Pricing Strategy skill for Claude Code

Done well means the price scales with a metric customers recognize as value, each tier has a clear buyer and a reason to upgrade, the numbers come from willingness-to-pay evidence and unit economics rather than competitor copying, and every change ships with a rollout plan for existing customers and a measurement of what it did to conversion, expansion, and churn.

Ground rules

• Packaging before price points. Wrong tiers at the right price still fail.
• Price on value delivered, bounded below by cost to serve and checked against alternatives customers actually compare you with (including spreadsheets and doing nothing).
• One recommended structure, not a menu of five. State the main risk and the fallback.
• Never A/B test different prices for the same plan shown to similar visitors at the same time without the user deciding that consciously: it creates fairness complaints, support load, and in some jurisdictions consumer-protection exposure. Prefer the tests in step 6.
• Price changes for existing customers, removed plans, and contract terms are business and legal decisions. Draft them, then get explicit user confirmation before anything is announced or changed in billing.

Workflow

- [ ] 1. Gather facts: segments, current plans, unit costs, conversion and churn by plan
- [ ] 2. Choose the value metric
- [ ] 3. Design packaging: tiers, fences, add-ons
- [ ] 4. Research willingness to pay
- [ ] 5. Set price points and the discount policy
- [ ] 6. Choose a test or rollout method
- [ ] 7. Plan the migration of existing customers
- [ ] 8. Write the pricing decision document and measurement plan

1. Facts first

Collect before proposing anything:

• Segments: who buys (solo, team, mid-market, enterprise), how they buy (self-serve card, sales-assisted, procurement), and what each currently pays (ARPA by segment).
• Funnel by plan: visitor to trial or free signup, trial to paid, free to paid, upgrade rate, logo churn and net revenue retention (NRR) by plan and cohort.
• Unit economics: cost to serve per account and per unit of usage (infrastructure, model inference, support), current gross margin.
• Discount reality: list price vs realized price across the last 50-100 deals.
• Alternatives: the 3-5 products or workarounds prospects mention in sales calls and churn surveys, with their real prices.

2. Value metric

The value metric is the unit the price scales with. Test candidates against four questions:

Test	Good sign	Bad sign
Customers understand it	They can predict their bill	They need a calculator and still get surprised
It grows with value received	Bigger outcomes mean more units	Heavy users pay the same as light users
It is hard to game	Sharing a login or batching does not cut the bill	One seat shared by 10 people
It is cheap to measure and bill	Already tracked reliably	Needs new metering you do not trust

Common choices and where they fit:

Metric	Fits	Watch out for
Per seat	Collaboration tools where each user gets value	Discourages adoption; seat sharing
Per active user	Tools with uneven usage across a company	Needs a clear “active” definition
Usage (API calls, messages, GB, minutes)	Infrastructure, APIs, communications	Bill shock; forecasting for buyers
Outcomes (invoices processed, leads, resolutions)	Clear, countable results	Disputes over what counts
Credits	AI products with varied cost per action	Opaque value; needs a published credit table
Flat tiers with limits	Simple SMB products	Leaves expansion revenue on the table

Hybrid is the 2025-2026 default for AI-heavy products: a platform fee or seats for access plus included usage, then metered overage or credit packs. Check that the heaviest 5 percent of accounts stay above the target gross margin at the proposed price.

3. Packaging

Default to three self-serve tiers plus enterprise, each with a named buyer:

Tier	Buyer	Job
Entry	Individual or small team proving value	Low-friction start; limited scale
Core (the target)	The typical buying team	Where most new revenue should land; visually recommended
Advanced	Teams needing control and scale	Admin, permissions, integrations, higher limits
Enterprise	Procurement-led buyers	SSO and SCIM, audit logs, data residency, SLAs, invoicing, custom terms

Fencing rules:

• Fence on things that correlate with willingness to pay: scale (usage limits, seats), control (roles, audit, SSO), and business-critical integrations. Do not fence core value so hard that the entry tier cannot show what the product does.
• Each tier should have 2-4 headline differences, not 30 checkmarks.
• Put features only a minority needs, with real cost behind them, into add-ons.
• Target mix after launch: roughly 50-70 percent of new paid accounts on the core tier. If most land on entry, the entry tier is too generous or the core tier too expensive.

Free trial vs freemium:

Choose	When
Free trial, 14 days	Value shows within a few sessions, self-serve product
Free trial, 30 days or sales-assisted pilot	Value needs setup, data import, or team rollout
Freemium	Low marginal cost, network or viral effects, large market, clear upgrade triggers
Reverse trial (full features, then drop to free)	You want freemium reach with trial urgency

Requiring a card at trial start typically cuts trial signups sharply but raises trial-to-paid conversion; decide on which number the funnel can afford to lose.

4. Willingness to pay

Use at least two methods; no single method is reliable alone.

Method	Use for	Minimum sample
Customer interviews (value, alternatives, budget owner)	Finding the value metric and fences	10-15 per segment
Van Westendorp price sensitivity (four price questions)	Acceptable price range for a defined package	100+ responses per segment
Gabor-Granger (purchase likelihood at set prices)	Demand curve around candidate points	100+ per segment
Conjoint or MaxDiff	Which features drive tier choice	200+, needs a survey tool that supports it
Win/loss and discount data	What buyers actually pay	Last 50+ deals

Describe the package concretely in surveys; asking “what would you pay for software that does X” without specifics produces noise. Survey answers skew low for consumers and sometimes high for B2B; treat them as ranges, not a point.

5. Price points and discounts

• Set the core tier first, from WTP research and the value delivered (a common sanity check: price at 10-20 percent of the measurable value or cost saved for the customer).
• Space tiers so each step is a meaningful jump, commonly 2-3x between adjacent tiers.
• Self-serve SMB: charm endings ($49) are fine. Enterprise and procurement buyers: round numbers.
• Annual billing discount: 15-20 percent (about two months free) is the norm; more than 25 percent trains customers to expect deep discounts.
• Gross margin floor: 75-80 percent for classic SaaS; AI-heavy products often plan for 55-70 percent and must model inference cost per heavy user.

Discount policy (write it down, enforce it in the CRM):

Discount	Who approves	Conditions
Up to 10 percent	Account executive	Annual prepay or multi-year
10-20 percent	Sales manager	Volume or competitive displacement, documented
Over 20 percent	Finance or founder	Strategic logo, written justification, expiry date

Trade every discount for something: longer term, upfront payment, case study, faster signature. Make discounts expire at renewal or step down over time, otherwise they become permanent.

6. Testing and rollout methods

Method	Use when	Caveat
New-customers-only change, before/after	Most self-serve changes	Control for seasonality and marketing changes
Geographic or segment rollout	Distinct markets with separate pricing pages	Leakage via VPNs and word of mouth
Packaging test (feature placement, not price)	Tier contents uncertain	Still measure revenue per visitor, not just conversion
Sales-led price test	Quoted deals	Track win rate and sales cycle by price band
Simultaneous price A/B	Huge traffic, consciously accepted risks	Fairness and legal risk; user must approve

Primary metric: revenue per visitor or per trial over a full sales cycle, not signup conversion alone. Guardrails: trial-to-paid, 90-day logo churn, refund rate, support tickets about pricing. Hand experiment mechanics (sample size, run length) to conversion-optimization.

7. Existing customers

• Grandfather existing customers for a fixed period (commonly 6-12 months) or permanently on legacy plans; decide deliberately, permanent grandfathering accumulates billing complexity.
• Give 30-90 days written notice for increases (check contract terms and local consumer law), explain what improved since they signed up, and offer annual lock-in at the old price.
• Never remove a feature from a paying customer’s plan silently.
• Expect some churn; model it. An increase is usually worth it if new_price x (1 - extra_churn) > old_price.

Gotchas

• Copying a competitor’s price imports their cost structure and positioning, not yours.
• Too many plans (more than four self-serve) slows decisions and raises support load.
• Usage pricing without spend alerts, caps, or budgets produces bill shock and chargebacks.
• Unlimited plans on AI products invite a few accounts to consume most of the inference budget.
• Changing price, packaging, and the pricing page design at once makes the result unreadable.
• Localized pricing needs purchasing-power and tax handling (VAT-inclusive display in the EU and UK for consumers); hand billing details to stripe-payments or merchant-of-record-billing.

Verification

• Every tier has a named buyer, 2-4 headline differences, and a modeled gross margin at the heaviest 5 percent of usage.
• The recommendation cites at least two WTP sources and the discount data.
• After rollout, compare revenue per visitor or trial, plan mix, NRR, and 90-day churn against the pre-change cohort; if revenue per visitor falls more than the guardrail for 4 weeks, revert for new customers and revisit packaging.

Output template

## Pricing recommendation — <product> (YYYY-MM-DD)

**Value metric:** <metric>, because <evidence>
**Model:** <seats / usage / hybrid / credits>

| Tier | Buyer | Price (monthly / annual) | Headline differences | Limits | Margin at p95 usage |
|---|---|---|---|---|---|

**Trial / free:** <choice and reason>
**WTP evidence:** <methods, sample sizes, ranges found>
**Discount policy:** <bands and approvals>
**Rollout:** <method>, metric, guardrails, decision date
**Existing customers:** <grandfathering, notice period, message draft>
**Main risk and fallback:** ...