14 April 20267 min read
Nobody selects perfectly. Interviews are a weak signal, references are a weaker one, and pre-vetting improves the odds without changing the fact that some proportion of matches will be wrong. The interesting question is not how to avoid ever being wrong. It is what being wrong costs, and how fast you find out.
On that question the two routes are not remotely symmetric, and the asymmetry is the strongest cost argument for outsourcing that most vendors never make properly, because they reach for the rate card instead.
What a wrong permanent hire costs
- The salary and loaded cost of the months before anyone admits the problem, which is usually longer than it should be.
- The manager's time: the extra supervision, the documentation, the difficult conversations, the process.
- Notice periods and, depending on jurisdiction, severance and a formal procedure with legal costs attached.
- Recruitment, paid a second time, plus the interviewing hours taken from the team a second time.
- The ramp of the replacement, from zero, months later.
- The work that did not happen throughout, which is where the role started.
Add these and the cost of a wrong hire is a multiple of the person's salary, which is why the expected cost of hiring is always higher than its nominal cost. That is not an argument against hiring people. It is an argument for pricing the risk honestly on the in-house side of a comparison rather than pretending the risk sits only on the vendor side.
The part nobody prices: not deciding
The largest component above is not on the list as a line of its own, because it hides inside the first item. It is the delay between knowing and acting.
Permanent employment makes that delay long by design and by decency. There is a probation mechanism, then there is a process, and in between there is the entirely human hope that another month of coaching will turn it around. Sometimes it does. When it does not, every month of hoping was paid at full loaded cost, and the roadmap absorbed it silently.
The expensive part of a bad hire is rarely the exit. It is the six months of hoping that preceded it.
What a wrong placement costs
A placement that is not working ends under a contract rather than under employment law. Typically that means a notice period measured in weeks, a replacement obligation on the provider, and a new person starting who is supported by whatever documentation and handover the provider arranges.
The costs that remain are real and worth naming. You paid for the ramp of someone who is leaving. Some context leaves with them. Your team spends time onboarding a replacement, and the management load resets to its front-loaded level for a few weeks. If the contract is written badly, you may also be paying notice on someone you no longer want.
So the honest comparison is not free versus expensive. It is weeks versus months, contract terms versus employment law, and a provider with an obligation to fix it versus a process you run yourself.
The clauses that determine the real number
Everything above is set by four contract terms, and they are far more worth negotiating than a small movement in the rate.
- The trial or initial period: how long, and on what notice either side can end it.
- The replacement obligation: whether the provider must supply a replacement, within what timeframe, and at whose cost.
- Who pays for the replacement's ramp: the provider, you, or split. This is the clause most often left silent and it is the one with money in it.
- Ordinary notice: symmetric, long enough for a real knowledge transfer, short enough that a wrong decision is correctable.
A provider that resists all four is selling you a placement rather than an outcome. A provider that accepts all four is exposed to its own selection quality, which is exactly where the exposure should sit, and it is a reasonable proxy for how seriously the pre-vetting is taken.
Finding out fast is a design decision
Whichever route you take, the cost is dominated by how long it takes to know. That is controllable, and it is controlled at the start of the engagement rather than at the end.
Give the first real task a definition of done that is checkable. Review the first piece of work yourself, closely, however senior you are. Watch how questions get asked in the first fortnight: good people ask early, precisely, and once. Compare the second week to the first, because the direction of travel matters more than the starting point when someone is still learning the domain.
None of these tell you about capability in week one. Together they tell you a great deal by week three, which is early enough for the answer to be cheap.
What this actually means for the decision
It does not mean outsourcing is safer than hiring. It means the two carry different failure costs, and the difference belongs in the comparison alongside the rate. Work that is genuinely differentiating is still worth hiring for at full local cost and full correction risk, because the upside of the right permanent person is enormous.
For everything else, being able to be wrong cheaply, and to find out in weeks, is worth a real amount of money. Put a number against it when you build the comparison, even a rough one, because setting it to zero is the same mistake as leaving recruitment fees out of the in-house column.