How does simulation-based learning compare to the case method, and where does each fail?

It is Sunday night. A marketing professor has a syllabus to file by Wednesday. They have read the dean’s email about “AI integration.” They have a class of eighty undergraduates, half of whom believe ChatGPT is a search engine and the other half who think Google Ads is run by a single person at Google. The professor has, by their own count, seven hours of weekend left, and they are deciding between two pedagogies. Cases or simulations.

This is the choice in front of every business school marketing professor in 2026, and the literature pretending to settle it is mostly written by people who have not stood in front of an undergraduate cohort in twenty years.

Both pedagogies are good. Both are flawed in specific, knowable ways. The thesis of this piece is straightforward. Each fails in opposite directions, and the failures explain why neither alone is sufficient for teaching marketing in an era when the tools are changing every six months and the buyer at graduation will be, increasingly, an AI agent.

A simulation that allows clickthrough success is not a simulation. It is a pdf with an academic veneer.

What is the case method, and what does it actually do?

The case method, in its modern form, was systematised at Harvard Business School in the 1920s and codified by C. Roland Christensen into the discipline most marketing professors will recognise. A thirty-page narrative, a set of exhibits, a class facilitated by Socratic questioning, the student-as-protagonist asked, “what would you do.”

The case method is excellent at three things. It teaches students to reason from incomplete information. It teaches them to defend a position out loud, in front of peers, under cross-examination. And it gives them a substantial library of decision-making analogies. A marketing professional who has worked through three hundred cases over an MBA has, in some real sense, been in the room for three hundred decisions.

These are not small things. The case method has produced two generations of senior marketers who can construct an argument and hold a meeting. That is a serious achievement, and it explains why the method has survived a century of pedagogical fashion.

Where does the case method fail?

Three places, mostly.

One. There is no consequence. A student in a case discussion can argue brilliantly for the wrong answer and pay no price. The wrong answer was, after all, a hypothetical. The case wraps up. The next case starts. A student who never genuinely felt the cost of a poor decision graduates with the rhetorical equipment of a strategist and the operational instincts of someone who has never had to run anything.

Two. The cases age badly. A case written in 2012 about Coca-Cola’s social media strategy is a museum piece. Some marketing cases age into period dramas: their arguments are intact, but their context is unrecoverable. The case method’s heaviest cost is its reading rate, which is roughly two cases per hour of class time and two hours of preparation per case. Refreshing a curriculum at speed is hard when the unit of work is so large.

Three. The case method assumes the student wants to be there. It is brutally effective for the engaged thirty per cent, and brutally ineffective for the seventy per cent who would rather be anywhere else. A case discussion in a 90-person undergraduate elective is, statistically, eight people talking and 82 people surviving.

I should be clear. I do not think the case method is dead. It is not. Used well, with the right cohort and the right professor, it remains the strongest single pedagogy for teaching strategic thinking in marketing. But the conditions for “used well” have narrowed.

What is simulation-based learning, and what does it actually do?

A simulation is a piece of software that asks the student to make decisions and shows them the consequences. The decisions can be tactical (set a Google Ads bid, choose a creative variant) or strategic (allocate a quarterly marketing budget across channels). The consequences are computed against a model that is meant to approximate what would happen in the world. (A definitional primer sits here.)

The good simulations do three things the case method cannot. They put the student inside the decision, not outside it. They allow consequences to compound: a poor week-three decision produces a worse week-five outcome, and the student feels the slippage. And they scale. A simulation that runs for one student runs for a thousand at the same fidelity.

The cohort attainment gap that opens up between simulation cohorts and lecture cohorts is sharper than most marketing professors expect, and it has been replicated across enough disciplines to be taken seriously. (I have written about how this shows up in the analytical thinking data.)

The good simulations also let students fail in private. This matters more than it sounds. The case method asks students to be wrong in front of a room. The simulation asks students to be wrong in front of a screen, learn from it, and come back to the next decision with new information. For the seven students in the room with strong opinions and the eighty-three without, this is a fundamentally different proposition.

Where does simulation-based learning fail?

Two places, severely.

One. The cheap simulations allow clickthrough success. This is the failure mode that has given simulation-based learning a bad reputation in some marketing departments, and it deserves more honesty than it gets. A simulation in which a student can press “next” repeatedly and earn a passing grade is not a simulation. Some of the largest commercial simulations on the market sit close to this line. (I have written elsewhere about why even “perfect” AI-driven simulations can miss the pedagogical point.) The students who buy these simulations learn to game them, not to learn from them. The professor who assigns them learns nothing about the student’s marketing instincts.

The diagnostic question is simple. Can a student get a high score in this simulation by setting their budget to zero and clicking through? If yes, the simulation is broken. If no, the simulation has a chance.

Two. The structural fidelity of the underlying model is mostly invisible to the buyer. A marketing professor evaluating a simulation cannot, in a thirty-minute demo, tell whether the model is a serious approximation of how Google Ads bidding actually works or whether it is a colourful Excel sheet with a logo. The asymmetry is enormous. A simulation that looks identical on the surface can have a fidelity gap of an order of magnitude underneath. The professor pays the price for the gap when a student says, “this isn’t how it works in real life,” and the professor cannot defend the assignment.

It is the territory where simulation providers genuinely diverge, and it is the question I would press any provider on hardest. Show me the model. Show me what changes in the model when the student does X. If the answer is hand-wavy, the simulation is hand-wavy. (On strategy under realistic models, the longer essay is here.)

What does productive failure actually mean?

The most useful single concept in modern learning science for marketing professors is Manu Kapur’s work on productive failure. The argument, condensed: students who are allowed to attempt difficult problems before being shown the correct method retain and transfer the learning at higher rates than students who are taught the method first and then practise. The productive failure is the failure that the student has to think their way out of.

Donald Schön, writing forty years before Kapur, made an adjacent argument. The reflective practitioner is the practitioner who learns from the gap between what they expected and what happened. In Schön’s framing, the case method gives the student the practitioner’s situation. The simulation gives the student the practitioner’s loop: the act, the consequence, the reflection, the next act.

If you want the philosophical lineage, it goes further back. Vygotsky’s zone of proximal development. Bruner on scaffolded learning. Dewey on experience as the basis of learning. (I have argued before, taking my cue from Spinoza, that knowledge proper begins where the learner is engaged with the world rather than receiving descriptions of it.) The marketing professor has not had time to read all of this, and that is fine. The summary is enough.

Real learning happens at the boundary of what the student can already do. Set the difficulty too low and the student is bored. Set it too high and the student gives up. The pedagogy of friction is the pedagogy that finds the boundary, holds the student there, and makes the discomfort productive.

This is not soft pedagogy. It is the opposite of soft. A simulation that allows clickthrough success has set the friction to zero, which is why students learn nothing from it. We have a name for what we do instead: praxis. Reflective doing. The act, the consequence, the reflection, the next act.

What does this mean in a 12-week marketing course?

The combination is the answer. Cases for the strategic register. Simulations for the operational register. Each compensates for the other’s failure mode.

A worked example. Imagine a 12-week digital marketing course. Weeks one and two are theory, set up with the relevant cases. Weeks three through ten are simulation: paid search, paid social, organic content, B2B with AI, the AI skills set, run as four two-week sprints, each one with budget, decisions, consequences, and a debrief. Weeks eleven and twelve return to cases, this time with the students having genuine operational experience to bring to the discussion. The cases now read differently. The students are not arguing in the abstract. They are arguing from what they have done.

I have published a working AI marketing syllabus that maps onto exactly this rhythm. (See the 2025 Novela AI marketing syllabus.)

The result is a cohort that has the rhetorical equipment of the case method and the operational instincts the case method cannot give them. This is what the syllabus ought to look like in 2026, and it is roughly what the better marketing programmes are already converging on.

What does Novela actually do that makes this work?

A short list. Each item is a deliberate design choice that came from teaching, not from a product workshop.

Every choice has a consequence

Set the Google Ads budget too low, and the campaigns underdeliver, the data is thin, the next week’s analysis suffers. Set it too high, and the spend outruns the conversions, the gross margin collapses, the student has to defend the decision in the debrief. There is no pressing “next” through this. The student is in the consequence loop. (Browse the simulation suite here.)

The model is a real approximation of the platform

Quality Score, learning periods, smart bidding limitations, the way Meta’s algorithm handles narrow audiences in 2026. We have built this from teaching the platforms and from running campaigns on them. It is not perfect. It is closer than the alternatives, and we will defend the modelling choices in any conversation a professor wants to have. (Five worked uses of the search marketing simulation, with screenshots.)

Designed by people who have taught and run campaigns

The single sentence that explains the difference between Novela and most of our competitors. The simulations are designed by practitioners who have also taught at universities. The choice is felt every time a student tries to game the simulation: the gaming gets caught because the underlying model has been built by someone who has had to run the same trick on a real campaign and watched it not work.

Where does the AI fit in?

AI does not replace simulation-based learning. AI is a category of capability that students need to learn how to wield, and a simulation is the right environment in which to learn that.

We are building an AI Search simulation precisely because the channel is changing fast enough that traditional case writing cannot keep up. The case for what is currently being called “GEO”, or “answer engine optimisation”, or whatever the agreed term is by 2027, will not be a case study any time soon. It will be a simulation, and it will need to be refreshed three times a year as the LLMs change. (We announced the AI Search simulation here.)

The AI Skills for Marketing simulation is already live. It teaches prompting, AI workflow design, evaluation of AI-generated work, and the boundary judgement that separates a marketing professional who can use AI from one who has been replaced by it. (See the AI Skills for Marketing simulation.)

What are the honest caveats?

Three, before I close.

First, simulation-based learning is not a panacea. Simulations are best at the operational and tactical layers and weakest at the genuinely strategic. A simulation cannot teach a student to decide whether a brand should reposition. The case method can. This is why the combination matters.

Second, the literature on simulation efficacy in marketing is thinner than it should be. There is good work in operations management (Sitzmann’s 2011 meta-analysis is a useful starting point), some in strategy (Faria), and the overall direction of the evidence is positive. Simulations produce stronger procedural knowledge and similar declarative knowledge compared with traditional methods. The marketing-specific evidence base is small, which is a fair caveat, and it is also where the next decade of marketing pedagogy research will land. Departments adopting simulations now are well placed to contribute to it.

Third, the cost of running simulations badly is higher than the cost of running cases badly. A poorly-run case discussion bores 80 students for 90 minutes. A poorly-run simulation produces 80 students with a transcript of demonstrably wrong learning. The pedagogy demands more from the professor: the debrief is where the learning is consolidated, and the debrief is where many simulation programmes fail.

I would rather Novela was having this conversation in public than the conversation about AI feature parity. The AI question is interesting. The pedagogy question is more interesting, and it is the question that determines whether students actually learn anything at all.

What does this mean for the marketing professor on Sunday night?

Three suggestions.

If you have been running cases and feel they are increasingly disconnected from how students will actually work after graduation, add one simulation in week six. Just one. See how the cohort responds. That is the cheapest, lowest-risk experiment available.

If you have been running simulations and feel the strategic register is missing, set a single rich case in the second-to-last week, with the students drawing on what they did in the simulation. This converts the simulation into raw material for the case discussion, and it is one of the most effective pedagogical moves available in our field.

If you are designing a new course from scratch, build it as a 70/30 split: simulation for the operational sprints, case for the strategic bookends. This is what the strongest marketing courses I have seen at INSEAD, Imperial, and Chicago Booth are converging on, and it is the working assumption inside our syllabus bundle.

The pedagogy of friction is the long, unfashionable answer to the question every marketing department is asking. Students learn marketing by making decisions, feeling consequences, and reflecting on the gap. Simulations are the cheapest way to put that loop in front of a cohort of any size. They are not the only thing a marketing course needs. In 2026 they are the closest thing to a non-negotiable.

Frequently asked questions

What is the difference between simulation-based learning and the case method?

Simulations put the student inside the decision and let them feel the consequences over time. The case method puts the student outside the decision and asks them to argue what they would do. Both have value. The combination is stronger than either alone.

Are marketing simulations effective?

The marketing-specific evidence base is small but consistent. Simulations produce stronger procedural knowledge (knowing how) and roughly equivalent declarative knowledge (knowing what) compared with traditional teaching. They are strongest at the operational and tactical levels.

Why do some marketing simulations get a bad reputation?

The cheap ones allow clickthrough success. A student who can earn a passing grade by clicking “next” is learning nothing. The diagnostic question is whether a poor decision in week three produces a worse outcome in week five. If yes, the simulation has a chance.

What is productive failure in marketing pedagogy?

The concept comes from Manu Kapur. Students who attempt difficult problems before being shown the correct method retain and transfer the learning at higher rates than students taught the method first. Friction is what makes the learning stick.

What should a 12-week digital marketing course look like in 2026?

Roughly: two weeks of theory and case set-up, eight weeks of simulation-based operational sprints, two weeks of case-based strategic synthesis. Cases for the strategic register. Simulations for the operational register.

How is Novela different from other marketing simulation providers?

Two structural differences. First, every choice in our simulations has a consequence; clickthrough success is not possible. Second, the simulations are designed by people who have run real campaigns and taught at universities. The combination is the difference.