Pricing a new product? Work through these in order — each step constrains the next, and each links to its derivation. A step that surprises you is the one to go read in full.

#The nine steps

• Arithmetic first. Estimate the free-user cost m (especially AI), plausible ARPU, churn class, CAC per channel. Does any configuration clear payback < 3 months solo (funded: 12–18 months, LTV/CAC ≥ 3)? If not, restructure before building. → the arithmetic
• Position. Whose alternative is expensive? Position against it; ceiling ≈ 10–30% of the quantified value. If the alternative is a competitor: premium-narrower or sideways, never cheaper-same-job; never match a funded competitor’s subsidies. → Decision 1
• Who pays. B2C / bottom-up / top-down → price against the expense thresholds; the price picks the sales motion. Stay out of the dead zone. → Decision 2
• Value metric. Hybrid default — flat base + prepaid credits on the cost-bearing dimension. Seats only if value is genuinely per-human. Something must be able to expand, or NRR is capped below 100% and the base can only shrink. → Decision 3
• Measure. 10–15 interviews — past purchases and current spending, never hypotheticals; sort leaders/fillers/killers; Van Westendorp as a prior; fake door with real prices for anything over two weeks of build. Days, not weeks. → Decision 4
• Model. Recurrence from the shape of value and cost: subscription default, hybrid credits when AI COGS, pay-once/Sketch-model only when your costs are one-time too. Start: trial by default; reverse trial if a free tier is justified; freemium only with a named, quantified mechanism. A free tier gets a dollar budget or doesn’t exist. → Decision 5
• Gates. Value → account → payment. Usage-gated trial per workspace, time backstop. Receiving side of collaboration free forever. Soft limits at the moment of impact. → Decision 6
• Tiers. Three plus an enterprise anchor, segment-shaped, leaders gating, expansion dimension chosen on purpose. Every limit passes the explain-it-to-their-face test. → Decision 7
• Plan the raise now. New customers continuously (your elasticity probe); existing customers annually, with justification, notice, and time-boxed grandfathering. Price toward value delivered, never toward cost of leaving. → Decision 8

#The thresholds

• Payback ceiling, solo: under ~3 months. Funded: 12–18 months, LTV/CAC ≥ 3.
• Price ceiling: 10–30% of the quantified value gap over the alternative.
• Approval bands (B2B, monthly): under ~$50 goes on a corporate card; $50–500 needs a manager; past $1–2k means procurement. Between card and procurement-worthy sits the dead zone — don’t price in it.
• Elasticity probe: raise new-customer prices 20%; if conversion drops less than 20%, you were underpriced — raise again.
• Free tier: rational only if m × months free < P(convert) × LTV + referral value. Give it a dollar budget or don’t have one.
• Interviews: 10–15 in the target segment, budgeted in days, never weeks.
• Tiers: three buyable, plus an “Enterprise — talk to us” anchor.
• Annual discount: 15–20% (≈ two months), with the toggle defaulted to annual.
• Raising prices: justification + 60–90 days notice + “lock your price by going annual.” Grandfather for a defined period (e.g. 12 months), never forever. Single-digit raises yearly beat a big correction after five frozen years.
• Regional pricing: tolerate VPN arbitrage below ~5%; check every regional price still clears your inference cost.

Dashboarding these against reality — conversion benchmarks, customer counts per revenue target, valuation multiples — is the benchmarks.