A Databricks contractor and a Databricks engineer on payroll are not two prices for the same thing. One buys you delivery starting next Monday. The other buys you a cheaper run-rate starting in about a quarter. Most hiring decisions get made as though the comparison is day rate versus salary, which is the one framing guaranteed to produce the wrong answer — because a day rate is a gross cost and a salary is a net one, and because the contractor is available now and the permanent hire is available after a notice period you don’t control. Here is the arithmetic, in full, with the break-even month written down.
The two numbers, side by side
Start with the published medians. The median advertised contractor day rate for Databricks skills in the UK is £535 over the six months to 16 October 2026. The median permanent salary advertised for the same skills is £80,000.
To compare them you need the same denominator. A contractor bills days worked. Take 260 weekdays, strip 11 bank holidays, assume zero bench and zero holiday — the most generous possible reading for the contract case — and you get 249 billable days.
| Line | Contractor | Permanent |
|---|---|---|
| Headline rate | £535/day median | £80,000 median salary |
| Statutory employer load | none | 15% employer NI |
| Annual cost, full utilisation | £133,215 | £92,000 |
| Monthly run-rate | £11,101 | £7,667 |
| Monthly difference | — | £3,434 cheaper |
That is the whole story in one row. The contractor costs about £11.1k a month. The same skill set on payroll, loaded with employer National Insurance, costs about £7.7k. There is no month in which the contractor is cheaper. Not month one, not month thirty.
Note that the £92,000 figure is deliberately unkind to the permanent case — UK employer NI applies to earnings above a secondary threshold, not to the full £80,000, so the real statutory load is lower than the flat 15% I’ve applied. I’ve rounded against permanent headcount and permanent headcount still wins.
What “loaded” actually means
Employer NI is the floor, not the ceiling. Pension, private medical, equipment, training budget, holiday and sick pay all sit on top. If you want a defensible upper bound rather than a guess, the US Bureau of Labor Statistics measures the split directly: for full-time private industry workers, benefit costs were 31.5% of total employer compensation costs, with wages and salaries the remaining 68.5%.
Apply that ratio to a £80,000 base and you get the fully-loaded ceiling. Here is the full ladder:
| Load model | Annual cost | Monthly |
|---|---|---|
| Salary only | £80,000 | £6,667 |
| Salary + employer NI at 15% | £92,000 | £7,667 |
| Salary + full benefits load (benefits = 31.5% of total comp) | £116,788 | £9,732 |
| Contractor at £535/day × 249 days | £133,215 | £11,101 |
So the honest permanent band is £7.7k to £9.7k a month depending on how generous your benefits are, against £11.1k for the contractor. The premium ranges from £1,369 to £3,434 a month. Even at the top of the benefits ladder — full private medical, meaningful pension, real training spend — permanent headcount is cheaper per month than the median day rate.
The denominator nobody checks
The same BLS release puts total employer compensation cost per hour worked at $54.00 for full-time private industry workers. The useful part isn’t the currency — it’s the phrase hours worked. Permanent cost per hour worked is already net of holiday, sickness, bank holidays and training days, because those hours are paid but not worked. My 249-day contractor model has none of that slack in it. In practice a contractor takes holiday too, at which point you either pay for cover or lose the delivery. Every adjustment you make for realism widens the gap, never narrows it.
The break-even month
If the run-rate comparison is this lopsided, why does anyone contract? Because the permanent hire does not start on Monday. A search, an offer and a notice period realistically puts a start date three months out, and full productivity a month or two after that. During that window the contract path delivers and the permanent path delivers nothing.
So the real comparison is cumulative cost against cumulative delivery. Three paths:
| Month | Contract throughout | Permanent hired now, starts month 4 | Contract now, convert at month 6 |
|---|---|---|---|
| 3 | £33,303 | £0 | £33,303 |
| 6 | £66,606 | £23,001 | £66,606 |
| 12 | £133,212 | £69,003 | £112,608 |
| 18 | £199,818 | £115,005 | £158,610 |
| 24 | £266,424 | £161,007 | £204,612 |
Read the first three months carefully, because that is where the entire argument for contracting lives. You spend £33,303 to get three months of delivery that the permanent path simply does not produce. Against three months of loaded payroll you’d otherwise be paying — £23,001 — the premium for buying that time is £10,302. That is the price of a quarter, and for a migration with a regulatory deadline or a platform cutover with a hard date, £10,302 is cheap.
After month three, the premium stops buying anything. It is just a higher price for the same output. And it compounds: at £3,434 a month, the cumulative premium crosses £80,000 — one entire median Databricks salary — somewhere in month 23.
Two years of contracting at the median rate buys you one engineer and pays for a second one you never hired.
The break-even that matters, then, is not a cost crossover. It is a decision date. The contract-to-conversion path overtakes contract-throughout the month after you convert, and the longer you leave the conversion, the more of a second salary you’ve burned. Convert at month six and you are £20,604 better off by month twelve than the open-ended contract. Convert at month twelve and you’ve already spent £41,208 of premium that bought you nothing after the first quarter.
Why the market still pays the premium
You might expect a rate this far above loaded payroll to be eroding. It isn’t, and the reason is volume on the demand side. Contract vacancies citing Databricks more than doubled year on year — 1,374 against 621, while the median day rate drifted from £550 to £535. Demand roughly doubled and the rate moved about 2.7% in your favour.
On the permanent side the movement is the other way: the median salary rose 3.23% year on year, from £77,500 to £80,000. Permanent pay is climbing slowly, contract rates are flat to slightly down, and the gap is narrowing — from the contractor’s end.
What that tells you about availability is more useful than what it tells you about price. A doubling of contract postings against a flat rate means supply is keeping up, which means the specialists exist — they are just mostly employed somewhere else, which is precisely the pool most job-board sourcing never reaches. If you want the mechanics of finding them, we’ve written that up separately in how to hire Databricks developers in 2026.
Price every contract with a conversion date attached
If the only defensible reason to contract is to buy delivery time, then an open-ended contract is a pricing error with a calendar attached. Build the conversion in from the start.
Write the conversion date into the statement of work
Not a vague intention to “review in six months” — an actual date, with the permanent requisition opened in parallel on day one. The three-month notice window you’re paying £10,302 to cover is the same window your permanent search needs. Run them concurrently and the premium buys exactly one quarter, which is what it should buy.
Decide upfront whether the contractor is the conversion candidate
Two different contracts. If you want the option to convert the person in the seat, price it in and agree it in writing before they start — retrofitting a conversion conversation onto month eight of a contract is how you end up paying a day rate for another year. If the contractor is explicitly bridge capacity and the permanent hire is someone else, then the handover is a deliverable, not an afterthought, and should appear in the SOW alongside the documentation and runbooks.
Separate capacity from capability
Some of what looks like a hiring problem is a delivery-sequencing problem, and a contractor is the most expensive way to discover that. If the backlog is a one-off migration, contract it and mean it. If the backlog is permanent and growing, you are using contract rates to fund a structural gap, which the table above prices at £3,434 a month in avoidable cost. We’ve unpacked that distinction in capacity versus adoption.
What the arithmetic changes about how you hire
The practical consequence is that the permanent requisition should almost always open first, and the contract should be sized to cover the gap it leaves. That inverts how most teams do it — contract now because it’s faster, think about permanent later, and discover in month fourteen that “later” never arrived.
It also raises the stakes on vetting, in both directions. A contractor you’re paying £11,101 a month needs to be productive in week one, not week five, because you are explicitly buying time and a slow start spends the thing you bought. A permanent hire at £7,667 a month loaded needs to survive two years of comparison against that number. Neither of those is a CV-screening exercise. If you want a repeatable technical bar, vetting a Databricks engineer in 90 minutes sets out the structure.
And price your own delay honestly. Every month the permanent requisition sits unopened while a contractor delivers is £3,434 of premium you’ve chosen to pay. That is a reasonable choice for a quarter and an expensive habit over two years. Put the conversion date in the contract, open the requisition the same week, and the premium does what it is supposed to do — buys a quarter of delivery and then stops.
If you’re sizing a Databricks build and want to see who is actually available against your stack before you commit to either path, the data engineering network is free to search.