How to simulate a pricing decision with Claude Code (free example)
Ask an AI "should I charge $29 or $99?" and you get a confident guess. What you actually need is a range: how often each price wins, how bad the bad case is, and which number matters most. That's what a Monte Carlo simulation gives you, and Claude Code can run one for you in seconds.
The idea in one minute
You don't know your exact conversion rate or traffic. Fine. Instead of one number, you give each unknown a range. The simulator picks random values from those ranges 100,000 times and plays out 100,000 possible futures. Then it counts: which price won most often? How often did you hit your goal?
The free scenario (copy it)
Save this as price-test.json:
{
"question": "Which price makes more money in 3 months?",
"unit": "$", "goal": 3000, "higher_is_better": true,
"shared": {"visits": "logn 1500 0.9"},
"options": {
"$29": {"vars": {"conv": "range 0.02 0.04"}, "formula": "visits*conv*29"},
"$49": {"vars": {"conv": "range 0.013 0.03"}, "formula": "visits*conv*49"},
"$99": {"vars": {"conv": "range 0.006 0.016"}, "formula": "visits*conv*99"}
}
}
visitsis shared: every price gets the same traffic luck in each future. That keeps the comparison fair.logn 1500 0.9means "around 1,500, but it could be much more or much less". Traffic and money behave like this.- Higher prices get a lower conversion range. That's the real trade-off.
What you get back
🏆 $99 wins 47% of runs · hits goal 22% · likely $1.5K (bad $436, good $4.8K) · biggest lever: visits · ⚠️ close call — get real data first
Read it like this:
- $99 wins 47% of the futures. Better than the others, but far from certain.
- Only 22% hit the $3,000 goal. The goal is a stretch at this traffic.
- Biggest lever: visits. Price matters less than getting more people to the page. That's often the most useful line of all.
- Close call means: don't bet everything. Run a cheap real test first.
Make it honest with real numbers
Once you have data, replace a guess with a measured rate: "conv": "rate 12/900" (12 sales from 900 visits). The simulator turns small samples into honest doubt automatically, so 1 sale out of 20 doesn't look like a 5% sure thing.
Ask Claude to attack it
The biggest risk in any sim is hopeful numbers. Ask Claude: "Act as a skeptic. Which of these ranges are too hopeful for a cold landing page?" Typical honest ranges: cold page conversion 0.5–3%, cold email replies 1–5%, refunds on digital products 3–12%.
Want this as a one-word command?
The Sim Your Decisions kit adds a /sim command to Claude Code, the simulator itself (pure Python, $0 per run), a Modeler, a Skeptic and a Scorekeeper agent, and 5 ready models (price, hiring, ad budget, launch timing, channels).
More guides · AI Crew Kits for Claude Code. Not affiliated with Anthropic.