Open courses or in-house training? A decision guide for a development budget

Open courses are the right buy when one person needs a new skill or a credential. In-house training wins the moment you want to change how a population behaves: cost per head drops and the context is yours. Neither format produces transfer on its own; only a capability system, with reinforcement and measurement, does.

This article is for whoever holds the development budget and has to allocate it: HR, L&D, a manager with a training line, or a founder staring at two very different quotes. We are not going to sell you one format. We will show you what each is genuinely good for, what neither can do, and how to tell (before you sign) what you have actually bought.

TL;DR

• A seat and a cohort are different products. An open course is priced per participant and grows one person. An in-house program is priced per group and changes how a group works. The invoices look alike; the questions differ.
• An open course is the right buy whenever the return is individual: someone stepping into a new role, someone who needs out of their own company's bubble, self-development a person asked for, or two people testing a provider.
• Two people out of forty do not move a norm, so an open course never reaches critical mass. Examples in a mixed room have to stay generic. And three people sent to three courses come back with three vocabularies.
• Do the break-even before you request a quote. Divide the cohort price by the seat price: that is how many open seats equal one in-house program. In the quotes we see, the number is lower than buyers expect.
• Five questions settle it: how many people must behave differently, skill or norm, is there a manager who can reinforce weekly, what you will measure at ninety days, and whether you are short on content or on architecture.
• The open versus in-house debate is really a debate about where the chairs are. Both options assume the product is the event. The third option moves the money from the event into the architecture around it.

What are you actually buying when you buy training?

One of two products sold under the same word. An open course sells a seat: one person's access to content and a trainer's time, priced per participant. An in-house program sells a cohort: design and delivery for your people on your cases, priced per group. They answer different questions, and most misallocated budget answers the wrong one.

Two completely different products are sold under one word.

An open course sells you a seat. You pay for one person's access to structured content and a trainer's time, in a room where everyone else comes from other companies. The unit of cost is the seat.

An in-house program sells you a cohort. You pay for the design and delivery of a shared experience for your people, on your cases, in your calendar. The unit of cost is the group.

The invoices look similar; the outcomes are not comparable, because they answer different questions. The first is "how does this person grow?" The second is "how does this team work differently?" Most misallocated training budget is an answer to the first question, bought by someone who actually had the second problem.

When is an open course genuinely the right call?

In five situations, and they share one thing: the return is individual. One person needs a specific skill. You need an externally recognized credential for an audit, a tender or a client. Someone benefits from a mixed room and an outside benchmark. It is self-development rather than organizational change. Or you are testing a provider cheaply before committing cohort money.

Open courses get a bad reputation in capability-development circles. Unfairly. They are exactly the right buy in a handful of situations:

• One person needs a specific skill. Someone steps into a new role or inherits an area nobody in the company holds. You do not build a program for one person.
• You need an externally recognized credential. When the value is the signature on the certificate (an audit, a tender, a client requirement), an accredited external provider is the only route.
• Someone needs out of the bubble. The real value of a mixed room is that your person spends two days next to someone from another industry solving the same problem differently. That informal benchmark is the one thing an in-house program cannot reproduce: a good facilitator brings counter-examples, but not the peer from another industry.
• It is self-development, not organizational change. Someone wants to grow and asks for support. An open course is a fast, cheap answer and a good retention signal.
• You are testing a provider. Send two people to a public session before you put cohort money on the table. Two seats is a cheap way to find out.

All of these share one thing: the return is individual. An open course is a good product with a clear beneficiary: the person. The problem starts when you buy it expecting a collective return.

What can't an open course do for an organization?

Six things. It never reaches critical mass, because two people out of forty do not move a norm. It cannot use your context, so examples stay generic. It produces three vocabularies rather than shared language. It has no Monday morning. It gives you attendance and satisfaction to measure, not transfer. And it socially exposes whoever applies it.

This is the heart of the decision, so we will put it without diplomacy.

It never reaches critical mass. Two people out of forty do not move a norm. Behavior in a team is not the sum of individual intentions; it is what is acceptable in Monday's meeting. The person who comes back with a new framework walks into an unchanged system, and that system corrects the deviation rather than adopting it.

It cannot use your context. In a room with eleven companies, the examples have to be generic or they work for nobody. Translating them into your own reality is homework, and homework rarely gets done.

It does not produce shared language. Three people sent to three courses come back with three vocabularies. They do not compound. An in-house program, however imperfect, at least gives a team a common set of words for naming what it does badly.

It has no Monday morning. The course ends Friday at 5pm and usually nothing follows: no practice, no routine, no conversation with a manager. Ebbinghaus measured the forgetting curve on nonsense syllables in 1885, but its shape has been replicated many times since: without retrieval, retention drops steeply in the first days. Training-transfer research from Kirkpatrick onward, and particularly Robert Brinkerhoff's work, keeps reaching the same conclusion: content delivered in a classroom rarely changes behavior at work, and the cause is not content quality. It is the absence of architecture around the content.

It gives you nothing to measure. You get an attendance list and a satisfaction score. Satisfaction is not transfer: someone can rate the course 10 out of 10 and return to exactly the same behavior.

It socially exposes the participant. The person who applies something the team never learned pays a price: they look like they are showing off. Wendy Wood's research on habits explains why: behavior is triggered by cues in the environment, not by intention. Change one person's intention, leave the environment intact, and the environment wins.

The detailed mechanics of that failure are in our article on why training fails. Here, only the budget consequence matters.

What do you gain, and what do you risk, with an in-house program?

You gain critical mass, your context, shared language and cover for whoever applies it, plus the option of involving the manager. You risk four things few providers name: an echo chamber where nobody challenges shared assumptions, attendance bought by obligation rather than interest, a program as disconnected from the work as any course, and fixed cost that punishes small numbers.

An in-house program solves four of the problems above: critical mass, context, shared language, and social exposure (if the whole team was in the room, nobody looks like they are showing off). It also opens the door to involving the manager. And you gain control: the calendar, confidentiality (you can discuss real cases, with names and numbers), and a topic that is yours rather than the market average.

The risks are real, and few providers name them:

The echo chamber. If everyone in the room shares the same assumptions, nobody challenges them. A good program deliberately imports counter-examples from outside and uses a facilitator who is allowed to contradict the group. A weak one confirms the existing culture and calls it development.

Mandatory attendance. An open course is chosen; an in-house program is often assigned. Obligation buys attendance, not learning. If people do not know why they are in the room, you have bought eight hours of polite silence.

Internal does not mean transferable. An in-house program can be just as disconnected from the work as an open course, only more expensive. Using your own examples does not mean anything changes on Monday.

Fixed cost punishes small numbers. Below a certain headcount, an in-house program is more expensive per person. That is not a moral dilemma, it is arithmetic.

What does the real cost look like: per seat vs per cohort?

An open course scales linearly: seat price times people, plus travel. An in-house program is a cohort price, largely flat between roughly eight and sixteen participants, so cost per person falls with each one you add. Divide the cohort price by the seat price and you have your break-even. Then add the cost no invoice shows: zero transfer.

An open course costs seat price × number of people. It scales linearly, plus travel and accommodation if the session is in another city. An in-house program costs a cohort price, largely unchanged between roughly 8 and 16 participants: cost per person falls with every participant you add.

The break-even takes two minutes to work out, and almost nobody works it out: divide the cohort price by the seat price. The number you get is how many open seats you have to buy before you have already spent the equivalent of an in-house program. In the quotes we see in the market, that number is lower than buyers expect.

Then there is the cost no invoice shows: the cost of zero transfer. Working days pulled out of production get paid for in either format, once as salary and once as work not done. If those days were paid for and behavior did not change, you bought goodwill at full price, and it is the one line that appears nowhere in the budget.

The third option most buyers never ask for

The open-versus-in-house debate is, at bottom, a debate about where the chairs are. Both options assume the product is the event. Our thesis, after 600+ programs and 20,000+ participants since 2008, is that behavior change does not fail because of weak content. It fails because of architecture: nothing exists between sessions, nobody reinforces, nothing gets measured.

A capability system moves the money from the event into the architecture. It has five components, and none is optional:

• Diagnostic. Before content, you find out where the actual constraint sits. Not every team has the same problem: sometimes Clarity is missing (people do not know what matters), sometimes Accountability (commitments never turn into follow-through), sometimes Dialogue (nobody says the useful thing), sometimes Learning (the same problems keep returning).
• Facilitated labs. Practice on the cluster where the constraint sits (Priority Translation, Commitment Discipline, Feedback Exchange, Reflection & Sensemaking) rather than a tour of every soft-skill topic.
• AI-personalized practice between sessions. This is where you work against the forgetting curve: short retrieval on each person's real case, in the weeks between labs.
• Manager reinforcement, roughly 15–20 minutes a week, and the cheapest of the five components to run. Gallup estimates that the direct manager explains around 70% of the variance in team-level engagement. That figure is about engagement, not transfer; we have simply never seen a cohort rise higher than its managers.
• Measurement and review. Observable behavior at 30, 60 and 90 days, not a satisfaction score at the end of day two.

None of this is an alternative to training. It is what turns training into capability. You can see the whole system in Capability OS.

Five questions to answer before you request a quote

1. How many people have to behave differently for this to matter? One to three: open course. A team or a management layer: cohort.
2. Do you want a skill or a norm? A skill belongs to a person; a norm is what is acceptable in the team. Norms cannot be bought one seat at a time.
3. Is there a manager who can reinforce weekly? If not, whatever you buy stays an event. And if the managers are the constraint, start there, for example with Manager OS.
4. What will you measure at 90 days? If the only answer is participant satisfaction, you bought a good day, not development.
5. Are you short on content or on architecture? If your people already know, in theory, what they should be doing and are not doing it, more content will not fix anything.

FAQ

So are open courses a waste of money? No. They are a good product for an individual problem: one person, one skill, one credential, one dose of outside perspective. They become a waste only when bought expecting a collective result: sending four people to four courses and expecting the team to work differently afterwards.

At what headcount does in-house become worth it? Divide the cohort price by the seat price. If the number of people you would send anyway exceeds that result, in-house is already cheaper, and it also buys you context, shared language, and somewhere for reinforcement to happen.

Can we start with a pilot? Yes, and it is usually the right move: one team, a diagnostic at the start, measurement at 90 days. The condition is that the pilot includes the reinforcement, not just the lab. Otherwise you are testing an amputated version and drawing the wrong conclusion.

Is it cheaper if our internal trainers deliver it? Long term, yes, and it is a good direction: internal trainers know the context and can reinforce continuously. What they usually lack is the architecture: the diagnostic, the rubrics, the practice sequence, the reinforcement materials. That is what the licensing model exists for: you keep the delivery, the system comes from us.

How do I compare two providers proposing apparently identical content? Do not compare agendas; they are nearly identical across the market. Ask three things: what happens between sessions, who reinforces afterwards, and what gets measured at 90 days. The difference between providers is almost never in the slides.

We have leftover budget at year end. Should we burn it on open seats? It is the most common decision and the weakest one. If the budget has to be spent, buy next year's diagnostic with it. Then you enter January knowing where the real constraint is, instead of entering with four certificates and no change.

Do certifications matter? They matter exactly as much as someone outside requires them: a client, an audit, a tender, a formal career gate. As evidence of changed behavior they are close to worthless: a certificate attests attendance and, at best, knowledge. Not practice.

Sources

• Ebbinghaus, H. (1885). Über das Gedächtnis (the forgetting curve and the effect of spaced retrieval).
• Kirkpatrick, D. (1959, later revised). The four levels of training evaluation: reaction, learning, behavior, results.
• Brinkerhoff, R. (2003, 2006). The Success Case Method and research on training transfer.
• Wood, W. (2019). Good Habits, Bad Habits (behavior as cue-driven rather than intention-driven).
• Gallup. State of the Global Workplace and State of the American Manager (the estimate that the direct manager explains roughly 70% of team-level engagement variance).
• Azimut internal program data: 600+ programs delivered, 20,000+ participants, 9.27/10 average trainer score (since 2008).

What you can do next

If you have an open-course quote and an in-house quote on the desk, the useful question is not which is cheaper. It is how many people have to work differently, and whether anyone will reinforce it afterwards.

Start with the Capability Snapshot, a free five-minute self-assessment that shows which of the four domains holds your constraint. For the full team-level picture there is the Capability Diagnostic. See how the system works in Capability OS, browse our programs, or, if budget allocation is your job, the page for HR and L&D. When you want to talk through your specific case, let's talk.

By Szilárd Kacsó