The decision gets made in a meeting where someone says training is cheaper and someone else says it takes too long, and both are guessing.
It is a modellable question. Not precisely, because several inputs are estimates, but well enough that the answer is usually clear once the real terms are on the page rather than the two that are easy to look up.
Why the usual comparison is wrong
The comparison most organisations run is the salary premium for an experienced hire against the price of a training program.
Those are the two most visible numbers and the two least decisive. The salary premium is recurring and often smaller than expected. The course fee is frequently the smallest line in the training column.
Everything that actually separates the two options is missing: the time either path takes to reach full productivity, the cost of the role sitting empty meanwhile, the probability of failure on each side, and the value of what an internal person already knows.
The two numbers that are easiest to obtain are the two that matter least. That is not a coincidence, it is why the decision gets made badly.
The full cost of a hire
Recruitment cost. Agency fees for specialist technical roles commonly run at a substantial share of first-year salary. Internal recruiting is cheaper in cash and not free: sourcing, screening, and the hours of engineering time spent interviewing, which for a senior technical role can run to dozens of hours across a panel.
Time to fill. Specialist roles in the domains where shortages are real routinely take months. That is months of the vacancy cost below.
Ramp to full productivity. This is the largest hidden line. An experienced external hire arrives with the skill and without the context: your systems, your processes, your product, your customers, who to ask. Estimates vary by role and seniority, but for a technical position it is commonly months rather than weeks before output matches the level the person is capable of.
Onboarding overhead. The colleagues doing the onboarding are less productive while they do it.
Early attrition risk. A meaningful share of hires leave within the first year, and the earlier the exit the higher the proportion of the investment that is written off entirely. Any honest model applies a probability here rather than assuming the hire stays.
The salary premium, recurring, and worth noting that it usually resets the internal pay band for that role too.
The full cost of training someone internal
Program cost, which is the visible number and usually the smallest.
Productive time. The real expense. If training takes a person out of their job, the cost is their loaded salary for that time plus whatever their output was worth. This is why format matters so much to the economics: training delivered in short daily blocks around the working day has a fundamentally different cost profile from a two-week residential course, even at the same list price.
Backfill. If the person is taken out of their existing role for a substantial period, that role needs covering.
The failure rate. Not everyone completes, and not everyone who completes reaches the standard. This is real and needs a number in the model rather than optimism. It is also the line most affected by program design, which is the subject of a companion article on completion.
Time to competence, which for a genuine step up in capability is not short.
The replacement chain. Promoting someone internally opens their previous role. Sometimes that role is far easier to fill than the one you started with, which is the whole point. Sometimes it is not, and the chain needs following to the end before the option looks as good as it did.
The cost of the vacancy itself, which sits on both sides
This is the number that changes decisions, and it is usually absent.
An unfilled role has a cost per month: output not produced, work redistributed onto colleagues at overtime or at the expense of their own, projects delayed, and in some settings revenue or capacity directly forgone.
In construction, the AGC and NCCER 2025 workforce survey found 45 percent of firms already running projects late because of workforce shortages. A late project has a cost that dwarfs most training budgets.
The vacancy cost applies to whichever path is slower, and it accrues during the deliberation. A decision deferred by two months is not a free option, it is two months of vacancy cost purchased in exchange for more certainty.
Why the format changes the arithmetic
The single largest cost in the training column is productive time, not program price. That makes lesson length an economic variable rather than a pedagogical preference.
Astra Trainer runs on five-minute lessons that fit a shift change or a commute, with practice in every one. A program delivered that way does not require pulling people off the line or backfilling their role, which removes the two largest lines from the training side of the comparison. The engine underneath is the one from the consumer app, where fifty thousand learners keep streaks running without anyone's manager asking twice, applied to industrial training tracks across ten domains. You can see how programs are structured here.
A worked comparison
A manufacturing firm needs an automation engineer who can work with industrial control systems and Industry 4.0 tooling. The site has two maintenance technicians with deep knowledge of these specific machines and no formal automation training.
The hiring path. Specialist automation engineers are in demand, so the role takes several months to fill at a salary premium over the internal band. Add recruitment cost, panel time, and a ramp period during which the new engineer learns a plant they have never seen. Apply a probability that the hire does not work out. Add the vacancy cost for the months before they start and the partial-productivity months after.
The training path. One of the technicians enters a structured automation track. Program cost is modest. Delivered in short daily blocks, they stay in their job throughout, so productive time lost is small and backfill is unnecessary. Time to competence is months, not weeks, and there is a real chance they do not reach the required standard. Set against that: they already know every machine on that floor, which is the part the external hire would have spent half a year acquiring.
What usually decides it. Not the totals, which are often closer than either advocate expects. It is the size of the gap. If the technician needs one structured step, training wins comfortably, because the context they hold is expensive and the distance is short. If the role needs a chip designer and you have a maintenance technician, no program bridges that in a useful timeframe and the hire is correct.
The other decider is whether the requirement is one person or twenty. Hiring twenty specialist engineers into a shortage market is a different proposition from hiring one, and at that volume the pipeline case gets much stronger.
Where hiring clearly wins
Four situations, stated plainly, because a training vendor that claims training always wins is not worth reading.
The capability is genuinely absent and the gap is large. Training closes a step, not a chasm.
The timeline is short and non-negotiable. A regulatory deadline or a plant commissioning date will not wait for a competence curve.
You need someone who has done it before. Judgement built from having seen the failure modes is not trainable at speed. For a first-of-a-kind project, experience is the thing you are buying.
The role requires a credential you cannot confer. This one is absolute. Professional engineering licensure, clinical qualifications, nuclear operator certification and similar are regulated. Training can build competence toward them. It does not substitute for the credential, and no honest provider will tell you otherwise.
On regulated roles. Several of the domains where shortages are most acute carry statutory competency requirements: professional engineering, nursing and clinical practice, nuclear operations, aviation maintenance. Structured training can prepare people for those pathways and can maintain and extend competence within them. It cannot grant the licence, and any workforce plan that treats a training program as equivalent to certification is building on a mistake that will surface at the worst moment.
The attrition argument, and what the evidence really says
The standard objection: why train people who will leave?
The standard rebuttal is a quotation about what happens if you do not train them and they stay. It is a good line and it is not evidence.
Here is the more careful version. Training does raise a person's market value, and that is not a trick of framing, it is the entire point and the employee knows it. Some of those people will leave.
What survey evidence fairly consistently shows is that development opportunity is among the factors employees cite when they stay and among those they cite when they leave for lack of it. That is correlational and self-reported, and it does not establish that a given program reduces your attrition.
The more defensible framing for a business case is this: you are comparing attrition among trained employees against attrition among external hires, which is itself substantial in the first year, and against the certainty that an untrained workforce cannot do the work at all. Absolute retention is the wrong comparison. Relative retention against the alternative is the right one.
And the practical lever is not the training decision, it is what follows it. Training someone into a higher capability band and leaving them in the same role at the same pay is a reliable way to hand a competitor a qualified candidate. The program and the progression have to move together.
What to take from this
Salary premium against course fee is the wrong comparison and it is the one almost everyone runs.
Put all of it in: recruitment, time to fill, ramp, early attrition risk on one side; program cost, productive time, backfill, failure rate and the replacement chain on the other. Then put the vacancy cost on top of whichever path is slower.
Training format is an economic variable. Time out of the job is the largest line in that column, so a format that does not remove people from their work changes the answer.
The gap size decides it more than the totals. One step favours training strongly because of the context an internal person already holds. A chasm favours hiring.
And if you train people into a higher band and do not move their role or their pay, the attrition objection becomes self-fulfilling.
Is training cheaper than hiring?
Often, but not because the program costs less than the salary premium. It is because an internal candidate already holds the domain context an external hire spends months acquiring, and because the vacancy cost is shorter. Where the capability gap is large, hiring is cheaper despite appearances.
What is the biggest cost people leave out?
The vacancy itself. An unfilled role has a monthly cost that frequently exceeds both options, and it accrues while the decision is being deliberated.
Why train people who will just leave?
The right comparison is not against perfect retention, it is against first-year attrition among external hires and against a workforce that cannot do the work. The practical lever is pairing training with progression, since training someone into a higher band and leaving their role unchanged does invite the outcome you feared.
Can training replace a professional licence?
No. In regulated fields including professional engineering, clinical practice and nuclear operations, training builds competence toward a credential and does not confer it. Any provider suggesting otherwise should be treated with suspicion.
How do I scope a program that survives contact with the operation?
Start from the roles you cannot fill and work backward. Astra Trainer builds programs with partners across ten industrial domains, sequenced with the partner's own specialists, and you can see the catalog here.