validate-wtp

Most founders guess their price. This skill reads it. Every price anyone mentioned in the relevant posts is pulled out by a script, sorted into would pay, already pay and too expensive, converted to a monthly figure, and turned into a range. Where there are too few prices, it says so instead of guessing.

What the user typed after /validate-wtp: $ARGUMENTS

The tool

python3 "${CLAUDE_SKILL_DIR}/wtp.py" [--dir validator] [--theme KEY] [--community NAME] [--all-posts]

If the path is not expanded, use the base directory Claude Code printed for this skill. It needs validator/labels.jsonl (/validate-cluster). It reads posts labelled relevant, not only complaints, because "I pay $99 a year for X and it's fine" is a price signal too.

Step 1: run it three ways

1. All relevant posts: wtp.py
2. The top theme from pain.md: wtp.py --theme <key>
3. The beachhead, if segment.md picked a community: wtp.py --community <name>

Each run overwrites validator/wtp.md and wtp.json, so run the one you will report on last, or copy the numbers you need as you go.

How it reads a price, say it once

• It finds an amount with a currency ($, €, £, ₹, "dollars", "bucks", USD...) and a period ("/mo", "a year", "per seat per month", "weekly").
• It converts to USD a month with the fixed rates printed in wtp.md. A yearly price is divided by 12. A lifetime price, or one with no period, is listed but kept out of the range.
• It sorts each price by the words around it: would pay ("I'd pay"), pays ("I'm paying", "costs me"), too expensive ("overpriced", "nobody would pay"), or a plain mention. Mentions of company funding ("raised $5 million") are never prices.
• The range is the middle half of what buyers pay or would pay. If three or more people object to a price, the top is capped where the objections start. It needs at least 5 buyer prices, or it gives no range at all.

Step 2: tell the user

1. The counts: would pay, pays, too expensive.
2. The range and its confidence, or that there are too few prices to call one.
3. What people pay today: the named tools and amounts, from the table.
4. Where the objections start.
5. How this compares with the price in the hypothesis in idea.json.

Quote only from the "In their words" section of wtp.md. Those snippets passed a word-for-word check. Every amount you name must be in the table.

When there are too few prices

That is a finding, not a failure: people complain about this problem without talking about money. Two ways to get more:

• Mine money talk directly with /validate-mine, for example '"would pay" <problem words>', '"<competitor>" pricing' and '"<competitor>" "too expensive"'. Then label the new posts (batches.py plan --add-new, run the waves, tally) and run wtp.py again.
• Ask in interviews (/validate-interview), where price questions belong anyway.

Never fill the gap with industry benchmarks or your own sense of what similar tools cost, unless you label it clearly as outside knowledge, not data.

Honest limits, say them once

What people say they would pay is not what they pay: treat "would pay" as an upper bound. Prices in complaints skew towards people who already find a tool expensive. A range from fewer than 15 prices is a sketch. The real test is a price on a landing page or a preorder (/validate-smoke).

Output

validator/wtp.md, validator/wtp.json.

Next: /validate-competitors shows what they pay for today and what is missing.