Headcount

What an engineering hire really costs over five years

By Assaf Schwartz · 10 September 2026 · 9 min read

An offer letter states one number and it is the smallest one in the decision. The commitment you are actually approving runs for years, rises every January whether or not anything else does, drags a set of licences behind it, and is far harder to reverse than it was to make. On the simulator's default company, one additional engineer costs $831,424 over 60 months — 5.7 times the salary on the letter.

What the model actually charges

The Fully-loaded salary control is annual, and the engine charges one twelfth of it every month, multiplied by a wage index that compounds:

payroll = headcount × salary ÷ 12 × (1 + salary inflation)^(month ÷ 12)

The word doing the work in that field label is fully-loaded. It is not base salary. It is base plus employer taxes, benefits, insurance, equipment, software that is not seat-priced, and the share of office and admin cost the person consumes. Depending on jurisdiction that is typically 1.25 to 1.4 times base pay, so an engineer on a $111,538 base is roughly the $145,000 default here. Entering base salary and calling it loaded is the most common way this model gets quietly wrong by a quarter.

The second half of the formula is the part people forget. At the default Salary inflation of 4.5%, the ninth engineer costs $12,128 in month one and $15,058 in month 60 — the same person, the same job, 24.2% more expensive. Nominal monthly cost at the offer figure is $12,083; you never pay that amount for long.

The seats a hire drags in

Software is bought for engineers and paid for by everyone adjacent to them: the designer in the same squad, the product manager, the analyst, the support lead who needs read access. The engine encodes that with a multiplier:

seats = round(engineering headcount × 1.8)

Going from 8 to 9 engineers takes seats from 14 to 16 — 2 extra seats at the default SaaS cost / seat / mo of $145, so $290 a month before price drift. Over the full projection, with Cost inflation at 3.4% a year, that hire adds $18,965 of licence spend.

On its own that is a rounding error next to payroll, and it is meant to be. The point is that it is not zero and it is not one seat. Every serious tool in the stack — observability, CI, the design suite, the analytics platform, the incident tooling — applies its own version of the same multiplier, and the one you are modelling here is only the platform you named in the strategy inputs. If the real per-head tooling bill is three or four times the single licence you are testing, raise the seat price rather than pretend the rest does not exist.

The five-year total

Put the two together for one hire on the default company: $812,458 of payroll and $18,965 of licences, $831,424 in total, against an offer letter reading $145,000.

Year five costs 19.1% more than year one. Set Salary inflation to zero and the payroll component of the same hire falls from $812,458 to $725,000: the drift alone is $87,458, or 12.1% of the flat-salary figure. Every plan built on today's comp is understating headcount by roughly that much, and the error grows with the horizon.

Eight, nine, twelve

Costs in isolation are not the decision. What matters is what the cash line does, because payroll competes with the Re-investment rate for the same operating profit. Here is the default company at three headcounts, plus one row with salary drift switched off.

Read the third row first. Four extra engineers do not make this company slower; they make it insolvent. Cash goes negative in month 12, break-even never arrives, and the projection ends at -$2M. That is the same product, the same market and the same funnel as the first row, which finishes with $1.3M in the bank.

The single-hire row is subtler and more useful. Adding one engineer costs $831,424 directly, but ending cash falls by $868,235 — $36,811 more than the hire itself. The gap is compounding: lower operating profit means less money routed into paid acquisition, which means less MRR, which means less gross profit next month. Ending ARR is $316.9K lower for the same reason. Break-even moves from month 24 to month 37, and the company stops repaying its cumulative losses inside the horizon at all.

The ratchet

Every other line in this model is reversible on a reasonable timescale. Cloud spend responds to an efficiency sprint. Licences lapse at renewal. Contractor hours stop when the purchase order does. Acquisition spend can be halted this afternoon.

Headcount does not behave that way. Reversing a hire costs notice, severance, legal review and, in much of Europe, a consultation process measured in months. It costs the institutional knowledge that left with them and the productivity of everyone who spent a fortnight worrying about who was next. In practice a team that grew to twelve does not go back to eight because the model asked it to; it goes back to eight after two quarters of denial, at a cost nobody modelled.

The asymmetry is the argument. The upside of a hire is uncertain and arrives slowly. The downside is contractual and starts on the first of the month. Treat the Engineering headcount control as the one input with a one-way door attached, and give it the scrutiny you would give a capital expenditure of $831,424, because that is what it is.

Does the hire pay for itself?

There is a defensible answer to this and it is not a feeling about velocity. In month one the ninth engineer costs $12,128. At 78% gross margin the revenue required to cover that is:

MRR needed = loaded monthly cost ÷ gross margin

which is $15,548 of new recurring revenue a month, or $186,581 of ARR — about 37 accounts at the default ARPA of $420. That is the hurdle. If you cannot describe, specifically, how this person's work produces that much revenue or removes that much cost within a year, you are not making an investment, you are increasing your fixed cost base.

• Revenue-side hires should be attached to a named piece of demand: a segment you are turning away, a feature blocking deals you can count, a plan tier nobody can ship.
• Cost-side hires should be attached to a line in this model that goes down. Infrastructure efficiency work is the cleanest case, because the saving is measurable and recurring.
• Margin-side hires pay off through the Gross margin control rather than the revenue line — support tooling, onboarding automation, per-customer cost reduction. The arithmetic is in SaaS gross margin explained.
• Everything else is a bet on capacity, which is fine, but should be financed with an explicit runway decision rather than smuggled in as an obvious necessity.

The comparison worth running before any of this is the one against not hiring at all. The same $831,424 routed through the Re-investment rate buys customers directly; routed to a vendor it buys a capability without the ratchet. That last option is the subject of the real cost of building software in-house, and the honest version of a hiring case includes it.

What the model leaves out

The $831,424 figure is the floor, not the ceiling. Three real costs sit outside the engine, and all three point the same way:

• Recruiting. Agency fees run 15 to 25% of first-year base; even in-house sourcing costs referral bonuses and a meaningful slice of engineering interview time. Budget 20% of the loaded salary as a one-time hit — around $29,000 here.
• Ramp. A senior engineer in an unfamiliar codebase is rarely at full output before three to six months, and is consuming somebody else's output while getting there. Three months at half productivity is another $18,192 of cost with no work attached.
• Management overhead. Somewhere between the sixth and eighth engineer, coordination stops being free. Every additional head consumes a fraction of a manager and adds to the communication load of everyone else — the reason a team of twelve is not one and a half times a team of eight.

The practical adjustment is to raise Fully-loaded salary rather than to bolt on a separate line. Adding 10 to 15% to that control approximates recruiting amortised over an expected tenure plus the ongoing management drag, and it keeps the compounding intact. If you want to model ramp explicitly, run the projection at the lower headcount for the first two quarters and compare it to the run at the higher one — the difference in ending cash is the true price of the delay.

None of this argues against hiring. It argues for knowing that the question on the table is not $145,000 a year. It is $831,424 of committed spend, $868,235 less cash at the end of five years, and 1y 1mo of delayed break-even, in exchange for capacity you should be able to describe in revenue.

Educational content, not financial advice. Figures here are illustrative and exclude taxes, financing and one-off items. Model your own numbers in the simulator and check them with a qualified accountant before acting.