The person advising you on whether to build in-house, whether to keep the current vendor, and how large the team should be, may earn more if the answer is “use our team.” Every reasonable firm in this position knows it. The ones worth hiring will raise it before you do.

Three requirements make it workable. Say which hat. Advice on staffing, vendor selection and build-versus-buy is given in the advisory role, and the person should say plainly when a recommendation would benefit them. Keep the exits open. No exclusivity, no requirement to use their delivery arm, no penalty for hiring someone else, and no notice period that makes leaving expensive. Separate the money. The advisory fee is not contingent on build work following. If it is discounted on the expectation of the build, the advice has been sold to you at a price that assumes a particular answer.

You can also simply ask: “What would you recommend if you were not able to do the work?” The quality of the answer, and how long it takes to arrive, tells you most of what you need.

When it works, and when it does not

It works when the problem is decisions rather than capacity. A founder who cannot evaluate a vendor's proposal. A team of six that keeps rebuilding the same thing because nobody chooses. A departure that left nobody senior. A rebuild-or-repair decision worth several hundred thousand dollars, where an experienced outside read is cheap insurance. A portfolio company where the quarterly technology update has never once been legible.

It does not work in four recognizable situations. When the real problem is that nobody has decided what the product should be — that is a product problem and no amount of technical leadership resolves it. When the person is given responsibility without authority, so their recommendations require someone else's approval that never comes. When you needed hands and bought a head, and the backlog stays exactly where it was. And when the person is spread across too many clients: at two days a week each, three clients is a full week, and four is a person who is unavailable to all of them. Ask directly how many clients they have and at what commitment. The arithmetic either works or it does not.

How it should end

A good fractional engagement is designed to shrink. The role exists because something is missing; when it is no longer missing, the role should reduce or stop, and that transition should be described at the beginning rather than negotiated awkwardly at the end.

Name the trigger in the agreement. The usual ones are a permanent head of engineering starting, the team passing six or eight engineers, or a defined program finishing. Then name the wind-down: a month or two at reduced hours to hand over, with the documents — plan, decision log, risk register, hiring materials — already written, because they have been written all along.

The test to apply at six months is simple. If the company would be in the same position on the day this person stopped as it was on the day they started, the engagement has been a subscription rather than a transition. That happens most often when the arrangement drifts into a standing advisory relationship that everyone is comfortable with and nobody reviews. Put a review date in the calendar at the start, and hold it.

Mistakes we see

• No written plan in the first month, so there is nothing to measure the engagement against
• Hired for hands when the backlog needed engineers, not decisions
• Responsibility without authority — recommendations that always need one more approval
• Too many clients, discovered when they are unreachable during the week you need them
• Advisory fee discounted against future build work, which prices the advice at one answer
• No decision log, so a year of reasoning leaves with the person
• No end condition, so a transitional role becomes a permanent line item nobody reviews
• Writing production code, creating a dependency on someone present one day a week

Before you sign

• You know which of the three jobs you are buying, and so do they
• The commitment is stated in days per week, not vaguely in hours
• A written technical plan is due in the first month
• A decision log, risk register and monthly written update are named deliverables
• You asked how many other clients they carry, and at what commitment
• Any conflict with a delivery arm is disclosed, and the fee is not contingent on build work
• No exclusivity, and a short notice period on both sides
• The end condition is written down at the start
• Hiring materials and interview participation are included if a team is being built
• A review date is in the calendar, six months out

Bottom line

Decide which of the three jobs you need before you talk about rates, because the same title covers a people manager, a technical judge and a diagnostician. Buy decisions, not hours, and insist that the decisions get written down — the plan, the log, the register, the monthly note. Expect judgement, hiring quality and the handling of vendors to improve quickly, and expect delivery speed and team morale to move slowly, because those need presence. Ask what they would advise if they could not do the work themselves. And write the ending at the beginning: the job is to become unnecessary, and an engagement that never shrinks was never really transitional.

Frequently asked questions