Full pricing strategy: Model Selection, Van Westendorp WTP, Competitive Pricing Map, Good/Better/Best Architecture, Sensitivity Analysis, and AI Cost Modeling — with evidence tiers and confidence scoring.
CodeLens AI's flat $49/month pricing is structurally broken: the top 10% of users (820 accounts, 500+ reviews/month) consume 62% of AI compute but pay the same as casual users doing fewer than 20 reviews/month H internal usage data, T1. At $0.03 per AI-assisted review, power users cost $15.00+/month in marginal AI inference alone — a 31% cost-of-revenue on a $49 price point — while casual users cost under $0.60/month H cost structure analysis, T1. The flat rate subsidizes heavy users at the expense of margin and undercharges relative to the value delivered: a single caught production bug is worth $2,000-$10,000 in incident response costs avoided M customer interviews, T3.
The recommended pricing model is a hybrid (base platform + usage overage) with a Good/Better/Best tier structure: Starter at $29/month (50 AI reviews included), Professional at $69/month (250 AI reviews included), and Enterprise at $149/month (1,000 AI reviews included), with overage at $0.12/review. This captures 54% more revenue in year one ($7.4M projected ARR vs. $4.8M current) while actually reducing the effective per-review price for casual users — Starter costs $29 vs. $49 today — lowering churn risk in the bottom 50% of the base M sensitivity model, T4-T5. The recommended action: launch the new pricing for new customers on Day 0, grandfather all existing customers at $49/month for 6 months, then migrate them to the tier matching their actual usage with a 20% loyalty discount for 12 months.
Before applying pricing frameworks, verify this is a pricing problem — not a product, activation, or positioning problem wearing a pricing mask.
| # | Gate Question | Assessment | Result |
|---|---|---|---|
| 1 | Is the core problem actually about pricing/monetization? | ✓ Yes — margin compression from AI costs at high usage, revenue under-realization from power users, flat-rate model misaligned with variable cost structure | Proceed |
| 2 | Do you have evidence about what customers value? | ✓ Yes — 8,200 paying customers, usage data segmented by review volume, 47 customer interviews conducted in Q4 2025 | Proceed |
| 3 | Do you know your cost structure (marginal cost per unit)? | ✓ Yes — $0.03/AI review (GPU inference), $0.002/static analysis check, $8.50/mo fixed infrastructure per account | Proceed |
| 4 | Is this a pricing decision or a positioning decision? | ✓ Pricing — competitive position is established (mid-market AI code review), product-market fit is validated (8,200 customers, 92% monthly retention), the question is monetization model and tier structure | Proceed |
Context Fitness: This is a packaging redesign with usage-based pricing design. Primary frameworks: Model Selection, WTP Assessment, Competitive Pricing Map, Package Architecture, Sensitivity Analysis, Revenue Impact Modeling. Full depth on AI/SaaS-specific patterns due to variable AI inference costs.
Flat $49/mo for all users. $4.8M ARR. 8,200 customers. 92% monthly retention. No usage limits. AI costs eating margin on power users.
Top 10% cost $15+/mo in AI compute (31% cost-of-revenue) while paying $49. Bottom 50% cost <$0.60/mo but find $49 expensive for <20 reviews/mo. Single price fails both segments.
Usage-aligned pricing that captures value from power users, reduces barrier for casual users, protects margins as AI usage scales, and minimizes churn during migration.
| Segment | % of Base | Accounts | Monthly Reviews | AI Cost/Account | Current Revenue | Margin/Account |
|---|---|---|---|---|---|---|
| Power Users | 10% | 820 | 500-2,000+ | $15.00-$60.00 T1 | $49/mo | $49 - $15 to -$11 = negative at top end |
| Regular Users | 40% | 3,280 | 50-200 | $1.50-$6.00 T1 | $49/mo | $34.50-$39.00 |
| Casual Users | 50% | 4,100 | <20 | $0.06-$0.60 T1 | $49/mo | $39.90-$40.44 |
Key insight: The top ~200 accounts (2.4% of base) doing 1,000+ reviews/month are margin-negative — each costs more in AI inference than they pay. These accounts generate the most value from CodeLens (catching critical bugs) and would pay more, but flat pricing gives them no reason to. Meanwhile, 4,100 casual users are cross-subsidizing power users while perceiving $49 as too expensive for their light usage.
The foundational structural decision — choosing the wrong model constrains every downstream pricing choice. Pricing Model Selection is the framework for determining whether to charge per seat, per usage, per outcome, or a hybrid.
| Criterion | Per-Seat | Flat-Rate (Current) | Usage-Based (Pure) | Hybrid (Base + Usage) |
|---|---|---|---|---|
| Value alignment | Partial — value is per-review, not per-seat T4 | None — same price for 5 and 2,000 reviews T1 | Strong — price scales with value delivered T4 | Strong — base covers platform value, usage covers AI value T4 |
| Cost structure fit | Poor — AI cost is per-review, not per-seat T1 | Poor — variable costs with fixed revenue T1 | Strong — cost and revenue scale together T1 | Strong — base covers fixed costs, usage covers variable T1 |
| Buyer predictability | High — fixed monthly cost T4 | High — same price every month T1 | Low — variable bills cause anxiety T3 | Medium — predictable base + capped overage T4 |
| Revenue predictability | High T4 | High T1 | Medium — usage varies monthly T4 | High — base provides floor T4 |
| Expansion mechanic | Add seats only T4 | None — no expansion within account T1 | Natural — more usage = more revenue T4 | Both tier upgrades and usage growth T4 |
| Competitive norm | Common in dev tools T2 | Common for simple tools T2 | Emerging norm for AI tools T2 | Industry trend (Copilot, Cursor) T2 |
| Overall Fit | Workable | Poor Fit | Workable | Strong Fit |
Selected model: Hybrid (Base Platform + Usage Overage) H — because CodeLens delivers two distinct types of value: (1) the platform itself (repository integration, dashboard, configuration) which justifies a base fee, and (2) AI-powered code reviews which have variable cost and variable value per unit, justifying usage pricing. Pure usage-based was rejected because it creates budget unpredictability for engineering managers and eliminates the revenue floor. Pure per-seat was rejected because value is per-review, not per-user.
Flat-Rate (Current Model): Structurally broken. Creates margin-negative accounts at high usage and overcharges casual users. No expansion mechanic. Only advantage is simplicity, but simplicity that loses money is not a feature.
Pure Usage-Based: Strong cost alignment but creates "meter anxiety" in engineering teams T3. 14 of 47 interviewed customers (30%) cited "predictable billing" as a top-3 purchase criterion. Engineering managers who face variable bills need budget approval monthly instead of annually — a procurement friction multiplier.
Per-Seat: Common in dev tools (GitHub, Jira), but CodeLens value is per-review-line, not per-developer. A team of 5 doing 2,000 reviews/month pays the same as a team of 5 doing 50 reviews/month. Ignores the cost structure entirely. Would work as a secondary metric (per-seat tiers) but not as the primary pricing axis.
Van Westendorp Price Sensitivity Meter — the four-question survey method for finding the acceptable price range — applied to 47 customer interviews and supplemented with competitive price inference.
Price ($/mo)
$200 ──────────────────────────────────────────────────────
│ Too Expensive ╲
│ ╱ ╲
$150 │ ╱ ╲
│ ╱ Expensive ╲ │
│ ╱ ╱ ╲ │
$100 │ ╱ ╱ ╲ │
│ ╱ ╱ PME=$119 ╲ │
│╱ ╱ ╲ │
$50 │ ╱ OPP=$58 IDP=$72 ╲ │
│ ╱ ╲ │
│╱ PMC=$31 Good Value ╱ Too Cheap │
$0 ─┼────────────────────────────────────────────────────────
$0 $25 $50 $75 $100 $125 $150 $175 $200| Van Westendorp Metric | Value | Interpretation | Evidence |
|---|---|---|---|
| Point of Marginal Cheapness (PMC) | $31/mo | Price floor — below this, buyers question quality of AI review | T3 n=47 interviews |
| Optimal Price Point (OPP) | $58/mo | Least price resistance — market "sweet spot" for mid-tier | T3 n=47 interviews |
| Indifference Price Point (IDP) | $72/mo | Market normative price — what customers consider "the going rate" | T3 n=47 interviews |
| Point of Marginal Expensiveness (PME) | $119/mo | Price ceiling — above this, demand drops sharply for standard tier | T3 n=47 interviews |
| Segment | Floor | Sweet Spot | Ceiling | Evidence | Confidence |
|---|---|---|---|---|---|
| Solo developers / freelancers | $19/mo | $29/mo | $49/mo | T3 12 interviews; compare GitHub Copilot $19/mo T1 | M |
| Small teams (2-10 devs) | $39/mo | $69/mo | $99/mo | T3 20 interviews; company spend benchmarks T4 | H |
| Mid-market teams (11-50 devs) | $89/mo | $149/mo | $249/mo | T3 11 interviews; procurement budgets T4 | M |
| Enterprise (50+ devs) | $199/mo | $349/mo | Custom | T3 4 interviews; enterprise dev tool budgets T4 | L EVIDENCE-LIMITED |
CodeLens is underpriced for its core market. The OPP ($58) is 18% above current pricing ($49), and the IDP ($72) is 47% above. Small teams — the largest paying segment — show a sweet spot of $69/mo, meaning the current $49 price is 29% below what the market considers normal for this value. We are leaving $1.6M+ in annual revenue on the table by pricing below the market's own willingness-to-pay anchor H.
Value anchor analysis: When asked "what would you do without CodeLens?", the three most common responses were: (1) add one more senior engineer to review rotation ($12,000/month fully loaded cost) T3, (2) use basic linting only and accept more bugs shipping ($2,000-$10,000/month in incident costs) T3, (3) use free open-source static analysis tools with no AI (CodeClimate free tier, SonarQube) T3. The most cited alternative cost ($12,000/month senior engineer) implies CodeLens at $69-$149/month is a 99% cost reduction — strong pricing power.
Where CodeLens sits relative to alternatives — and what that positioning implies for pricing power. Competitive Pricing Map is the framework for plotting price-value positioning against all alternatives including "do nothing."
| Competitor | Model | Entry Price | Mid Tier | Top Tier | AI Reviews? | Positioning | Evidence |
|---|---|---|---|---|---|---|---|
| SonarQube Cloud | Tiered flat + LOC | Free (up to 500 LOC) | $30/mo (100K LOC) | $450/mo (20M LOC) | ✗ Static only | Penetration | T1 |
| Codacy | Per-seat | Free (up to 5 users) | $15/user/mo | Custom | ⚠ Basic ML | Parity | T1 |
| DeepSource | Per-seat | Free (open-source) | $12/user/mo | $35/user/mo | ⚠ Autofix (basic) | Penetration | T1 |
| Snyk Code | Per-seat + usage | Free (limited scans) | $52/dev/mo | Custom (enterprise) | ✓ AI security | Premium | T1 |
| GitHub Copilot (Code Review) | Per-seat (bundled) | $19/mo (Copilot Pro) | $39/user/mo (Business) | $39/user/mo | ✓ PR review (new) | Bundled premium | T2 |
| Cursor (code review add-on) | Usage-based | $20/mo (500 requests) | $40/mo (unlimited slow) | $40/mo + $0.04/fast | ✓ Full AI | Premium | T1 |
| CodeLens AI (Current) | Flat-rate | $49/mo | $49/mo | $49/mo | ✓ Full AI | Parity | T1 |
| CodeLens AI (Proposed) | Hybrid | $29/mo | $69/mo | $149/mo | ✓ Full AI | Premium |
Price ($/user/mo equivalent)
High ───────────────────────────────────────────────────
$150 │ ● CodeLens Enterprise
│ (proposed)
$100 │
│ ● Snyk Code
$75 │ ● CodeLens Pro ● GitHub Copilot Biz
│ (proposed) (bundled value)
$50 │ ● CodeLens
│ (current) ● Cursor ($40)
$40 │
│ ● SonarQube ● CodeLens Starter
$30 │ ($30) (proposed)
│
$15 │ ● Codacy ● DeepSource
│
Free │ ● SonarQube Free ● Codacy Free
─┼─────────────────────────────────────────────────
Low Static Analysis AI-Assisted AI-Native
Only (basic) (full)
Perceived Value →Reference Price Effect: Engineering managers evaluating CodeLens compare against two anchors: (1) GitHub Copilot at $19-$39/mo — but this is a coding assistant, not a review tool, so the comparison is imprecise; (2) the cost of a senior engineer's time on code review — typically 20-30% of their work week, or $2,400-$3,600/month. CodeLens at $69/mo (proposed Pro) is 97-98% cheaper than the human alternative. The reference price strongly supports premium positioning. T3
Good/Better/Best is the tier design framework where the middle tier is the conversion target, the bottom tier drives acquisition, and the top tier captures expansion revenue. Package architecture is conversion engineering, not feature listing.
| Dimension | Starter ($29/mo) | Professional ($69/mo) | Enterprise ($149/mo) |
|---|---|---|---|
| Target persona | Solo devs, freelancers, small open-source maintainers | Small-to-mid engineering teams (2-20 devs) — TARGET TIER | Mid-market and enterprise engineering orgs (20+ devs) |
| AI reviews included | 50/month T4 | 250/month T4 | 1,000/month T4 |
| Overage rate | $0.15/review | $0.12/review | $0.08/review |
| Repositories | 3 repos | Unlimited | Unlimited |
| AI model | Standard (GPT-4o-mini equiv.) | Advanced (GPT-4o equiv.) | Advanced + custom fine-tuned |
| Languages | 5 languages | All 20+ languages | All + custom language support |
| Review depth | Single-file analysis | Cross-file + dependency analysis | Full codebase context + architecture review |
| Security scanning | ✗ | ✓ OWASP Top 10 | ✓ Full SAST + compliance (SOC2, HIPAA) |
| Team features | ✗ | ✓ Team dashboard, shared rules | ✓ + SSO, SAML, audit logs, RBAC |
| Integrations | GitHub, GitLab | + Jira, Slack, VS Code, JetBrains | + custom webhooks, API access, on-prem option |
| Support | Community + email | Priority email (24h SLA) | Dedicated CSM + Slack channel (4h SLA) |
| Upgrade trigger | Hits 50-review limit; needs more repos | Needs SSO/compliance; hits 250-review limit; 20+ devs | N/A (expansion via usage growth) |
50 AI reviews on Starter: Enough for a solo developer doing 2-3 PRs/day with AI review on the most important ones. The limit is visible — a developer hitting it weekly understands what more reviews would do. 50 reviews at $0.03 cost = $1.50/month in AI cost, well within the $29 price. Margin: 89%. T1
250 AI reviews on Professional: Covers 80% of small-team usage (median team does 120-180 reviews/month based on internal data). Teams that exceed 250 are high-value power users who should either pay overage or upgrade to Enterprise. The $69 price at 250 reviews = $0.276/review effective price — 4.6x markup on AI cost, healthy margin. T1
1,000 AI reviews on Enterprise: Covers all but the top 2% of accounts (only ~164 accounts currently exceed 1,000 reviews/month). At $149 with $0.08 overage, even a 2,000-review account pays $149 + (1,000 x $0.08) = $229/month — still margin-positive since AI cost is $60 and infrastructure is $8.50, totaling $68.50. Margin: 70%. T1
Security scanning gated to Pro/Enterprise: Security is the highest-value feature (prevents production incidents worth $2K-$10K each). Gating it creates the single strongest upgrade trigger from Starter to Pro. 68% of interviewed customers cited security as "must-have" T3. This is not "feature hostage" — Starter still delivers the core job (AI code review); security is an expansion job.
| Metric | Starter ($29) | Professional ($69) | Enterprise ($149) |
|---|---|---|---|
| Included AI cost (max) | 50 x $0.03 = $1.50 | 250 x $0.03 = $7.50 | 1,000 x $0.03 = $30.00 |
| Infrastructure cost | $8.50 | $8.50 | $12.00 (dedicated resources) |
| Support cost | $0.50 | $2.00 | $15.00 (dedicated CSM allocation) |
| Total COGS | $10.50 | $18.00 | $57.00 |
| Gross margin | 64% | 74% | 62% |
| Margin at max usage | 64% (capped at 50) | 74% (at 250) | 62% (at 1,000) |
| Margin at 2x included | N/A (overage kicks in) | $69 + (250 x $0.12) = $99 / COGS $25.50 = 74% | $149 + (1,000 x $0.08) = $229 / COGS $87 = 62% |
All three tiers maintain 62%+ gross margin even at maximum included usage. Overage pricing is set at 4x AI cost (Starter), 4x (Pro), and 2.7x (Enterprise) — each profitable on a marginal basis. The current flat-rate model has a blended margin of 52% but hides margin-negative accounts in the top decile. The new model eliminates margin-negative accounts entirely.
A pricing recommendation without "what happens if we are wrong" is a guess, not a strategy. The sensitivity analysis stress-tests the recommended pricing against volume, conversion, and churn variations.
| Scenario | Pro Price | Est. Pro Customers | Pro Revenue/Mo | Total ARR (all tiers) | vs. Current $4.8M |
|---|---|---|---|---|---|
| -30% (aggressive penetration) | $48/mo | 3,600 | $172,800 | $5.4M | +12% |
| -20% | $55/mo | 3,400 | $187,000 | $5.9M | +23% |
| -10% | $62/mo | 3,300 | $204,600 | $6.6M | +38% |
| Base (recommended) | $69/mo | 3,100 | $213,900 | $7.4M | +54% |
| +10% | $76/mo | 2,800 | $212,800 | $7.2M | +50% |
| +20% | $83/mo | 2,500 | $207,500 | $6.9M | +44% |
| +30% | $90/mo | 2,200 | $198,000 | $6.5M | +35% |
Revenue is robust across the range. Even at +30% above recommended ($90/mo Pro), total ARR is still $6.5M — 35% above current. Even at -30% ($48/mo Pro), total ARR is $5.4M — 12% above current. The pricing recommendation has a wide "safe zone": any Pro price between $48-$90 produces meaningfully higher ARR than the current flat $49. M
| Variable | Base Assumption | If Wrong by 20% | Impact on Recommendation |
|---|---|---|---|
| Churn rate on migration | 8% of existing base churns during transition | If 10% churn: ARR drops to $7.0M (still +46%) | Recommendation holds — extend grandfathering by 3 months |
| WTP estimate | OPP at $58, IDP at $72 | If OPP is actually $48: reduce Pro to $59, ARR ~$6.4M | Pro price may need reduction — run A/B test in first 60 days |
| Usage-tier mix | 40% Starter, 38% Pro, 22% Enterprise | If 55% Starter, 30% Pro, 15% Enterprise: ARR drops to $5.8M | ARR still above current; adjust feature gates to drive more users to Pro |
| AI cost per review | $0.03/review | If $0.036/review: margin drops 2-3 points per tier | Recommendation holds — margin still above 60% on all tiers |
| Competitive response | No major competitive price cut in 6 months | If GitHub Copilot adds full AI review at $39/user/mo | Pro tier must drop to $49-$59 to compete; emphasize depth/security as differentiation |
$8.2M ARR. 5% churn on migration. 35% Pro, 25% Enterprise. 15% overage revenue. New customer acquisition up 20% due to lower Starter entry point. L
$7.4M ARR. 8% churn on migration. 38% Pro, 22% Enterprise. 10% overage revenue. New customer acquisition flat. M
$5.6M ARR. 15% churn on migration. Heavy downshift to Starter (55%). GitHub Copilot competes directly. Still 17% above current. L
Revenue Impact Modeling quantifies the financial effect of pricing changes on existing customers — including churn risk, grandfathering cost, expansion uplift, and net 12-month impact.
| Metric | Current (Flat $49) | Projected (Hybrid) | Delta |
|---|---|---|---|
| Total customers | 8,200 | 7,544 (8% migration churn) | -656 accounts |
| ARPU | $49/mo | $81.72/mo (blended) | +$32.72 (+67%) |
| Monthly revenue | $401,800 | $616,414 | +$214,614 |
| ARR | $4,821,600 | $7,396,968 | +$2,575,368 |
| Gross margin (blended) | 52% | 68% | +16 pts |
| Margin-negative accounts | ~200 (2.4%) | 0 | Eliminated |
| Tier | Projected Customers | % of Base | Revenue/Mo | Overage Revenue/Mo | Total/Mo |
|---|---|---|---|---|---|
| Starter ($29) | 3,018 | 40% | $87,522 | $2,114 | $89,636 |
| Professional ($69) | 2,867 | 38% | $197,823 | $18,635 | $216,458 |
| Enterprise ($149) | 1,659 | 22% | $247,191 | $63,129 | $310,320 |
| Total | 7,544 | 100% | $532,536 | $83,878 | $616,414 |
| Impact Dimension | Estimate | Confidence | Evidence |
|---|---|---|---|
| Existing customer churn risk | 656 accounts (8% of base) lost during migration T4 | M | Industry benchmark: 5-12% churn on pricing changes T4; mitigated by grandfathering |
| Grandfathering cost | $0 for 6 months (existing customers keep $49), then $196,800/year in loyalty discounts (20% off for 12 months on ~1,640 accounts) T5 | M | Revenue deferred, not lost — loyalty discount expires after 12 months |
| Expansion revenue uplift (tier upgrades) | +$1.8M/year from power users moving to Pro/Enterprise T4 | M | 820 power users currently at $49, projected 70% move to Pro ($69) or Enterprise ($149) |
| Overage revenue | +$1.0M/year from usage above included allotments T5 | L EVIDENCE-LIMITED | Based on current usage patterns projected onto new tiers; actual overage behavior uncertain |
| New customer acquisition uplift | +10-15% from lower $29 Starter price reducing entry barrier T5 | L EVIDENCE-LIMITED | Directional: $29 vs. $49 = 41% lower entry; comparable products saw 15-25% acquisition lift on tier unbundling T4 |
| Net revenue impact (12-month) | +$2.58M (+54% ARR uplift) | M | Triangulated across tier modeling, churn estimates, expansion projections |
| Phase | Timeline | Action | Owner | Risk |
|---|---|---|---|---|
| Announce | Day 0 | Email all 8,200 customers. New pricing for new customers immediately. Existing customers grandfathered at $49 for 6 months. | Product + Marketing | Negative PR on social media (dev community sensitive to pricing changes) |
| Grandfather period | Months 1-6 | Existing customers stay at $49/mo flat. Usage tracking visible in dashboard ("you used X reviews this month — see which tier fits you"). | Product + Eng | Customers entrench at $49; resistance to any change |
| Migration communication | Month 5 | Send personalized emails: "Based on your usage (X reviews/mo), you'd be on [Tier] at [Price]. Here's your 20% loyalty discount." | Product + CS | Sticker shock for power users seeing $149+ for first time |
| Migration execution | Month 7 | Move all existing customers to tier matching their median 6-month usage. Apply 20% loyalty discount for 12 months. | Eng + Billing | Billing system errors; incorrect tier assignment |
| Loyalty discount expiry | Month 19 | Loyalty discount expires. Full pricing in effect for all customers. | CS + Billing | Second churn wave when discount expires — monitor closely |
| Monitor | Ongoing | Track: churn by tier, overage adoption, tier upgrade/downgrade rates, NPS by tier, support tickets mentioning pricing | Product + Data |
AI products have non-trivial marginal costs that traditional SaaS pricing ignores. This section applies AI-specific pricing patterns to CodeLens's cost structure.
| Pattern | Assessment | Recommendation |
|---|---|---|
| Value metric alignment | Price should scale with AI reviews (the unit of value delivered). Each review catches 0.3-1.2 issues on average T1. Value is highly variable — catching a critical security flaw is worth $10,000; catching a style nit is worth $0. | Use review count as the metering unit, not tokens or compute time. Customers understand "reviews" — they do not understand tokens. |
| Marginal cost structure | $0.03/review (GPU inference + LLM API call). This is 6.1% of the current $49 flat price at median usage (100 reviews/mo = $3.00 cost) but 31%+ at power-user usage (500+ reviews = $15+). T1 | Usage-based component mandatory. At $0.03 marginal cost, pure flat-rate is sustainable only below ~400 reviews/month. Above that, each incremental review erodes margin. |
| Usage metering approach | Per-review is clean and customer-friendly. Alternatives: per-LOC (penalizes large PRs), per-token (opaque), per-finding (unpredictable, penalizes clean code). T4 | Per-review metering. One PR = one review, regardless of size. Simple to understand, simple to meter, correlates with value delivery. |
| GPU cost trajectory | AI inference costs declining ~30% annually as hardware improves and models optimize T4. Current $0.03/review may be $0.02 in 12 months and $0.014 in 24 months. | Do NOT pass cost savings through automatically. Capture declining costs as margin improvement. If reviews currently cost $0.03 and pricing is $0.12 overage (4x), a drop to $0.02 cost with $0.12 pricing gives 6x margin — reinvest in product quality. |
| Bundle vs. unbundle AI | AI review IS the product — cannot unbundle. But the quality tier (standard vs. advanced model) can be gated by tier. T4 | Bundle AI into all tiers. Gate model quality (standard/advanced/custom) and review depth (single-file/cross-file/architecture) by tier. This gives each tier a differentiated AI experience. |
| Monthly Reviews | AI Cost | Infrastructure | Total COGS | Tier | Revenue (incl. overage) | Gross Margin |
|---|---|---|---|---|---|---|
| 10 | $0.30 | $8.50 | $9.30 | Starter | $29.00 | 68% |
| 50 | $1.50 | $8.50 | $10.50 | Starter (at cap) | $29.00 | 64% |
| 100 | $3.00 | $8.50 | $12.00 | Pro | $69.00 | 83% |
| 250 | $7.50 | $8.50 | $18.00 | Pro (at cap) | $69.00 | 74% |
| 500 | $15.00 | $8.50 | $25.50 | Pro + overage | $69 + (250 x $0.12) = $99.00 | 74% |
| 1,000 | $30.00 | $12.00 | $57.00 | Enterprise (at cap) | $149.00 | 62% |
| 2,000 | $60.00 | $12.00 | $87.00 | Enterprise + overage | $149 + (1,000 x $0.08) = $229.00 | 62% |
| 2,000 | $60.00 | $8.50 | $68.50 | Current flat-rate | $49.00 | -40% |
The critical comparison: A 2,000-review/month account on the current flat $49 plan has a -40% gross margin — CodeLens pays $19.50 per month to serve this customer. On the new Enterprise tier + overage, the same account generates $229/month at 62% margin. That is a $180/month swing per account across ~200 accounts = $432,000/year in recovered margin from the heaviest users alone.
| Scenario | Cost/Review | Impact on Starter Margin | Impact on Enterprise Margin | Action Required |
|---|---|---|---|---|
| GPU costs drop 30% (12 mo) | $0.021 | 68% → 72% | 62% → 69% | None — capture as margin improvement |
| Model upgrade increases cost 50% | $0.045 | 64% → 58% | 62% → 53% | Gate upgraded model to Pro+ tiers; keep Starter on standard model |
| Token prices spike 2x (supply shock) | $0.060 | 64% → 51% | 62% → 41% | Raise overage rates by $0.03; add usage cap to Starter |
| Tier | Count | Examples |
|---|---|---|
| T1 (Transaction/usage data) | 14 | Internal usage distribution, AI cost per review, infrastructure costs, current revenue, competitor pricing pages, review volume by account |
| T2 (Primary research n>=100) | 4 | GitHub Copilot pricing announcements, Cursor pricing changes, competitor feature comparison matrices, industry AI pricing reports |
| T3 (Interviews n>=10) | 18 | 47 customer interviews (Van Westendorp, feature importance, churn risk, competitive alternatives, value quantification) |
| T4 (Competitive inference) | 12 | Tier mix projections, churn rate estimates, competitive positioning assessment, model fit analysis, feature allocation rationale |
| T5 (Expert estimation) | 6 | Overage revenue projections, new customer acquisition uplift, grandfathering cost, AI cost trajectory, loyalty discount impact |
| T6 (Model inference) | 0 | Not used |
Total evidence points: 54 T1-T4 : 48; T5: 6; T6: 0
Triangulation: All pricing recommendations cite minimum 2 evidence tiers. The core pricing model decision (hybrid) is supported by T1 (usage data + cost structure), T2 (competitive norms), and T3 (customer interviews). The weakest area is revenue projections (T4-T5) — these are directional, not precise. Run A/B pricing tests in months 1-3 for T1 validation.
| # | Assumption | Framework | Confidence | Evidence | What Invalidates This |
|---|---|---|---|---|---|
| 1 | Power users (500+ reviews/mo) will accept $149+ pricing because the value justifies it | WTP, Package Architecture | M | T3 11 of 15 power users said they'd pay $100-$200/mo; T4 reference price = $12K/mo senior eng time | If >25% of power users churn within 3 months of migration, WTP overestimated for this segment |
| 2 | Casual users will not churn en masse when they see usage limits for the first time | Package Architecture, Revenue Impact | M | T3 8 of 12 casual users said $29 was "more reasonable" than $49; T4 50 reviews/mo covers median casual usage | If casual user churn exceeds 15% in months 7-9, usage limits are triggering loss aversion even though effective price dropped |
| 3 | AI inference costs will not increase materially in the next 12 months | Model Selection, AI Patterns | H | T4 GPU costs declining ~30% YoY; LLM inference pricing trending down across all providers | If a major AI supply shock occurs (GPU shortage, model licensing changes) and costs double, Enterprise tier margin drops to 41%, requiring overage rate increase |
| 4 | GitHub Copilot will not ship a full-featured AI code review product at $39/user/mo within 6 months | Competitive Map, Sensitivity | L | T2 Copilot code review is in beta, limited to PR suggestions; no pricing announced for standalone review | If Copilot ships full AI review bundled in Business ($39/user), CodeLens Pro must drop to $49-59 and differentiate on depth, security, customization |
| 5 | The 6-month grandfather period is sufficient to absorb migration shock | Revenue Impact, Migration | M | T4 Industry benchmarks show 3-6 months is standard; T3 customers said "at least 3 months" in interviews | If NPS drops >15 points during grandfather period, extend to 9 months and increase communication cadence |
What this strategy assumes: Usage-based overage pricing captures value from power users without changing their behavior.
What could be wrong: Metered pricing creates "meter anxiety" — developers consciously or unconsciously reduce AI review usage to avoid overage charges. If CodeLens's growth thesis depends on teams reviewing MORE code over time (expanding the habit), overage pricing creates a counter-incentive. The very mechanism designed to capture value may suppress the behavior that creates value. In 14 of 47 interviews, customers cited "predictable billing" as a top-3 criterion. T3
Mitigation: Set included allotments high enough that 70-80% of users on each tier never hit overages. Only 20-30% of users should see overage charges — enough to capture value from power users, not enough to create widespread anxiety. Monitor: if average reviews/account drops >10% post-migration, overage pricing is suppressing usage.
What this strategy assumes: 40% Starter / 38% Pro / 22% Enterprise tier distribution.
What could be wrong: CodeLens has never had tiers. The projected distribution is based on current usage patterns mapped onto new tier thresholds. But behavior changes when pricing changes — some users will downshift usage to stay on cheaper tiers (usage compression), some will overshoot into higher tiers than expected, and some will simply leave. The 40/38/22 split is an educated guess, not a measured outcome. If the actual split is 55/30/15 (heavy Starter skew), projected ARR drops to $5.8M instead of $7.4M. T5
Mitigation: Run the first 90 days as a pricing A/B test: 50% of new signups see old $49 flat pricing, 50% see new tiered pricing. Measure conversion rate, tier selection, and 30-day retention by cohort before committing to full migration. This gives T1 evidence on tier mix before risking the existing base.
What this strategy assumes: Starter at $29 primarily attracts NEW customers who find $49 too expensive, and casual existing users who are currently overpaying.
What could be wrong: Some regular users currently paying $49 (doing 50-200 reviews/month) may discover they can survive on 50 reviews/month by being selective about which PRs get AI review. They downgrade to Starter and reduce usage — losing $20/month per account. If 20% of regular users (656 accounts) do this, that is $157,440/year in revenue loss not accounted for in the model. The Starter tier's 3-repo limit is the primary gate, but teams with a single critical repo can operate entirely within Starter. T5
Mitigation: Make the Starter-to-Pro upgrade trigger irresistible: gate security scanning, cross-file analysis, and team dashboard to Pro. Users who need any team features or security coverage must upgrade. Monitor the downgrade rate from current $49 to new $29 — if >25% of the base downgrades, the feature gate between Starter and Pro is too porous.
| Contradiction | Framework A Says | Framework B Says | Resolution |
|---|---|---|---|
| WTP vs. Competitive Map on Pro price | WTP (Van Westendorp): OPP is $58, IDP is $72 — price at $58-72 | Competitive Map: Snyk Code at $52/dev, Copilot at $39/user — price at $39-52 for parity | Weight WTP. CodeLens is differentiated (full AI review vs. partial); the reference price from customer interviews ($58-72) already incorporates competitive awareness. Price at $69 (within WTP sweet spot) and compete on value, not parity. |
| Package Architecture vs. AI Cost Patterns on Starter allotment | Package Architecture: Starter should deliver enough value to create habit — higher allotment (100+ reviews) | AI Cost Patterns: Marginal cost at 100 reviews = $3.00, but Starter at $29 needs 64%+ margin — cap at 50-70 reviews | Weight AI Cost Patterns. Set Starter at 50 reviews (still enough for solo dev habit formation at ~2-3 reviews/day on work days) while maintaining 64% margin. If conversion data shows Starter users don't hit the limit enough to trigger upgrade, raise to 75. |
| Sensitivity Analysis vs. Migration Strategy on timing | Sensitivity: Move fast — every month at $49 flat is $214K/month in foregone revenue | Migration: Move slow — 6-month grandfather reduces churn risk from 12%+ to 8% | Weight Migration. The $1.3M in foregone revenue during the 6-month grandfather is a customer retention investment. Losing 4% more of the base (328 accounts x $49/mo x 12 = $193K/year in permanent revenue loss) costs more over the customer lifetime. |
| Trigger | What to Re-assess | Timeline |
|---|---|---|
| GitHub Copilot ships full AI code review at <$50/user | Pro tier price, competitive positioning, feature differentiation | Within 30 days of Copilot announcement |
| Migration churn exceeds 12% | Grandfather period length, loyalty discount amount, migration communication | Monthly check during months 1-9 |
| AI inference costs increase >30% | Overage rates, included allotments, model quality gating | Quarterly cost review |
| >40% of Pro users consistently hit overage every month | Pro allotment (may need increase from 250 to 400) | Monthly after migration |
| Starter-to-Pro conversion <5% in 90 days | Feature gate between tiers, Starter allotment (may be too generous) | 90-day checkpoint |
Analysis Date: March 2026 Evidence Points: 54 T1-T4 : 48; T5: 6; T6: 0 Frameworks Applied: Pricing Model Selection, Van Westendorp WTP, Competitive Pricing Map, Good/Better/Best Package Architecture, AI/SaaS Pricing Patterns, Sensitivity Analysis, Revenue Impact Modeling License: MIT PM Skills Arsenal: pricing-packaging
Apply this framework to your own pricing strategy in 4 steps:
Output: A board-ready pricing strategy with evidence-graded confidence in ~2.5 hours.
From this analysis to next steps: - See Competitive Analysis use case for positioning strategy that informs pricing decisions - See Metric Design use case for measuring the impact of pricing changes post-launch - See Discovery Research use case for running WTP studies to feed into pricing strategy
Real-world skill chains: - Run Discovery Research first to gather T2-T3 WTP data, then feed into this skill's WTP Assessment - After pricing launch, use Metric Design to build a measurement framework for conversion, churn, and ARPU by tier - Use Narrative Building to communicate the pricing change to customers, investors, and internal stakeholders