Capacity is a bet on a forecast.
Three plants, eight rounds, one growing market. The cheapest plant takes three rounds to build a line and two more to ship what it makes, so by the time you know you needed the capacity it is five rounds too late to start. Commit anyway.
How this game works
Every round you decide two things at each of three plants: how many capacity lines to start or mothball, and how many units to produce on the lines you already have. Then demand for the round is revealed and you sell what has reached the market.
- Build lead. A line started this round comes online after the plant's build lead — one round at Kenosha, two at Monterrey, three at Penang. Until then it costs you the capital and produces nothing.
- Ship lead. What a plant makes reaches the market after its ship lead — immediately from Kenosha, one round from Monterrey, two from Penang. Produce at Penang in round 3 and it sells in round 5.
- Fixed cost. Every line online bills you every round whether or not it produces anything. Idle capacity is not free.
- Overtime. Each plant can run above its nameplate for an extra premium per unit. It is the fastest capacity you have and the dearest.
- Mothballing pays a refund now and takes the line offline from the next round — you finish the run you are in the middle of.
- The forecast is not the demand. Future rounds show a forecast. It is wrong in a different way in each scenario, and that is the point.
Cash is never a constraint: you can always afford the build. That is deliberate, so the exercise stays about lead time rather than about the budget.
Round 1 of 8
In playWhere you stand
Demand by round
Last round
Round by round
| Round | Forecast | Demand | Sold | Lost | Fill | Capacity | Produced | Stock | Profit | Cumulative |
|---|
How each plant was used
| Plant | Produced | Share | Utilisation | Lines built | Mothballed | Cost per unit |
|---|
The year in review
| Plan | Profit | Fill rate | What it does |
|---|
What the model is doing
The order of a round
The sequence is fixed, and it is the part worth reading twice, because every apparent unfairness in the game comes from it:
Step 2 is why a line you pay for this round cannot produce this round, and step 3 is why production is not the same thing as supply. A unit made at Penang in round 6 is a unit sold in round 8, and if the market has turned by then it is a unit of inventory instead.
What a round costs
The lost-sale penalty sits on top of the margin you did not earn. It stands for the order that went to somebody else and the customer who remembers, and it is why running short is dearer than the gross margin alone suggests.
The three plants
Kenosha costs the most per unit and is the only plant whose output is on the market the same round it is made. One round to build a line. It is the plant you use when you were wrong and need to be less wrong quickly, and using it is the tuition fee for the forecast.
Monterrey is the middle of every axis: two rounds to build, one round of transit, a unit cost between the other two. It rarely wins an argument on its own and it is frequently the right answer.
Penang is a third cheaper per unit than Kenosha and five rounds of lead time away when you count the build and the boat. Committing to it is committing to a view about round 8 in round 3.
Why the forecast is wrong on purpose
Each scenario carries two demand series: the one you are shown for rounds that have not happened, and the one you are graded against. In Steady climb they are close, and the exercise is purely about lead time. In Spike and settle the forecast misses a two-round surge by 45 percent and then refuses to believe it, so the capacity that would have caught the peak has to be committed on a rumour. In Boom, then bust the forecast follows demand down a round late, which means the signal to stop building arrives after the commitment is made — and the lines you built in round 3 come online in round 6, into a market that has halved.
That last case is the one worth sitting with. Nothing in it is a mistake in the ordinary sense. Every decision was defensible with the information available at the time, and the year still ends badly. Capacity planning is largely the business of being wrong at a manageable cost.
The two benchmarks
Never build keeps the opening four lines all year and produces to the forecast, cheapest plant first. It is the floor: what the year looks like if capacity is simply not a decision anybody makes. It does surprisingly well in the bust scenario, for the obvious reason.
Best fixed plan, with hindsight searches every reachable line count at every plant, commits to it in round 1, and then produces against demand it already knows. No manager has that information. It is not a target so much as a ceiling on what a single up-front commitment can be worth — and a player who beats it has done something a static plan cannot do, which is change their mind partway through.
Assumptions, stated plainly
- Cash is unlimited. Real capacity decisions are usually rationed by capital, and adding that constraint would change which answer is right. It is left out so the lesson stays on lead time.
- One product, one market, one price. No mix, no segments, no price response to a shortage.
- A line is a line. No learning curve, no ramp-up, no difference in quality or yield between plants, no hiring or training time.
- Demand not served is gone. No backorders and no queue: a customer you cannot supply this round does not come back next round.
- Transit is reliable. Ship leads are exact. Nothing is late, lost or stuck at a port — which is the subject of a different game on this site.
- The horizon ends at round 8 and inventory left at the end is worth nothing. That is what makes the last two rounds different from the first six, and a rolling business does not work that way.
Where this sits in the literature
- The trade-off between a plant's cost position and its response time is the substance of the capacity cushion and capacity timing discussion in any operations strategy text — whether to lead demand, lag it, or straddle it. This game is a lead-versus-lag argument with the lead times made explicit rather than assumed away.
- The structure of committing scarce, long-lead capacity against an uncertain forecast is the same structure as the newsvendor problem, repeated over several periods and with the order quantity replaced by a durable asset. If the single-period version is unfamiliar, play that one first.