pricing-experiments — test what to charge

When to use

The user asks "what should I charge?", "can I raise prices?", "how do I test a new plan?", or is deciding between seat / usage / outcome-based pricing. Pricing is the highest-leverage growth lever and the most under-tested.

Process

1. Pick the model first (this constrains the experiment):
   • per-seat — predictable, good for collaboration tools
   • usage — aligns cost with value, good for infra/APIs
   • outcome-based — charge per result delivered; strongest NRR in 2026, but you must be able to measure the outcome
2. Find willingness-to-pay before A/B testing prices:
   • Van Westendorp 4 questions (too cheap / cheap / expensive / too expensive) → plot to find the acceptable price band.
   • Or fake-door / Stripe pricing-page test with real "Buy" intent.
3. Run the test cleanly:
   • Test ONE variable (price OR packaging), not both.
   • Randomize by new visitor, never switch an existing customer's price mid-flight.
   • Decide sample size and stop rule up front; watch conversion AND revenue per visitor (a higher price with slightly lower conversion often wins on revenue).
4. Grandfather existing customers when you raise prices; announce early.

What to produce

• A recommended model with a one-line rationale.
• A concrete experiment: variants, the single variable, the metric (revenue per visitor, not just conversion), sample size, and stop rule.
• A van Westendorp question set if WTP is unknown.

Pitfalls

• Optimizing conversion instead of revenue-per-visitor → you leave money on the table.
• Testing price and packaging at once → uninterpretable result.
• Surprise price hikes on existing users → churn and trust damage.

Report back

State the chosen model, the experiment design, and the decision criterion you'll use to call a winner.