The cheap levers need notice.
Ten weeks of a supply crisis and six ways to respond. Pre-building is cheapest and has to happen before the thing you are responding to. Air freight is fast and capped. The backup supplier is dearest and takes two weeks to wake up. Doing nothing and doing everything are both computed for you, and neither is close to right.
How this game works
Two distribution centres are fed by one plant. Each week you split that week's plant output between them and choose which response levers to pull. Everything you order arrives the following week, which is why timing is the whole game.
- Pre-build — extra output from the plant's spare capacity at $3.40 a unit. The cheapest unit you can buy, and only available while the plant has spare capacity, which is exactly when you do not yet feel you need it.
- Transfer — move stock between the two regions at $2.10 a unit, up to 3,000 a week. It creates nothing; it only puts what you have where the demand is. In transit it serves neither region.
- Air freight — $6.20 a unit, 2,500 a week. Fast, capped, and the lever most likely to be the right one.
- Backup supplier — $120,000 to activate, then two weeks before anything can be ordered, then $8.90 a unit up to 4,000 a week. Activating it in week 1 is expensive insurance; activating it in week 6 is a receipt for nothing.
- Allocation — what share of this week's plant output goes west. The only lever with no cost and no lead time attached.
- Rationing — proportional shares a shortage across the whole order book in proportion to what each part of it asked for. Protect-key serves the key accounts first and pushes the entire shortfall onto ordinary demand.
Key accounts are worth $21 a unit against $11 for ordinary demand, and an account served below 80 percent for three weeks leaves — taking its volume for the rest of the game and a $145,000 exit charge with it. The strain counter next to each account is the thing to watch.
Week 1 of 10
In playWhere you stand
The two regions
Plant output by week
Last week
Service, week by week
Bars are the overall fill rate. Amber is below 95 percent, red below 80.
| Wk | Output | Demand | Fill | Key fill | Stock | Mitigation | Shortage | Week cost |
|---|
The crisis in review
| Response | Total cost | Fill | Accounts kept | What it does |
|---|
What the model is doing
The six levers, on three axes
Every response lever in this game differs from every other on cost, on the notice it needs, and on how much of it you can have. That three-way spread is the whole design, because it is what makes the right answer a combination rather than a choice.
Transfer looks like the cheapest lever on the table and it is the weakest, because it creates nothing. It moves units you already own from a region that has them to a region that does not, and while they are moving they serve neither. It is the right lever for an imbalance and useless against a shortage.
Pre-build is the cheapest lever that creates units, and its ceiling is the plant's spare capacity — which is largest when the plant is running normally and zero when the plant is in trouble. You can only stock up before you need to. That is not a quirk of the model; it is the defining property of buffer stock, and it is why supply chain risk work is about early warning rather than about response capacity.
How a week resolves
Step 4 before step 6 is why the levers always feel one week late. Nothing you decide this week can serve this week's demand. The only decision with effect today is the allocation split and the rationing rule, and neither of them creates a single unit.
Key accounts and the strain counter
Each region has key accounts alongside ordinary demand — three in the west, two in the east, 3,500 units a week between them. They are worth $21 a unit against $11 for everyone else, and they keep a memory. Any week a region's key demand is filled below 80 percent, every active account there gains a strike. Any week it is filled at 98 percent or better, they lose one. At three strikes the largest account in that region leaves: its volume disappears for the rest of the game and it costs $145,000 on the way out.
The largest account goes first because the largest customer is the one with the alternatives, the procurement team and the lawyers. It is also the one whose departure hurts most, which makes the counter worth watching rather than merely worth having.
This is where protect-key rationing earns its place and where it costs you something real. Serving the key accounts first means deliberately failing ordinary customers who are, in the model, simply cheaper to fail. The model does not pretend that is comfortable; it prices it, at $11 a unit, and leaves the judgement where it belongs.
Why both benchmarks are bad
Most games give you a floor to beat and a ceiling to approach. This one gives you two floors, in opposite directions, and the interesting region is between them.
Ride it out pulls no levers at all and splits output between the regions in proportion to what each is short of. On the port closure it costs about $406,000, almost all of it shortage, and loses a key account.
Throw everything at it activates the backup supplier immediately and then fills the air freight, backup and spare-capacity ceilings every single week while protecting the key accounts. It reaches 100 percent service and costs about $992,000 — two and a half times as much as doing nothing — because it buys units into weeks that did not need them and pays a premium for all of them.
A response that sizes each lever to the gap it is actually covering comes in around $170,000 to $250,000 on the same scenario, at high service. The gap between that and either benchmark is not cleverness. It is arithmetic that nobody does in a crisis, because in a crisis the two available modes of thought are denial and panic, and the model is built so that you can watch both of them lose.
The three scenarios, and the one where the answer is to do less
- The port closes. A supply shock with real warning signs in weeks 2 and 3. Output falls to a third in week 4 and bottoms out in week 5. This is the scenario where pre-building pays, because the plant is at full capacity while the warning is arriving and the spare capacity is there to use.
- A demand surge. The plant is fine all the way through; western demand nearly doubles for four weeks. Here the instinct to activate a backup supplier is simply wrong — there is nothing wrong with supply. Air freight covers most of the gap for a fraction of the cost, and there is a real argument, which the numbers will support, for letting some of the surge go unserved rather than buying units at a premium to sell into a spike that is about to end.
- A quality escape. The hardest one. In week 3 a third of the stock sitting in both regions is quarantined, and in weeks 4 and 5 the line is down for re-validation. Pre-building before week 3 protects nothing, because the buffer you built is in the same quarantine as everything else. It is the scenario that punishes the reflex the first scenario rewards, which is the point of having all three.
Assumptions, stated plainly
- Nothing is random. Each scenario is a fixed path. Two players making the same calls get the same result, which is what makes them comparable — and which also means doing well here is not evidence that anyone can read a crisis in real time.
- No backorders. Demand you cannot serve this week is gone, not queued. Real shortages produce a backlog that makes the following week worse, and leaving that out makes every scenario here gentler than its equivalent would be.
- No price response and no substitution. You cannot raise prices to ration demand, and a customer you fail in the west does not buy from the east.
- Telling the truth early is not a lever, because it cannot be represented as a quantity. In practice it is often the cheapest thing available — a customer told in week 2 that they will be short in week 5 can plan; the same customer told in week 5 cannot. Read the absence of that lever as a limitation of the model, not as a claim about what matters.
- The backup supplier is perfectly reliable once live, which is generous. A backup qualified in a hurry during a crisis is not usually the most dependable thing in the network.
- Capacity ceilings are hard and constant. No amount of money buys 2,600 units of air freight in a week where the cap is 2,500.
Where this sits in the practice
The frame here is time-to-recovery and time-to-survive: how long a node takes to come back, against how long the network can keep serving without it. The levers in this game are the standard ones for closing that gap — buffer, flexibility, redundancy and allocation — and they carry their standard property, which is that the cheap ones must be bought before the event and the ones available during it are rationed. The analytics report the week service first fell below 95 percent and the week it came back above 98 for good, because the shape of that curve, rather than the total cost, is usually what a review is actually arguing about.