litmus

Price is the fastest lever on the margin and the one founders guess most. This skill sets it from three things you can check: what buyers said, what competitors charge, and what the business needs to earn. Only one thing settles it: a real buyer paying it.

Read ${CLAUDE_SKILL_DIR}/../receipts/rules.md first. If the ledger has real price evidence (a deposit, a paid pilot, a buyer telling you what they pay today), it outranks every simulated answer below.

Tool in this folder:

python3 "${CLAUDE_SKILL_DIR}/van_westendorp.py" founder/panel/answers --price 6.50 --out founder/pricing-curve.md --handoff founder

Handoff contract (stage 04 in founder/CONTEXT.md)

This skill sits between the panel and the CFO, which is where a simulated number most easily turns into a "fact". Rule 10 in rules.md: a number takes the weakest label of anything it was computed from.

Inputs

Process: read the range, set it against competitors and the margin, pick a price.

Outputs

Step 1: what buyers said

If /truth-serum ran, every buyer answered four price questions. Run the tool on founder/panel/answers (or a CSV with the columns too_cheap, cheap, expensive, too_expensive from real surveys, which beat simulated answers every time). It reports:

PMC to PME is the acceptable range. If the panel has not run, say so and offer to run /truth-serum --quick first. Do not invent a range.

Step 2: what competitors charge

From founder/competitors.md (or look them up now, from their own public price pages, with links): the price of the nearest substitutes for the same job. Note what each price includes, so you compare like with like. A matcha latte is compared with other matcha lattes and with the coffee it replaces, not with a tea bag.

Step 3: what the business needs

From founder/numbers.json: the variable cost per unit and the break-even. Run the CFO's tool at two or three candidate prices to see the margin and break-even at each:

python3 "${CLAUDE_SKILL_DIR}/../burn-test/unit_economics.py" founder/numbers.json --price 7.00

Step 4: decide

Write founder/pricing.md:

0. The hard truth: how much of this price rests on real buyers, and its label. For example: "This price comes from 20 simulated buyers and two competitor pages, so it is SIMULATED, and so is every margin worked out from it. No one has paid it yet."
1. The price, and why: where it sits in the acceptable range, against the competitors, and the margin it gives (from the tool, not a guess). Front end or back end: say which this price is for. The first product exists to win customers (Ready, Fire, Aim, rule 11 in rules.md): when testing its price, lean towards the lower end of the range, because more customers sooner usually beats more margin per first sale, as long as the price stays above the variable cost and what a customer is worth covers what it costs to win one (/burn-test, "What one customer is worth"). Higher prices belong to the later products sold to customers who already trust you.
2. The ladder, if the business can have one: a good / better / best (size, bundle, subscription, membership) with a reason to step up each rung.
3. The opening offer: an intro price or bonus for launch that does not train buyers to wait for discounts, with an end date.
4. What to test with real buyers: two prices inside the range and how to test them (two landing pages, a pre-sale at each, A/B on the menu board). For a business sold to businesses, the first price is often a pilot: a small deposit credited to the first month, or free for a fixed time in exchange for named things (their numbers, staff time, a quote if it works). A large setup fee asked of a stranger with no references is the most common reason for an early no; say so if the plan has one.
5. The panel's verbatim price objections (top three), so marketing can answer them.

Rules

• Price answers from simulated buyers pick what to test. They are not proof.

• Log the range first, then the recommended price computed from it, so the price inherits the weakest label instead of being called PLANNED:

  python3 "${CLAUDE_SKILL_DIR}/../receipts/ledger.py" add --claim "Panel acceptable price range" --kind price --status SIMULATED --use PRIVATE --number "$5.80 to $7.40"
  python3 "${CLAUDE_SKILL_DIR}/../receipts/ledger.py" add --claim "Recommended price" --kind price --use PRIVATE --number "$6.50" --from c07,c03

  The second line has no --status: the ledger gives it the weakest label of c07 and c03 (here SIMULATED), and refuses a stronger one. Then write the new id into founder/numbers.json as "sources": {"price": "c08"}.

• A price picked by hand inside a simulated range is still computed from it. Rounding it, or calling it "our decision", does not make it PLANNED.
• The price becomes MEASURED only when a real buyer pays it: log that sale as its own claim with an evidence id and point sources.price at it.
• Never price below variable cost to "buy" customers without saying so, and how long the CFO says the business can afford it.
• No fake anchors: no "was $X" price that was never charged.

Output

founder/pricing-curve.md, founder/pricing.md, the range and price claims in the ledger, and sources.price in founder/numbers.json. If the price changed, update founder/numbers.json and re-run /burn-test. Next: /dealbreaker.