Hire an engineer, or engage a firm?
Almost everyone models this as salary versus rate. That comparison is wrong in a predictable direction, and it is wrong by enough to flip the decision. Here is the model with the missing terms put back — bring your own numbers.
- Salary versus rate is not the comparison. It omits everything that makes a hire expensive before anyone writes code.
- The seven usually-missing terms: recruitment, vacancy, ramp, employer overhead, equipment, management time, and mis-hire risk.
- Time is the term that moves the answer most — and it is the one nobody prices.
- The decision is rarely about cost at all. It turns on permanence, direction and risk.
- This page has no benchmark numbers on purpose. Yours are the only ones that matter, and industry averages in this category are mostly unsourced.
The hire-versus-buy question is normally settled by putting an annual salary next to an agency rate and seeing which is smaller. On that comparison a permanent hire almost always wins, which is why it is the comparison most often presented — by internal advocates of hiring and by recruiters alike.
It is wrong because it compares a fully-loaded external price against an unloaded internal one. The external number contains recruitment, employment overhead, equipment, management and idle capacity, because a vendor cannot survive otherwise. The internal number usually contains none of them, because they are absorbed by budgets nobody attributes to the role.
The correct model is not complicated. It is the salary plus seven terms, compared against the engagement price plus the internal time the engagement still consumes. Below is each term, what to put in it, and where the number usually hides in your own organisation. The output is not a recommendation — it is your own arithmetic, made complete. If it points at hiring, hire.
Compare
The seven terms that go missing
| Term | Where it hides | How to put a number on it |
|---|---|---|
| Recruitment | Agency fees, job board spend, and your engineers' interview hours | Agency fee as a percentage of first-year salary, plus interview hours × the loaded hourly cost of everyone in the loop |
| Vacancy | Nowhere — this is the term nobody books | Weeks from decision to first commit × what the delayed work is worth per week. Usually the single largest term. |
| Ramp | Absorbed into 'onboarding' and never measured | Weeks to full productivity × salary, plus the senior time spent teaching |
| Employer overhead | Finance, as a blended percentage nobody applies per-role | Your actual statutory and benefits loading on top of base |
| Equipment & licences | IT capex and a dozen per-seat SaaS lines | Machine, amortised, plus every per-seat tool the role needs |
| Management | Your calendar | Hours per week of direction and review × your own loaded rate. Applies to both options — an engagement does not remove it, and any vendor implying it does is overselling |
| Mis-hire risk | Discovered later, attributed to something else | Probability × (months of salary + the second recruitment + the delay). Model it explicitly rather than assuming zero |
Two honest notes on this table. Management does not go away when you engage a firm — it shrinks. Put a real number in that row for both columns. And mis-hire risk is not zero for either option; the difference is how quickly and cheaply it is reversible, which is a contract question, not a cost question.
Run it over the actual horizon
A permanent hire amortises its fixed costs over years; an engagement does not have most of them. So the answer depends heavily on how long the work lasts, and the mistake is to model a permanent hire against a horizon you have not committed to.
Run the model at three, twelve and thirty-six months. If the work is genuinely permanent, hiring usually wins at thirty-six and the model will show you that clearly. If you cannot honestly say the work exists in year two, the twelve-month figure is the one to trust — and modelling it at thirty-six is how organisations end up carrying headcount for work that finished.
The three questions that actually decide it
Cost is the tiebreaker, not the decision. In most real cases the arithmetic lands closer than either side expects, and the decision turns on these instead:
- Is the capability permanent? If this skill is core to what you sell and you will need it every year, hire. Renting a core capability indefinitely is worse than paying for it, whatever the spreadsheet says.
- Do you have someone to direct the work? Both options need an internal owner with real capacity. If nobody has it, you do not want capacity at all — you want someone accountable for the deliverable. Those are different purchases.
- How fast do you need to be wrong? If the work might be cancelled, descoped or invalidated in a quarter, reversibility is worth more than unit cost. A notice period is cheaper than a redundancy.
The comparison a firm should lose
If your work is permanent, you have a technical owner with time to direct it, and you can wait out a hiring cycle without the delay costing you anything — hire. No engagement beats that combination, and a firm that tells you otherwise is selling rather than advising.
The cases where engaging is genuinely better are narrower and more specific than the category likes to admit: a skill you need now and not permanently, a deadline a hiring cycle cannot meet, work you want to be able to stop, or a first version whose shape you expect to change before it stabilises.
That is the honest boundary, and we would rather state it here than discover it with you in month three. If the model points at hiring, we will tell you the same thing on a call.
If you do engage a firm, price the review
One term applies to the engagement column and is almost always set to zero: who reviews the work. Capacity without a quality gate is not output — unreviewed code is a liability that arrives faster the more capacity you buy.
Ask any vendor to name the person who reviews before code reaches your branch, and to say what standard they review against. If the answer is a process rather than a person, put your own review hours back into the model, because they are about to become yours. That question is the reason our engagements are structured the way they are.
Common questions
Is it cheaper to hire an engineer or use an agency?
Over a long enough horizon and for permanent work, hiring is usually cheaper — once the work reliably exists year after year, fixed costs amortise and no engagement competes. Over shorter horizons, or where the work might stop, the comparison frequently reverses because recruitment, vacancy and ramp are paid up front regardless of how long the role lasts. The honest answer is that it depends on the horizon, and that anyone giving you a universal answer is selling one of the two options.
What costs do people forget when budgeting a hire?
Seven, reliably: recruitment, vacancy while the seat is empty, ramp to full productivity, employer overhead, equipment and per-seat licences, ongoing management time, and the probability-weighted cost of a mis-hire. Vacancy is usually the largest and is almost never booked anywhere, because no budget line exists for work that did not happen.
Does engaging a firm remove the management overhead?
No — it reduces it. Someone on your side still sets priorities, answers questions and accepts the work. Any vendor implying you can hand over a problem and stop thinking about it is describing an outcome-owned engagement, which is a different purchase with a different price. Put a real number in that row for both options.
How long does it take to hire an engineer versus start an engagement?
Use your own history rather than an industry benchmark — your last three hires, decision to first commit, including notice periods. That is the number to put in the vacancy row. Engagement start times are quoted by vendors and are worth verifying against a reference, but they are typically shorter, which is precisely why the vacancy term matters so much to the outcome.
When is hiring clearly the right answer?
When the capability is core to what you sell and permanent, when you have a technical owner with capacity to direct and grow the person, and when the hiring delay costs you nothing you care about. If all three hold, hire — the model will agree, and we will say the same on a call.
Why does this page not include salary or rate benchmarks?
Because they would be wrong for you and probably unsourced. Compensation and time-to-hire benchmarks in this category are among the most freely invented numbers in circulation, and a model built on someone else's averages produces a confident answer to the wrong question. Your salary bands, your hiring history and your cost of delay are the only inputs that make this decision correctly.
Want a second opinion on the model?
Bring your numbers to a call and we will run it with you, including the case where the answer is hire. If that is where it lands, you will have a defensible model to take to whoever signs off — which is worth having either way.
Related: What is staff augmentation? · Staff augmentation · AI consulting & implementation