Business Simulation

Designing a Market Your Teams Can Actually Learn From

Singapore sells this family in two very different shapes, and the gap between them is wider than any other format in the corporate market. At one end are outdoor acquisition races in which teams move around the island buying fictional plots and locations, published at 2 to 4 hours for 30 to 3,000 participants at $40 to $60 per head, above an event fee floor of roughly $2,000. At the other end are seated, facilitated learning simulations that occupy a half day, take a room of forty at most, and sit in the $120 to $250 specialist band, with one published local half-day simulation workshop priced at $406 per participant. Counting the trading and property-dealing versions, at least five local suppliers offer something in this family, and several of them publish a list of game titles with no duration, no headcount and no rate attached to any of them.

One disclosure before anything else. PLAYON holds no facility of any kind for this format and does not offer it.

That is the one line it deserves. Everything below concerns the activity itself, which is worth getting right and which almost nobody writes down properly. Every listing you will read describes the theme. None of them describe the round loop, the resolution model, how long the lag is between a decision and its consequence, or why that lag is the single setting that determines whether your group learns anything. Those are mechanism design questions, and a simulation is nothing but mechanism. Get them wrong and you have bought a board game with a facilitator standing next to it. This page works through each of them and finishes with a complete five-round market, numbers included, that you can run as written.

The simulation as Singapore buys it, and the four shapes it arrives in

A business simulation is a competitive exercise in which teams operate a fictional enterprise or trade in a fictional market, submit decisions on a fixed schedule across several rounds, and receive results produced by a model rather than by a referee's judgement.

The phrase carrying the weight is produced by a model. It is what separates this family from everything adjacent. In a station-rotation format a facilitator scores a task. In an escape room a lock either opens or it does not. Here, a set of numbers goes in, a rule set converts them, and a different set of numbers comes out, and the same decision can be brilliant in round two and ruinous in round four because four other teams also moved. Teams are not competing against a task. They are competing against each other through an intermediary that neither side controls.

Local listings fall into four shapes, and which shape you get matters far more than the name printed on the brochure.

The seated enterprise simulation. Teams run a fictional firm across several decision rounds, choosing production, pricing, hiring, investment or supply commitments each round. A facilitator resolves the market between rounds and publishes standings. This is the version that carries actual learning weight, it is the one that costs the most per head, and it is the one Singapore operators publish least about.

The trading variant. Teams hold commodities, contracts or cards whose value differs between teams, and the round is a bounded negotiation window rather than a decision submission. Score comes from the value of what you finish holding. This version lives or dies on information asymmetry, discussed at length below, and it is the one most often sold under a single operator's own product badge.

The property-dealing variant. Teams acquire locations, plots or assets and collect income from them across rounds. Locally this is usually delivered outdoors as an acquisition race across the island, with teams travelling between sites and earning the right to buy through short physical or mental tasks. Published figures for that delivery sit at 2 to 4 hours, from 30 participants up to headline capacities in the thousands, at $40 to $60 per person. It is a hybrid: a route format with an economic scoring layer bolted on, and it behaves much more like a race than like a seated simulation.

The specialist domain simulation. A model of one specific thing, usually built for training rather than for team building, and often licensed from a publisher. Urban redevelopment, supply chain, project portfolio and hospital operations all exist in this shape. One local half-day version of the redevelopment type publishes 4.5 hours and $406 per participant, which tells you where the ceiling is when the model is the product.

Read those four together and the price gap explains itself. The $40 to $60 versions are paying for logistics, marshals and a route. The $120 to $250 versions are paying for a model that somebody spent months building and validating, plus a facilitator who can run a debrief off it. The middle of the local market, where a supplier lists a dozen business game titles with no figures published against any of them, is where you have to ask questions, because a title alone tells you nothing about which of the four shapes you are buying.

For the rest of this page, "the simulation" means the seated version, with the trading variant treated wherever its design differs.

The decision loop, and the four things every round must contain

A round is not a turn. A turn is a thing you take. A round is a machine with four moving parts, and if any one of them is missing the exercise degrades into something else.

1. The decision window. A fixed, visible period in which each team must produce a small number of committed choices, written on a slip and handed in. Three decisions per round is the working maximum for a corporate group. Two is often better. The moment a team has five or six levers to move, the window fills with arithmetic and nobody talks about strategy.

2. Resolution. The facilitator collects all slips and runs them through the model. Every team's outcome depends on every other team's submission, which is the property that makes it a market rather than a puzzle. Resolution has to be mechanical and fast, and the single most useful thing you can build before the day is a spreadsheet that takes five sets of three numbers and produces five results in under ninety seconds.

3. Feedback. Results are published. What gets published, and how much of it, is a design decision and not an afterthought. The minimum is each team's own position. The maximum is every team's full decision set, which turns the next round into a game of reading opponents.

4. The delay. Between a team's decision and the moment its consequence is visible, one or more rounds pass. This is the part that gets left out, and it is the whole exercise.

Around those four, two supports that people skip and then regret.

A practice round. Round zero, run publicly on the board with fictional submissions the facilitator invents on the spot, resolved in front of everybody, with the arithmetic shown. It costs eight minutes and it removes the single most common complaint, which is that teams did not understand what the model was doing until round three.

A written reason. One line on every decision slip saying why. Nobody scores it. It exists so that the debrief has something to point at, and its value at the end is out of all proportion to the six seconds it costs.

Feedback delay is the dial that decides what the room actually learns

If you change nothing else after reading this page, change this.

Zero delay. A team decides, the result appears immediately, and the next decision is a correction. Round two adjusts round one, round three adjusts round two. What the room practises is tuning: read the last signal, move the lever, read again. This is a real and useful skill, and it is the skill of an operator. It is also very satisfying, because the loop closes fast and teams feel competent by round three.

One-round delay. A decision in round one shows up in round two. Teams get a taste of committing before knowing, but the horizon is short enough that memory covers it, and most groups treat it as tuning with a lag.

Two-round delay. A decision in round one lands in round three. This is where the exercise changes character. A team must now hold two open commitments at once whose results it has not seen, decide whether to add a third, and do all of this while the standings are being driven by other teams' commitments it also cannot see. Nobody can tune their way through it. The only way to play well is to form a view about where the market is going and back it before there is evidence.

That is planning under uncertainty, and it is a different cognitive activity from optimisation. It is also, for most corporate groups, closer to what their actual jobs are made of. The finance team commits headcount for a year. The product team commits a roadmap for two quarters. Nobody in either of those situations gets to see the result before the next decision is due.

Three-round delay and beyond. Now the causal thread snaps. Teams cannot remember what they did three rounds ago, cannot connect the outcome to it, and start attributing results to luck. Once a room believes results are random, it stops deliberating and starts guessing, and every design property you built collapses at once.

So: two rounds, in almost every case. One if the group is inexperienced or the session is short. Three only if you are running eight rounds or more and you have a visible decision log on the wall that teams can consult.

Two practical corollaries fall out of this.

The delay sets your round count. A two-round lag means an investment made in round one is only visible in round three and only pays back in rounds four and five. Anything committed in round four cannot resolve inside the session at all. That arithmetic is what makes short simulations useless, and it is the subject of its own section below.

The delay has to be stated, not discovered. Tell teams in the brief that investments take two rounds to land. Hiding the lag does not make it more interesting, it makes it feel arbitrary, and a team that discovers in round four that its round two decision has silently been working the whole time does not feel clever, it feels cheated. The uncertainty you want is about the market, not about the rules.

Fixed rules or hidden variables: what each model teaches

The second design choice is how much of the resolution model teams can see. There are two clean positions and one that is usually a mistake.

Fully published rules. Every formula, every table, every parameter is on the brief. Teams can, in principle, compute the optimal move.

What this produces is arithmetic, and arithmetic has a winner: whoever is fastest and most careful with numbers. It is not a bad exercise, and it holds one real virtue: nobody can claim they lost because the facilitator moved something. It is, however, closed. Within two rounds a sharp team finds the dominant line, and from there the exercise is execution rather than judgement. It is also, and this matters, the version where the two people in the room with a finance background quietly take over, because the game has been converted into a task they are better at.

One hidden variable. Everything is published except a single unknown that materially affects results. Total market demand each round is the natural candidate, because it is the one thing a real firm never knows and always has to estimate.

This is the recommended setting, and the change it produces is immediate. Because no team can compute the answer, every team has to form a belief, state it, and act on it. Beliefs are arguable in a way that arithmetic is not, so the quiet person who thinks demand is about to fall has standing to argue. The team's decision becomes a negotiated position rather than an output, and that is exactly the behaviour the debrief will be about.

Give the unknown a shape rather than leaving it blank. Announce a range to the teams each round, or last period's figure, or a one-line market outlook, phrased vaguely and honestly: "analysts expect continued growth into the second half". Teams can then be wrong for interesting reasons instead of being blind.

Several hidden variables. Now avoid this. Two or more unknowns, especially interacting ones, and teams cannot attribute any outcome to any decision. The room concludes that the model is a black box, engagement drops off a cliff, and your debrief is left with nothing usable, since no one in the room can say why anything happened. One unknown produces uncertainty. Three produce noise, and noise teaches nothing.

The test to run on your own model. After you have built it, sit down and ask: if a team plays reasonably well, can it tell at the end which of its decisions was the good one? If the honest answer is no, you have too many hidden variables, or your delay is too long, or both. Fix it before the day.

Why unequal information turns arithmetic into negotiation

This is the lever behind every trading variant, and it is worth understanding even if you never run one, because it is also the cheapest single upgrade you can make to a plain enterprise simulation.

Start with the plain case. Every team holds the same information and faces the same rules. When two teams talk, there is nothing to discover: both already know everything the other knows, so a conversation can only be about splitting a fixed quantity. That is haggling, and it is zero sum, and after one round teams stop bothering.

Now give each team a private value sheet. The same five commodities are on the table, but every team values them differently, and no team can see anyone else's sheet.

Team 1 holds cobalt it scores 2 for. Team 3 holds copper it scores 3 for. A straight swap moves Team 1 from 2 to 8 and Team 3 from 3 to 9. Both teams are better off by six points, and thirteen points of value that did not exist a minute ago now exist.

Nothing about that is discoverable by computation. Team 1 cannot deduce Team 3's sheet. The only route to it is asking, and asking well: what are you short of, what would you give up, what do you actually need this for. The exercise has stopped being arithmetic and become negotiation, which is a completely different set of muscles and, not incidentally, a set that the quantitative people in the room hold no advantage in.

Four rules make it work.

1. Values must genuinely differ, and differ in both directions. If one team's sheet is uniformly higher, it just wins. Build the sheets so every team is strong on two commodities and weak on two, and no two teams share a profile.

2. Sheets stay private, permanently. Not private until round three. Private at the end too, revealed only in the debrief. The moment a sheet is public the gains from trade become computable and the negotiating stops.

3. Trades must be declarable but not enforceable in advance. Teams may promise anything. Only the completed exchange counts. This introduces the question of whether a team's word holds across rounds, which is the most interesting thing this variant produces and the thing the debrief should go after.

4. Only one nominated person per team may leave the table. Rotate them every round. Without this rule, one confident negotiator becomes the team's entire external interface for the whole session, and four people spend three hours watching.

Bolting asymmetry onto an enterprise simulation. You do not need a full trading game to get this effect. Give each team one piece of private market information at the start of each round: one team learns next round's demand direction, another learns a competitor's capacity, another learns that a cost input is about to move. Allow teams to trade information for information, verbally, with no enforcement. The economics of the simulation are untouched, and the room acquires a second layer in which the currency is what you know. It costs five index cards.

How many rounds before learning happens, and why three is not enough

Suppliers quote three rounds constantly, because three rounds fits an afternoon and looks like a complete arc. It is not enough, and the reason is arithmetic rather than taste.

Take the recommended two-round lag. A decision in round one becomes visible in round three. So in a three-round simulation:

• Round one decisions resolve in round three, the final round. Teams see the result once, with no opportunity to act on it.

• Round two decisions resolve in round four, which does not exist. That decision never resolves at all.

• Round three decisions resolve in round five, which also does not exist.

Two of three rounds of decisions are therefore inconsequential, and the teams work this out somewhere in round two. What they do next is entirely rational and completely ruins the session: they stop investing and start harvesting, because only immediate returns can score. A three-round simulation with a two-round lag is a game about short-term extraction, which is the precise opposite of what anybody books it for.

Five rounds is the working minimum. Round one and round two decisions resolve in rounds three and four, leaving two and three further rounds respectively for them to compound. A team can commit early, watch the commitment land, and see it pay. That is the full causal arc, and it is the smallest structure that contains one.

Six or seven rounds is better if the clock allows, because it lets a team make a mistake in round two, recognise it in round four, correct it in round five and recover by round seven. The recovery arc is the most valuable thing this format produces and five rounds barely fits it.

Eight or more and attention becomes the binding constraint rather than design. Rounds start to feel repetitive, and unless the model introduces genuinely new decisions later on, the last three rounds are the first three again with bigger numbers.

The endgame problem, which everybody hits. In the final round, no investment can pay back, so the rational move is to strip the position: stop investing, cut quality, dump inventory, price for cash. If your score is cumulative cash, the final round inverts the standings and rewards the team that abandoned the strategy the session was meant to be about.

Two fixes, and you should use one of them.

• Value the assets at the close. Final score equals cash plus a published valuation of remaining capacity, quality level and any other durable position. Now stripping costs you exactly what it gains you.

• Do not announce the final round. Tell teams the session runs between five and seven rounds and that the stop depends on the clock. Uncertainty about the horizon reproduces the real condition, which is that nobody knows when the game ends.

Announcing the round count and using cumulative cash is the combination to avoid, and it is the most common one in circulation.

Five rounds of a components market, with the arithmetic in full

Here is a complete simulation, small enough to read end to end and large enough to run for real. Five teams, five rounds, three decisions per round, one hidden variable, a two-round investment lag. Change the names and the product freely. The structure is the point.

Starting position, identical for all five teams

The three decisions each round

1. Units to produce, between zero and current capacity. Production cost is charged whether or not the units sell.

2. Selling price per unit, any whole dollar figure.

3. One investment, chosen from the four options below, or none.

An investment bought in round one is live from round three. Say this in the brief.

How the market resolves

Total market demand is the hidden variable. Teams are told round one demand and given a one-line outlook each round afterwards. They are never told the actual figure in advance.

Each round, every team receives an attractiveness score:

Score = (quality rating x 20) minus price

Teams are ranked by score, highest first. Demand is allocated down that ranking, each team selling up to the units it has available, until demand runs out. Teams below the cut sell nothing. Unsold finished units carry to the next round at a holding cost of $5 per unit and can be sold later.

The score formula tells teams something true and useful: one point of quality is worth exactly $20 of price. A team that lifts quality from 3 to 4 can charge $20 more and hold the same position in the queue. Publishing that formula is deliberate. The formula is not the uncertainty. Demand is.

The hidden demand schedule

Combined starting capacity is 1,000 units, so the market is short from round three onwards. Teams that expanded early can fill the gap. Teams that did not cannot, and they discover this in round three, which is exactly when the round one investment lands.

Round 1, demand 850

All five teams are at quality 3, so the ranking is purely by price. Supply is 900 against demand of 850, and D is last in the queue with 50 units left unsold.

Standings after round 1: A $38,400, C $31,360, E $31,100, B $29,800, D $26,030.

Team A took the largest profit available in round one by the simplest possible route: produce at full capacity, price at the front of the queue, spend nothing. It leads by more than $7,000, which in round one feels decisive. Every other team is carrying an unresolved commitment that has cost it cash and returned nothing.

This is the moment to note, because everything that follows is a consequence of it.

Round 2, demand 950

No investments have landed yet. D carries 50 finished units from round one.

Supply is 960 against demand of 950. C sells 150 of 160.

D produced only 150 units because it already held 50, so its production cost is $6,000 rather than $8,000, and it posts the best round in the game so far while sitting third.

Standings after round 2: A $47,200, B $38,400, D $36,030, E $33,500, C $28,660.

Team A now leads by nearly $9,000 and has raised its price, because winning round one taught it that its approach works. It has still invested nothing. This is the point at which a facilitator watching the room should be quietly pleased, because the trap is fully set and the team has walked into it for entirely defensible reasons.

Round 3, demand 1,150

Round one investments land. B is at quality 4. C is at unit cost $34. D and E are at capacity 280.

Supply is 1,160 against demand of 1,150. A is last in the queue and sells 190 of 200.

Look at B. Quality 4 gives it a score of 80 minus price, so at $95 it sits at the front of the ranking while charging $11 more than anyone else. That is the round one investment arriving.

Standings after round 3: B $49,400, A $48,490, D $47,790, E $45,540, C $37,860.

A has lost the lead. It invested for the first time in round three, which will land in round five, and it paid $7,000 for the privilege in the same round its capacity ceiling started to bite. Three teams are now within $2,000 of each other and A is no longer one of the top two.

Round 4, demand 1,400

Round two investments land. C is at quality 4. E is at unit cost $34.

Supply is 1,160 against demand of 1,400. Everything sells, including A's full run, and the queue position no longer costs anyone volume. It costs them price.

Standings after round 4: B $60,600, D $58,990, E $58,140, A $57,290, C $49,460.

There it is. Team A, which took the largest profit in the market in round one and led outright for two rounds, is fourth of five. It has done nothing wrong in any individual round. Every single decision it made was locally sensible: produce to capacity, price to clear, avoid spending cash you have not earned yet. The reason it is fourth is that in a market with a two-round lag and rising demand, "locally sensible" compounds into a ceiling, and three of its competitors bought their way past that ceiling in round one while A was banking the difference.

Note also what has happened to A's round result. It is still making $8,800 a round, which is almost exactly what it made in round two. Nothing has gone wrong. It is standing still while the market grows around it, and standing still in a growing market is a form of decline that does not show up anywhere in a team's own numbers. Teams almost never see this from inside their own sheet, and when the standings are projected in round four they are visibly surprised. That surprise is the most valuable fifteen seconds of the session.

Round 5, demand 1,500

A's round three capacity investment lands, taking it to 280.

Supply is 1,240 against demand of 1,500. Everything sells again.

Final standings: B $72,200, D $70,470, E $70,460, A $68,490, C $61,460.

Three findings worth walking the room through, and none of them is the obvious one.

A led for two rounds and finished fourth. Total spent on investment: $7,000, and that late. It never had a bad round and it never recovered.

C spent the most and finished last. Investment total $15,000, across two consecutive rounds. The second one landed in round four and had only two rounds to earn back. C was right about what to buy and wrong about when, and it is important that the winning story on this page is not simply "investing is good". The horizon has to cover the payback, and C's did not.

B, D and E all finished within $1,750 of each other by three completely different routes: B bought quality and priced high on 200 units, D and E bought volume and priced low on 280. There is no dominant strategy in this model, which is deliberate, and it is what stops the debrief becoming a lecture about the right answer.

If you run the closing valuation rule, add the value of remaining assets. At $7,000 per capacity step and $9,000 per quality point, A's late capacity purchase and C's quality position both look considerably better than the cash column suggests, and the ordering tightens. Deciding in advance which of those two scoreboards you are using is part of the design, not an afterthought.

How many people a market needs before it behaves like a market

Team size: four to six. Below four, the roles described later cannot all be filled and one person makes the decision while the others watch. Above six, someone is not touching the sheet. Five is the working default.

Minimum teams: three. With two, there is no market. Each allocation reduces to a head-to-head between the only pair present, price discovery becomes a private staring contest, and if the trading variant is in play, every negotiation is with the only counterparty available, so nobody has an alternative and reciprocity means nothing.

Maximum teams in one market: eight. Past eight, three things break simultaneously. Resolution slows to the point where the gap between rounds exceeds the decision window. The standings board becomes unreadable at the back of a function room. And the effect of any one team's decision on the market gets diluted until teams cannot detect their own influence, which removes the reason to think hard about it.

So a single market carries 12 to 40 people, with 20 to 30 being the shape it is built for.

Above 40, run parallel markets. Identical starting conditions, identical demand schedule, no contact between rooms or table clusters, and compare final positions at the end. This is genuinely interesting rather than merely a workaround: two independent markets given the same conditions rarely produce the same winner, and showing the room that fact is a debrief point you cannot get any other way. Two markets need two facilitators and two resolution sheets. Budget for that rather than hoping one person can run both.

How to compose teams. Deliberately, and against the grain of the org chart. This format has a specific equity problem, addressed below, and the composition step is where most of the fix lives. Split people with finance, analytics, pricing or trading backgrounds across every team so that no team has two and no team has none. Split seniority the same way, because a table containing one director and four juniors will produce the director's decision every round, and you will spend the debrief talking about that instead of about the market.

Do not let people choose their own teams. In an office group, self-selection reliably produces one table of the numerically confident, and that table wins, and everybody learns the wrong lesson.

Round length, total length, and the compressions that do not cost you the lesson

A single round: 12 to 18 minutes. Roughly eight to twelve minutes of decision window and three to six minutes of resolution and publication. Round one always overruns, so brief it as twelve and expect eighteen. By round three, teams typically want less time rather than more, and you should give it to them: shortening the window from round three onwards raises the pressure at exactly the point where the lag makes decisions harder, which is the right shape.

A full session: two and a half to three and a half hours.

That lands between 2 hours 35 minutes and 3 hours 5 minutes. Published local durations for the event-style variants run 2 to 4 hours and one seated half-day workshop publishes 4.5 hours, so this sits comfortably inside what the market already sells.

The shortest honest version is 2 hours. Five rounds at 12 minutes, a 10 minute brief with no practice round, and a 25 minute debrief. It works if the group has done something like this before. It does not work cold, because without the practice round the first two live rounds become the practice round, and you have effectively bought a three-round simulation.

What to compress, in order.

1. Decisions per round, from three to two. Drop the investment menu to two options rather than four. This cuts the decision window by a third and costs almost nothing pedagogically.

2. The break. Painful for a three-hour session but recoverable.

3. The number of teams. Four teams resolve faster than seven, in the window and at the board.

What never to compress.

The round count. This is the one everybody reaches for first and it is the one that destroys the exercise, for the arithmetic reasons set out above. If your slot cannot hold five rounds, run shorter rounds with fewer decisions, or run four rounds with a one-round lag and accept that you are teaching tuning rather than planning. Do not run three rounds with a two-round lag and tell yourself it is the same thing smaller.

The debrief. Forty minutes, and it is not the part you trim when the rounds overrun. A simulation with no debrief is a board game with a scoreboard. Protect it by starting the compression at round four, not at the end.

What to print, what to project, and what the room has to do

Per team

• Decision slips, one per round per team, pre-printed with the round number, the fields, a line for the written reason and two signature boxes. Numbered, so the facilitator's log builds itself.

• A position sheet, updated by the team each round: cash, capacity, unit cost, quality, stock carried.

• Role cards, five of them, described in the equity section below.

• A private value sheet, if you are running the trading variant or the information layer.

• A calculator or a laptop. Allow it. Banning tools does not make the exercise about judgement, it makes it about mental arithmetic, which is the opposite of the intent.

For the room

• The resolution spreadsheet. This is the entire product. It takes the submitted numbers and produces the allocation, the results and the standings, and it must be built and tested before the day.

• A projected standings board, refreshed between rounds. Cumulative position, current round result, and nothing else. Legible from the back.

• A visible round clock, counting down the decision window.

• A printed rule sheet per team, including the score formula and the investment table. Teams will read it four times in round one and never again.

• The facilitator decision log, described in the debrief section. This is the second most important artefact after the spreadsheet.

The venue

A flat room with round or cabaret tables, one per team, far enough apart that a team cannot overhear its neighbours' deliberation. A projector or a large screen. Power. That is all. It is a meeting room, which is why this format is weather-proof, quiet and unusually easy to place, and also why it feels to some groups like a longer meeting.

What cannot be improvised

The resolution model. Everything else on the list can be bought in a stationery shop on the morning. The model cannot, and building it is the whole job.

An honest estimate for a competent spreadsheet user building one from scratch:

Three to four days, and the rehearsal is not optional. Every homemade model has a degenerate strategy in it, usually a price point at which producing nothing and holding stock beats playing, and you will not find it by yourself because you built the thing and you think in its assumptions. Give it to four colleagues, tell them the prize is breaking it, and watch what they do.

Set against a $120 to $250 per head quote for thirty people, three to four days of one person's time is comfortably cheaper, provided you intend to run the thing more than once. For a single date, especially a close one, buy it.

Stopping the three people with spreadsheets from playing on everyone's behalf

Every format has an equity problem. The physical ones sideline anybody who does not want to run. This one sidelines everybody who is not quick with numbers, and it does so quietly, without anybody deciding it, in about eleven minutes.

The mechanism is worth naming precisely, because the fix follows from it. The decision window is short. The decision has a numerical output. One person at the table can produce that output faster than the group can discuss it. So they do, and they are helpful, and the table is grateful, and by round two the other four have settled into a supporting role they never agreed to and will not now leave. Nobody is at fault. The structure did it.

Five interventions, in order of how much they achieve.

1. Rotating roles, one per person, changing every round. This is the big one. Print five cards.

• Chief Executive. Signs the slip and breaks ties. Does not compute.

• Analyst. Owns the numbers and produces the options. May not choose between them.

• Market Officer. The only person permitted to leave the table, negotiate or gather information from other teams.

• Recorder. Writes the one-line reason on the slip and maintains the position sheet. Reads the reason aloud in the debrief.

• Challenger. Must state one objection to the proposed decision before the slip may be signed. Not optional and not a formality.

The split that does the work is Analyst and Chief Executive. Separating "who produces the options" from "who chooses" is the single structural change that stops numerical fluency from converting directly into control. The quick person is still valuable, still busy, still visible, and no longer deciding.

Rotate every round. Over five rounds each person holds each card once, and the numerate colleague spends four rounds not being the Analyst.

2. Make the reason mandatory and the signature dual. Two signatures on every slip, and one written line saying why. It takes six seconds and it forces a sentence that at least two people have agreed to. Tables that were not talking start talking, because you cannot co-sign a reason you have not heard.

3. Use invented units and a fictional market. Do not call it EBITDA. Do not denominate it in a real currency the finance team reasons in fluently. Widgets, units, credits, a made-up commodity. This costs nothing and it removes the domain advantage while leaving the reasoning advantage intact, which is the correct split: you want the person who thinks clearly to do well, not the person who already knows what a contribution margin is.

4. Show the teams the score formula. Counter-intuitive, but sound. If the resolution rule is secret, the people who can reverse-engineer it from three rounds of results acquire an enormous edge, and it is precisely the numerate people. Share the formula with every team and the advantage disappears, leaving the uncertainty where you want it, in the hidden demand variable, which nobody can compute their way to.

5. Compose the teams yourself. Covered above and worth repeating here, because if two of your quantitative people end up at the same table, no role card will save the other three.

What to do when it happens anyway. It will, at one table. Do not announce it to the room. Walk over between rounds and hand that table an extra constraint for the next round only: the Analyst may not speak during the decision window and may only answer direct questions in writing. It is a blunt instrument, it lasts one round, and it usually resets the table permanently, because the other four discover they can do it.

Scoring the outcome and the reasoning, in two variants

Cumulative cash is the obvious scoreboard and it has two known defects: it inverts in the final round unless you value assets, and it rewards outcome over process, which means a team that reasoned badly and got lucky beats one that reasoned well and got unlucky. In a five-round market with a hidden variable, that happens often enough to matter.

Variant one: the balance sheet close. Final score is cash plus assets, valued at the close.

Assets are valued at what they cost, not at a mark-up, so buying them is never free score. What this achieves is that a round five investment is neither penalised nor rewarded, which removes the endgame distortion completely and lets teams keep playing the game they were playing.

Announce this in the brief. Announcing it at the end, after teams have stripped their positions, is the fastest way to lose a room.

Variant two: outcome plus reasoning, out of 100.

This is the version to use when the session is genuinely developmental rather than competitive, when the group contains large seniority differences, or when you expect one table to run away with the market. A team can lose the market and win the exercise, and it can see exactly why.

The round three self-assessment is the component worth stealing even if you use variant one. Ask every team, in writing, at the midpoint: what did you buy in round one, what do you expect it to do, and in which round. Collect the slips. Read them out at the end against what actually happened. It takes four minutes and it converts a vague sense that planning matters into a specific, personal, documented prediction that either held or did not.

The tie-break. Score ties happen. Break them on the round three self-assessment, not on speed of submission. Speed rewards the team that deliberated least, which is precisely backwards.

Basic, advanced and punishing: which parameter moves between them

One dial at a time. Turning three at once produces a session where nobody can tell what went wrong.

Basic. Four teams, five rounds, two decisions per round (produce and price, with a single yes-or-no investment option), a one-round lag, demand published one round ahead, score formula published, scoring on cash plus asset close. Suits a first-timer group, a post-lunch slot, or a room of very mixed confidence. The one-round lag is the concession that makes it basic, and it is the right one to make, because everything else stays intact.

Standard. The worked market above, exactly as written. Five teams, five rounds, three decisions, four investment options, a two-round lag, one hidden variable with a vague outlook each round, published formula, cash plus assets.

Advanced. Move exactly two of these five, not more.

• Lengthen the lag to three rounds, and simultaneously run seven rounds with a visible wall-mounted decision log. Without the log this is not harder, it is random.

• Add the information layer. One private piece of market intelligence per team per round, tradeable verbally, unenforceable. This is the highest-value single addition on the list and the cheapest to build.

• Add a shock. One announced but undated event: a demand collapse, a cost spike, a new entrant taking a fixed slice of demand. Tell teams in the brief that exactly one shock will occur between rounds three and six. Announcing that a shock exists while withholding when is the correct amount of information, because it changes behaviour without producing helplessness.

• Add a second hidden variable. Reluctantly, and only for a group that has run the standard version. Read the warning in the model section first.

• Remove the published score formula. Only with the information layer switched on, so that the formula becomes something teams can trade for rather than something the quickest analyst reverse-engineers alone.

Do not make it harder by shortening the rounds. Time pressure does not deepen judgement in this format. It converts deliberation into whatever the fastest person at the table says, which is the exact failure the role cards exist to prevent, and you will have engineered it deliberately.

Six ways a simulation day comes apart, and the guard against each

A team goes bankrupt in round two and stops playing. The most damaging failure, because it removes a fifth of the room for two hours and the people affected are usually the ones the session was booked for. Prevention is structural: publish a credit facility from the start, so cash below zero is a loan at a stated cost rather than an exit, and never eliminate anyone. Cure on the day, if it happens anyway: rule that the team has been recapitalised by its board, restore it to a viable position, and say so publicly without irony. A team that cannot win can still play for a placing, and nobody learns anything from three hours of watching.

Resolution takes longer than the decision window. The facilitator is still typing when the next slips arrive, the clock slips ten minutes a round, and round five gets cut. Prevention is the spreadsheet, tested against five sets of realistic inputs and timed. If it takes you more than ninety seconds, simplify the model. Cure on the day: drop to two decisions per round from the next round, and say why.

Teams conclude the model is arbitrary. Symptom is teams submitting quickly and without discussion by round three, and a flat, polite debrief. Cause is almost always too many hidden variables, or a lag longer than the room can hold. Prevention is the practice round with the arithmetic shown publicly, plus one hidden variable and no more. Cure: publish the full resolution for one round on the screen, line by line, showing exactly how each team's number was produced. It costs five minutes and it usually brings the room back.

One table takes an unassailable lead by round three. The other four disengage. Prevention is parameter tuning during your rehearsal: if any single strategy produces a runaway in your solo playthrough, the model needs a brake, usually a capacity ceiling or a demand allocation rule that caps any one team's share. Cure on the day: introduce the shock event immediately and aim it at the leader, which is legitimate if you announced in the brief that a shock would occur.

The last round becomes a liquidation. Covered above. Prevention is the asset valuation, announced at the start. There is no cure once teams have stripped their positions, which is why announcing it late is worse than not having it.

Nothing survives for the debrief to point at. The facilitator asks teams why they did what they did, and round one has gone from everybody's memory. This is the commonest failure of the six and the cheapest to prevent, at a cost of one sheet of paper per team, which is the next section.

A seventh, worth mentioning. Somebody who plays this genre recreationally recognises the model and plays it well immediately. Rarer here than in other formats but not rare. Ask at the briefing, and if you have one, make them the Recorder for the first two rounds rather than pretending it is not happening.

Tracing a late outcome back to an early decision, question by question

This is the section that makes the difference between a simulation and an afternoon of arithmetic, and it depends on something you have to do during the rounds, not after them.

The decision log

The facilitator keeps one sheet per team, filled in as slips come in. It takes about twenty seconds a round.

The stated reason column is the one that matters and it is the one people leave out. Copy the team's own words from the slip, verbatim, including the vague ones. At the end, hand each team its own completed log.

Without this, the debrief is a conversation about feelings. With it, the debrief is a conversation about evidence, and every claim anybody makes can be checked against a line on a page they wrote themselves.

The sequence

Forty minutes, in four movements, in this order.

1. Read round one back, before showing anything else. Ten minutes. Each team reads out its round one decision and its own written reason. Nothing else happens. No commentary, no scores, no reaction from the facilitator. Just five teams saying what they did first and why. The room is usually very quiet by the third one, because people are hearing their own reasoning from three hours ago, before they knew anything, and it sounds different.

2. Show the round four standings, then the final. Five minutes. Round four first, deliberately, because that is where the reordering happened. Ask the room to say, out loud, which round one decisions are visible in that table. They will find it, and they should find it rather than being told.

3. Trace one chain per team. Fifteen minutes, three minutes each. Each team picks one decision from rounds one or two and walks its consequence forward to the final standings using its own log. Not the outcome. The chain: what we chose, what we expected, when it landed, what it did, what we did next. The facilitator's only job is to keep pushing the question "and then what did that make possible or impossible in the next round".

4. Move it across. Ten minutes. The questions below.

The questions

1. Read your round one reason again. What did you know then, and what were you assuming without knowing you were assuming it?

2. Which round did your team first realise that its round one decision had been the wrong size, too big or too small? What told you?

3. Team A took the largest profit in the market in round one and finished fourth. Nothing they did was foolish. What exactly went wrong, and in which round did it become unrecoverable?

4. When did your team last change its mind about something, and what evidence did it take? Was it new information, or was it the standings board?

5. Who at your table proposed the round three decision, and who proposed the round one decision? If it was the same person both times, why?

6. What did the Challenger object to that you overruled, and were they right?

7. You made a written prediction at round three about what your investment would do. Read it out. Was it right, and if it was wrong, was it wrong about the size or about the timing?

8. Which piece of information would have been worth the most to you in round two, and what would you have given up to get it?

9. If you had known the demand schedule from the start, what would you have done differently in round one? Now: what is the equivalent thing you do not know at work?

10. In your actual job, which decisions have a two-round lag, and are you treating them as if they have none?

Questions three and ten carry the session. Question three works because Team A did nothing wrong, which is the honest and uncomfortable version of the lesson: a run of individually sensible decisions can compound into a bad position with no error anywhere in the chain. Question ten moves it out of the fiction, and it should be last, and it should be left hanging rather than answered in the room.

Protect question six as well. Teams overrule their Challenger constantly, and about a third of the time the Challenger was right and everybody at the table now remembers it.

Decision-making under incomplete information, and everything this format cannot touch

What the mechanism produces.

Deciding without adequate information. The mechanism is the hidden demand variable combined with the lag. A team must commit money in round one to an outcome it cannot see until round three, in a market whose size it does not know, against four competitors whose plans it also does not know. There is no way to compute past that, so the team has to form a view and act. Then it has to live with it for two rounds while the evidence arrives slowly. That is a precise structural model of a real planning decision, and it is very hard to reproduce in any format where feedback is immediate.

Negotiating trade-offs against a budget. The mechanism is the investment menu. Four options, one slot, finite cash. Choosing capacity means not choosing quality, and the case for each is genuinely arguable because the model has no dominant strategy. What happens at the table is a real negotiation about priorities under constraint, conducted by people who have to co-sign the outcome. That conversation is the format's actual product, and it is worth noticing that most of the value is generated before the decision is submitted, not by the result.

Distinguishing a good decision from a good outcome. The mechanism is the hidden variable, which guarantees that at least one team will reason well and be unlucky. If you run the reasoning-weighted scoring variant, this becomes explicit and scored. Very few workplace exercises make the distinction concrete, and fewer still let a team lose the market and still be shown to have played well.

What it does not produce, and be honest about this in the brief.

Anything physical. Nobody moves. Nobody touches anything except a pen. A group that wanted to be tired at the end will be tired in the wrong way.

Anything social, in the sense most people mean. It is a poor icebreaker. It opens with a rule sheet, and ten silent minutes of reading is a difficult way to start a room of people who have not met. Bonds do form, but they form between people who solved something together at a table, which is a slower and narrower kind of connection than the one a group gets from doing something ridiculous in public. If the brief was "the new joiners do not know anybody", this is the wrong format.

Trust, in the interpersonal sense. The seated version pits tables against each other and every figure can be checked, so no one is required to take anybody on faith. The trading variant is the exception and it is a real one: unenforceable promises across rounds do generate genuine questions about whose word holds, and that is the one place this family touches trust directly.

And the honest structural limit. For a group whose real work is decision-heavy, this is close to the best format available, and the transfer to the job is more direct than any physical activity can manage. For a group whose work is not decision-heavy, it is a three-hour meeting with a scoreboard, and they will experience it as one no matter how well you run it. That is not a facilitation problem and no amount of energy at the front of the room fixes it. Choose the format against the work, not against the mood.

One more thing worth saying plainly: it rewards comfort with numbers and with written English, so where language backgrounds or educational routes vary widely across a group, a paper-and-figures format can shut out part of the room without anybody noticing. The role cards, the invented units and the published formula all mitigate it. None of them eliminate it.

Every label this family wears, and the proposals that are not simulations at all

One underlying structure carries a great many local names: business simulation, strategy simulation, business game, market simulation, trading game, commodity game, enterprise game, corporate strategy game and a spread of property, tycoon and acquisition themes. As with most formats in this market, one skeleton carries at least ten different labels, and a fair share are product labels invented by whichever single supplier sells that one. A few carry a proprietary badge that is trademarked to one operator, which means the badge tells you who you are buying from and nothing whatsoever about the mechanism.

One real distinction does hide inside the labels, and it is worth learning. A listing that leads on simulation or business game generally means seated, several rounds, a model doing the resolution. A listing leading on challenge, race, tycoon or a place name generally means the outdoor acquisition format, where the economic layer sits on top of a route and the actual activity is travelling and completing tasks. Both are sold to the same buyer and they are not remotely the same product. The price gap, $40 to $60 against $120 to $250, is your first clue and it is not a reliable one.

Six questions cut through all of it, and you can ask them in one email.

1. How many decision rounds are there? Below five, with any meaningful lag, you are buying a demonstration rather than a simulation.

2. How long is the gap between a decision and its visible consequence? If the answer is "immediate", the session teaches optimisation, which may well be what you want, though you should choose it knowingly rather than by default.

3. Is the resolution model published to participants, and how many variables are hidden? One is right. Zero is arithmetic. Three is noise.

4. Do all teams receive identical information? If yes, the trading and negotiation layer does not exist regardless of what the brochure says.

5. Is a decision log kept per team, and do we get it at the end? This is the question that separates suppliers who debrief from suppliers who announce a winner. Very few say yes.

6. What stops the two most numerate people in the room deciding everything? If the answer is a shrug or "we find it works itself out", it does not.

Question five is where the answers get vaguest, and question two is the one most likely to reveal that a supplier has never thought about it at all.

Now the neighbouring formats that get quoted against this one.

Route-based races. Teams move between locations completing scored tasks, and the design work sits in the route and the rotation. When an acquisition theme is layered onto one of these, the economics are a scoring skin: teams are not running an enterprise across rounds, they are earning purchase rights through tasks. A useful test is whether a decision made at the first location changes what is possible at the fourth. In a race it usually does not. In a simulation it always does.

Parallel station formats. Teams rotate through simultaneous physical events and scores are summed. No compounding, no model, no lag. Different exercise in every respect and frequently proposed as a substitute when a brief says "indoors, half a day".

Escape rooms. Lock-gated and sequential. Progress is mechanical, the score is time, and there is no market and no other team affecting your outcome.

Investigation and deduction formats. Everything needed sits on the table from early on, and the work is deciding what it means. There is a right answer, which a simulation does not have.

Case study workshops. Genuinely adjacent, and sometimes the better buy. Teams analyse a real situation and present a recommendation. No rounds, no compounding, no competition through a model, but a much more direct connection to the specific decision your group actually faces. If what you want is for the team to think hard about one real problem, this is the format, and a simulation is the wrong tool wearing a more exciting badge.

Negotiation workshops. The trading variant's near neighbour. Structured bilateral negotiation exercises with private briefs, run in pairs rather than as a market. Deeper on negotiation technique, with none of the compounding or the market dynamic.

PLAYON has no version of this, and why the page continues anyway

PLAYON runs no simulation of any kind. The venue is built around indoor physical attractions that keep their own scores, and this format wants tables, a projector, quiet, a facilitator and three uninterrupted hours of argument. None of that exists here, and pretending otherwise would only waste your time.

Two points are still worth making, because an organiser who has read this far needs them.

Where the brief that brought you here reads "our people make decisions for a living and I want them practising it together", this is your format, and you want the seated version rather than the outdoor one, whether you build it or buy it. Arenas and lanes answer a completely different question, and enthusiasm will not convert one into the other.

Where the brief was vaguer, and what was actually wanted was something indoors and rain-proof, half a day, a clean result and thirty people who had a good time, a station-rotation format in a venue whose attractions keep their own score answers that instead. Call it an alternative and not a replacement: it develops other things entirely, and it feels nothing like three hours at a table with a spreadsheet.

The two combine well across a full day, in that order. Simulation in the morning while people can still concentrate, physical stations after lunch when they cannot. The reverse does not work, because nobody plans well after two hours of running around, and the debrief is the part you lose.

Simulation queries that arrive with the budget request

How much does a business simulation cost per person in Singapore? Published local rates split in two. Event-style acquisition and trading variants publish $40 to $60 per head for 2 to 4 hours, while seated facilitated learning simulations sit in the $120 to $250 specialist band, with one published half-day workshop at $406.

How many decision rounds does a business simulation need? Five is the working minimum for a two-round feedback lag, and six or seven is better. Three rounds cannot teach planning, because an investment made in round one lands in round three with nothing left to prove it.

What is feedback delay in a business simulation? The number of rounds between a decision and the moment its result becomes visible. Zero delay teaches optimisation, because teams tune the next move against the last score. A two-round delay teaches planning, because teams must commit before knowing.

How long does a business simulation take? Three hours for five rounds with a proper debrief. Budget twenty minutes to brief, a practice round, twelve to eighteen minutes per live round including resolution, and forty minutes at the end for the causal walk-back.

What is the smallest group a business simulation works with? Twelve, arranged as three teams of four. A market needs three participants at minimum: with only two tables, every trade has a single possible counterparty, and price discovery shrinks into a private negotiation nobody can walk away from.

How do you stop the finance people from taking over a simulation? Rotate five named roles each round, and give only the market officer permission to leave the table. Require a written one-line reason on every decision slip, signed by two people, and use invented units rather than real financial terms.

Should the market model have hidden variables? One hidden variable, not several. Fully published rules produce arithmetic and a solved optimum. A single unknown, such as total demand each round, forces teams to form and revise a belief, which is the behaviour worth practising.

Why does the trading variant need unequal information? Because gains from trade only exist when teams value the same items differently and cannot see each other's tables. Give every team a private value sheet, and the game stops being arithmetic and becomes asking the right questions.

What does a business simulation actually build? Decision-making with incomplete information, and negotiating trade-offs under a budget constraint. The mechanism is the feedback lag: teams must commit resources before they can see whether the last commitment worked, which is what real planning feels like.

What does a business simulation not build? Anything physical or social. Nobody moves, nobody touches anything, and it is a poor icebreaker because it opens with reading a rule sheet. Teams that already dislike meetings will experience this as a longer meeting with a scoreboard.

What should a simulation debrief cover? The causal chain from an early decision to a late outcome. Read each team's round one written reason aloud, then show the round four standings, then ask what information would have changed that first decision and when it arrived.

How many teams can one market carry? Four to eight, with four to six people each, so sixteen to forty people in a single market. Beyond that, run parallel markets with identical starting conditions and compare final positions, rather than adding a ninth table.

How do you stop the last round turning into a liquidation? Value the assets at the close. Count remaining capacity, quality and cash into the final score, or do not announce which round is the last. Otherwise every team strips its position in the final round and the standings invert.

Can a team that wins the first round still lose? Routinely, and that is the point. In the worked five-round market on this page, the team that takes the largest round one profit by skipping investment leads for two rounds and finishes fourth of five.

Can a company build its own business simulation? Yes, and the resolution spreadsheet is the whole job. Budget three to four days: the model, the parameter tuning, a solo dry run of five rounds, and a live rehearsal with colleagues who will try to break it.

How does a business simulation differ from a route-based race? A race scores tasks completed at stations along a route, and the clock decides it. A simulation has no route: teams sit, submit decisions on a schedule, and a model converts those decisions into a position that compounds across rounds.