Monte Carlo for Total Rewards.
Everything from the session, in one folder. Both prompts, the intake sheet, the validation checklist, a glossary, a worked example with every input labeled measured or assumed, and the scripts if you have Claude Code. None of it requires you to write code.
The take-home pack
ZIP, about 1.2 MB. Nothing to sign up for, nothing to install to read it.
Read the assumption list first, then the answer.
Every model makes choices you did not give it. The prompts in this pack end with one line that forces those choices into the open: "list every assumption you made that I didn't give you." That single sentence is the difference between a number you can defend and a number you are hoping is right.
Six things, none of them code.
Start with the intake sheet. It walks you through describing your own plan in plain English, which is the actual work. The rest supports it.
Both prompts
The relative TSR prompt and the bonus plan prompt, ready to paste. They read like a brief you would hand a consultant, not like code.
prompts/rtsr-psu-prompt.txtprompts/bonus-plan-prompt.txt
The intake sheet
Describe your own plan: the curve, the threshold, the cap, the comparator group, the averaging convention. Fill this in and you have your prompt.
intake-sheet.md · .docxThe validation checklist
The three checks to run before you trust any number, plus what a failure actually means. Run them every time.
validation-checklist.md · .docxA glossary
Plain English for the terms that get used loosely: fair value versus forecast, risk-neutral drift, percentile method, correlation.
glossary.md · .docxThe worked example
The full Chipotle run: result card, distribution chart, comparator chart, every measured input separated from every assumed one, and the fact-check behind it.
worked-example/result_card.htmlThe scripts
The engine, the what-if flags, and the bonus plan model. Optional. Everything above stands on its own without them.
scripts/run_demo.py · whatif.pyOne award, the whole shape.
A hypothetical three-year relative TSR award on public Chipotle data, ranked against 16 peers. The committee asked for one number. Here is what the distribution actually looked like.
| What we asked | What came back |
|---|---|
| Grant date fair value | $44.97 per target share, 135% of the stock price |
| Chance it pays nothing | 29% of futures |
| Chance it pays target or better | 49% |
| Chance it pays the maximum | 27% |
| If peers moved independently | $47.39, up 5.4% |
Fair value probabilities are valuation measures, not forecasts. The pack labels both, and so should you. Everything here is public data and a demonstration, not a critique of any company's program.
Want to talk it through?
If you are trying this on your own plan and get stuck, or you want a second set of eyes before it goes to a committee, I am happy to help.