---
name: "Biz pricing strategy"
repo: https://github.com/asgard-ai-platform/skills
source: https://github.com/asgard-ai-platform/skills/blob/HEAD/biz-pricing-strategy/SKILL.md
license: "MIT"
install: "npx skills add asgard-ai-platform/skills --skill biz-pricing-strategy"
page: https://100xlab.com/resources/agent-skills/asgard-ai-platform-skills--biz-pricing-strategy--gsvp19
---

# Pricing Strategy

## Overview

Pricing is the only marketing mix element that generates revenue — all others are costs. This skill covers five pricing approaches (cost-plus, value-based, competitive, penetration, skimming) plus psychological pricing techniques. The right approach depends on the product lifecycle stage, competitive landscape, and customer price sensitivity.

## When to Use

**Trigger conditions:**
- User setting prices for a new product
- User evaluating whether current pricing is optimal
- User asks "how much should we charge?" or "why are our margins low?"
- User needs to choose between pricing models (subscription vs one-time, freemium vs premium)

**When NOT to use:**
- For comprehensive financial analysis → use financial ratios or DCF
- For customer segmentation → use STP
- For cost structure analysis → use Value Chain

## Framework

```
IRON LAW: Price Communicates Positioning

Price is not just economics — it's a signal. Lowering price to compete
can permanently reposition a brand as "cheap." Raising price without
value justification creates distrust.

Every price change must be evaluated through BOTH a financial lens
(margins, volume) AND a positioning lens (what does this price say about us?).
```

### Step 1: Understand the Three Price Anchors

Every pricing decision sits between three constraints:

| Anchor | What It Sets | Method |
|--------|-------------|--------|
| **Cost floor** | Minimum viable price | Cost analysis — below this, you lose money |
| **Competitor reference** | Market expectations | Competitive benchmarking — what alternatives cost |
| **Customer ceiling** | Maximum willingness to pay | Value research — what the customer thinks it's worth |

### Step 2: Choose a Pricing Approach

| Approach | How It Works | Best When |
|----------|-------------|-----------|
| **Cost-Plus** | Cost + fixed margin % | Commodity products, government contracts, stable costs |
| **Value-Based** | Price based on customer's perceived value | Differentiated products, strong brand, measurable customer benefit |
| **Competitive** | Match or undercut competitor prices | Undifferentiated market, price-sensitive customers |
| **Penetration** | Start low to gain market share, raise later | New market entry, network effects, high switching costs |
| **Skimming** | Start high, lower over time | Innovation leader, early adopters willing to pay premium |

### Step 3: Apply Psychological Pricing Techniques

| Technique | How It Works | Example |
|-----------|-------------|---------|
| **Charm pricing** | End in 9 or 99 | NT$299 instead of NT$300 |
| **Anchoring** | Show a higher price first, then the actual price | "Was NT$1,200, now NT$799" |
| **Decoy effect** | Offer three options where the middle is the intended choice | Small NT$99, Medium NT$149, Large NT$159 (Large looks like a deal) |
| **Bundle pricing** | Combine products at a discount vs individual purchase | "All 3 for NT$999" (vs NT$450 each) |
| **Freemium** | Free basic tier, charge for premium features | Spotify, Notion, Canva |

### Step 4: Validate with Price Sensitivity Analysis

Before committing:
- **Van Westendorp**: Survey-based method — ask customers "at what price is this too expensive / too cheap / a bargain / getting expensive?"
- **Gabor-Granger**: Show a price, ask if they'd buy. Vary the price across respondents.
- **A/B test**: If possible, test two price points with real transactions

### Step 5: Monitor and Adjust

After launch:
- Track **price elasticity**: % change in demand / % change in price
- Monitor **competitive response**: Did competitors match your price?
- Watch **customer perception**: Did the price signal what you intended?

## Output Format

```markdown
# Pricing Strategy: {Product/Service}

## Three Anchors
- Cost floor: {$X} (based on: {cost breakdown})
- Competitor reference: {$X range} (competitors: {list})
- Customer ceiling: {$X} (based on: {value metric})

## Recommended Approach
**{Approach name}** — {rationale}

## Price Point
- Recommended price: {$X}
- Expected margin: {X%}
- Positioning signal: {what this price says about the brand}

## Psychological Techniques Applied
- {technique}: {how applied}

## Sensitivity Analysis
| Price Point | Est. Volume | Revenue | Margin | Risk |
|------------|------------|---------|--------|------|
| {low} | {high vol} | {$X} | {X%} | {positioning risk} |
| {recommended} | {med vol} | {$X} | {X%} | {balanced} |
| {high} | {low vol} | {$X} | {X%} | {volume risk} |

## Monitoring Plan
- Review frequency: {monthly/quarterly}
- Key metrics: {elasticity, competitive response, perception}
```

## Examples

### Correct Application

**Scenario:** Pricing a new SaaS project management tool for SMBs in Taiwan

**Three anchors:**
- Cost floor: NT$150/user/month (server + support costs)
- Competitors: Asana NT$350/user, Monday.com NT$300/user, Trello Free-NT$170/user
- Customer ceiling: NT$400/user (based on 30 customer interviews — value of time saved)

**Approach**: Value-based with decoy pricing
- Basic: NT$199/user/month (limited features — the decoy)
- Pro: NT$299/user/month (full features — the target)
- Enterprise: NT$499/user/month (with SSO, audit logs — anchor)

**Why**: Pro at NT$299 looks like great value vs Enterprise at NT$499, and much better than Basic at NT$199 for only NT$100 more.

### Incorrect Application

**What went wrong:**
- Set price at cost + 20% (NT$180/user) without checking competitor reference or customer ceiling → Left NT$120+/user of value on the table. Customer would have paid NT$299.
- Cut price from NT$299 to NT$149 to match a new budget competitor → Signaled "we're a budget tool now," causing premium customers to leave. Violates Iron Law: price communicates positioning.

## Gotchas

- **Cost-plus is a fallback, not a strategy**: Cost-plus only makes sense when you can't measure value or differentiate. In most cases, value-based pricing captures more margin.
- **Penetration pricing requires a plan to raise prices**: If you start low, you need a clear path to profitability. "We'll raise prices later" without a mechanism (switching costs, network effects) is wishful thinking.
- **Discounts are addictive**: Frequent discounts train customers to wait for sales. Use selectively and time-limit them.
- **B2B vs B2C psychology differs**: B2B buyers evaluate ROI rationally (though with organizational politics). B2C buyers are more susceptible to psychological pricing. Calibrate techniques to the buyer.
- **Free is not a price — it's a category change**: Moving from paid to free (or vice versa) changes the product category in the customer's mind. The shift from "paid product" to "free with ads" is a complete repositioning.

## References

- For Van Westendorp and Gabor-Granger methodology details, see `references/price-sensitivity.md`
- For SaaS-specific pricing models, see `references/saas-pricing.md`

---

## Supporting file: examples/sample_scenario.md

# Example: 台灣健身 App 從免費轉付費定價策略

## Scenario

**公司：** FitFlow（台灣新創，2023年成立）
**情境：** FitFlow 是一款針對台灣用戶的居家健身 App，目前擁有 12 萬名免費用戶。
創辦人 Amy Chen 準備在 2026 年 Q2 推出付費版，來自投資人的壓力要求 2026 年底
前達到 MRR NT$500 萬。

**用戶對話：**
> "我們有 12 萬個免費用戶，想開始收費。競品 MindBody 一個月要 NT$590，但我覺得
> 我們的用戶不會付那麼多。我打算定 NT$99/月，應該夠低了吧？"

---

## Analysis

### Step 1：三個定價錨點

**成本底線（Cost Floor）**

| 項目 | NT$/月/用戶 |
|------|------------|
| AWS 主機 + CDN | 8 |
| 影片儲存（教練課程） | 12 |
| 客服 + 人工成本攤提 | 15 |
| **總計** | **35** |

→ 成本底線：**NT$35/用戶/月**（付費版）

**競爭對手參考（Competitor Reference）**

| 競品 | 月費 | 特色 |
|------|------|------|
| MindBody（國際） | NT$590 | 預約健身房課程 |
| Keep（中國，台灣版） | NT$168 | 訓練計畫為主 |
| Nike Training Club | 免費 | 無個人化功能 |
| 本土健身房 App X | NT$299 | 含直播課程 |

→ 市場區間：NT$168–599，主流付費點落在 **NT$200–350**

**顧客天花板（Customer Ceiling）**

Amy 手上有一份 800 人問卷（Van Westendorp 四問法）：

| 問題 | 中位數結果 |
|------|-----------|
| 「太貴，我不考慮」 | NT$450 |
| 「開始覺得貴，但可接受」 | NT$280 |
| 「划算，覺得合理」 | NT$199 |
| 「太便宜，品質堪憂」 | NT$59 |

→ **可接受區間：NT$59–450**，最佳定價帶 **NT$199–280**

---

### Step 2：選擇定價方式

Amy 原本想定 NT$99，但三個錨點顯示問題：

- NT$99 已低於顧客「太便宜會起疑」的底線 NT$59 邊緣
- 競品 Keep 定 NT$168，NT$99 會被市場解讀為「次等品」
- 更關鍵：**Iron Law 違反** — NT$99 會將 FitFlow 永久定位為折扣 App，未來漲價困難

**選定策略：Freemium + 價值導向定價（Value-Based）**

FitFlow 的核心差異化：AI 個人化課表 + 本土教練影片（非國際化通用內容）。
這是可量化的感知價值，應用價值定向定價，而非跟隨競品或成本加成。

---

### Step 3：方案設計（誘餌效應）

設計三層方案，讓 **Pro 方案** 成為明顯最佳選擇：

| 方案 | 月費 | 功能 | 角色 |
|------|------|------|------|
| **Free** | 免費 | 基礎訓練庫（50 支影片）、無個人化 | 導流入口 |
| **Pro** | **NT$199/月** | AI 課表、全庫 800+ 影片、進度追蹤 | **目標方案** |
| **Elite** | NT$399/月 | Pro + 每月一次線上教練諮詢（30 min） | 錨點 + 高端 |

**心理定價技巧：**
- Charm pricing：NT$199、NT$399（非 200/400）
- 誘餌效應：Elite NT$399 讓 Pro NT$199 顯得非常划算
- 年繳折扣：Pro 年繳 NT$1,788（相當於 NT$149/月，省 25%）→ 提升 LTV

---

### Step 4：敏感度分析

假設付費轉換率情境（現有 12 萬免費用戶）：

| 情境 | 轉換率 | 付費用戶 | ARPU | MRR | 達標 NT$500 萬？ |
|------|--------|---------|------|-----|-----------------|
| 保守（Pro only） | 3% | 3,600 | NT$199 | NT$717,000 | ✗ |
| 基準（Pro 80% + Elite 20%） | 5% | 6,000 | NT$239 | NT$1,434,000 | ✗（需靠成長） |
| 積極（同上 + 年繳推廣） | 8% | 9,600 | NT$255* | NT$2,448,000 | 需 Q3 加速 |

> *年繳用戶佔 40% 時，月均 ARPU 略升（現金流）

→ 結論：NT$199 定價正確，但 **NT$500 萬 MRR 目標需要靠新用戶獲取（MAU 成長），
不是靠降價**。Amy 原本的 NT$99 方案即使全部轉換，MRR 上限也只有 NT$1,188,000。

---

## Result

```markdown
# Pricing Strategy: FitFlow 付費版（2026 Q2 上市）

## Three Anchors
- Cost floor: NT$35/用戶/月（主機 + 影片儲存 + 客服）
- Competitor reference: NT$168–599（主流帶 NT$199–350）
- Customer ceiling: NT$280（Van Westendorp 最佳定價帶上限）

## Recommended Approach
**Value-Based + Decoy Effect（三層 Freemium）**
FitFlow 的 AI 個人化是可感知的差異化價值，不應用成本加成或跟競品削價。

## Price Points
- Free：NT$0（保留免費用戶基礎，維持網路效應）
- Pro：NT$199/月（目標方案）；年繳 NT$1,788
- Elite：NT$399/月（錨點 + 高端教練服務）

預期 Pro 毛利：(199 - 35) / 199 = **82%**
定位訊號：「專業但親民的本土健身夥伴」，非折扣品牌

## Psychological Techniques Applied
- Charm pricing：199、399（感知低於整數）
- Decoy effect：Elite NT$399 讓 Pro 顯得划算
- 年繳折扣：NT$1,788（鎖定用戶、提升 LTV）

## Sensitivity Analysis
| Price Point | Est. 轉換率 | 付費用戶 | MRR | 風險 |
|------------|------------|---------|-----|------|
| NT$99（Amy 原提案） | 15% | 18,000 | NT$1,782,000 | 品牌定位損傷，難以回頭 |
| NT$199（建議） | 5–8% | 6,000–9,600 | NT$1.4–2.4M | 需搭配用戶成長計畫 |
| NT$299 | 3% | 3,600 | NT$1,076,000 | 轉換率太低，天花板受限 |

## Monitoring Plan
- 審查頻率：每月（上線後前三個月每兩週）
- 關鍵指標：Free→Pro 轉換率、Pro→Elite 升級率、年繳比例、Churn rate
- 警示線：月 Churn > 8% → 調查是否為價格問題或功能缺口
```

**對 Amy 的一句話建議：**
NT$99 不是「夠低」——它是一個會損傷品牌的訊號，而且即使 100% 轉換也達不到
投資人目標。NT$199 配合積極的用戶成長才是正確路徑。

---

## Supporting file: references/price-sensitivity.md

# Price Sensitivity Analysis Methods

Three validated research methods for measuring customer willingness-to-pay before committing to a price point.

---

## Van Westendorp Price Sensitivity Meter (PSM)

### What It Measures

PSM identifies an **acceptable price range** and an **optimal price point** by asking customers four questions at four psychological thresholds.

### The Four Questions

Ask each respondent about your specific product:

1. **Too cheap** — "At what price would this product be so cheap that you'd question its quality?"
2. **Cheap/bargain** — "At what price would this be a bargain — great value for the money?"
3. **Expensive/getting pricey** — "At what price would this start to feel expensive, though you might still buy it?"
4. **Too expensive** — "At what price would this be too expensive — you definitely wouldn't buy it?"

### Analysis: The Four Cumulative Curves

Plot cumulative % of respondents at each price on the x-axis:

| Curve | Direction | Source Question |
|-------|-----------|----------------|
| **Too Cheap** | Ascending (L→R) | % who say price is ≤ their "too cheap" threshold |
| **Cheap** | Descending | % who say price is ≥ their "cheap/bargain" threshold |
| **Expensive** | Ascending | % who say price is ≥ their "expensive" threshold |
| **Too Expensive** | Descending | % who say price is ≤ their "too expensive" threshold |

### The Four Intersection Points

| Intersection | Name | Meaning |
|-------------|------|---------|
| Too Cheap ∩ Too Expensive | **Optimal Price Point (OPP)** | Fewest rejections from both ends |
| Cheap ∩ Expensive | **Indifference Price Point (IDP)** | "Normal" market price; 50% say cheap, 50% say expensive |
| Too Cheap ∩ Expensive | **Lower Acceptable Price (LAP)** | Bottom of acceptable range |
| Cheap ∩ Too Expensive | **Upper Acceptable Price (UAP)** | Top of acceptable range |

**Acceptable Price Range**: LAP to UAP  
**Target**: OPP if maximizing adoption; IDP if matching market perception

### Worked Example: SaaS project management tool for Taiwan SMBs

Survey of 80 respondents. After collecting all four thresholds per respondent, count cumulative % at each NT$ price point:

| Price (NT$/user/mo) | Too Cheap (↑) | Cheap (↓) | Expensive (↑) | Too Expensive (↓) |
|--------------------|--------------|----------|--------------|-----------------|
| 99  | 5%  | 98% | 4%  | 2%  |
| 149 | 15% | 90% | 9%  | 5%  |
| 199 | 30% | 78% | 18% | 10% |
| 249 | 44% | 62% | 35% | 18% |
| 299 | 55% | 48% | 52% | 30% |
| 349 | 65% | 35% | 67% | 44% |
| 399 | 75% | 22% | 80% | 58% |
| 449 | 84% | 12% | 90% | 72% |
| 499 | 91% | 6%  | 95% | 85% |

**Reading the intersections:**
- Too Cheap ∩ Too Expensive → ~NT$249 (OPP)
- Cheap ∩ Expensive → ~NT$295 (IDP)
- Too Cheap ∩ Expensive → ~NT$199 (LAP)
- Cheap ∩ Too Expensive → ~NT$365 (UAP)

**Interpretation**: The acceptable range is NT$199–NT$365. A price of NT$299 sits close to the IDP — customers perceive it as "normal market price." The OPP at NT$249 minimizes resistance but may sacrifice margin relative to what customers would pay.

**Decision**: The SKILL.md example chose NT$299 for the Pro tier — within range, above OPP, close to IDP. Consistent.

### Sample Size

| Use Case | Minimum n | Notes |
|----------|-----------|-------|
| Early validation | 30–50 | Direction only; curves will be noisy |
| Confident range | 100–150 | Standard for B2C products |
| Segment comparison | 150+ per segment | If comparing SMB vs enterprise, etc. |

### PSM Limitations

- Measures **stated preference**, not revealed preference (actual purchase behavior)
- Respondents imagine buying; real WTP is often lower
- Does not model volume — no demand curve output
- Less reliable for products respondents have never bought before

---

## Gabor-Granger Method

### What It Measures

Gabor-Granger builds a **demand curve** by measuring purchase intent at multiple price points. Output: estimated revenue-maximizing price.

### Protocol

Show each respondent a **single price** (or a descending sequence), then ask:

> "At NT$X per month, how likely would you be to subscribe?"  
> Scale: Definitely would / Probably would / Probably not / Definitely not

Assign binary purchase intent: "Definitely would" + "Probably would" = **buyer**

**Two design variants:**

| Variant | Method | Tradeoff |
|---------|--------|----------|
| **Between-subjects** | Each respondent sees one price | Cleaner; requires larger n (50+ per price point) |
| **Sequential descending** | Each respondent sees prices from high to low | Smaller n; risk of anchoring bias |

Sequential descending is standard; start 30–50% above your expected ceiling.

### Analysis: Demand Schedule

| Price (NT$/user/mo) | % "Would Buy" | Implied Demand Index |
|--------------------|--------------|---------------------|
| 499 | 18% | 18 |
| 399 | 31% | 31 |
| 299 | 52% | 52 |
| 249 | 64% | 64 |
| 199 | 78% | 78 |
| 149 | 87% | 87 |

### Revenue Index Calculation

```
Revenue Index = Price × % Would Buy
```

| Price | % Would Buy | Revenue Index |
|-------|------------|---------------|
| 499  | 18%  | 89.8  |
| 399  | 31%  | 123.7 |
| 299  | 52%  | **155.5** ← peak |
| 249  | 64%  | 159.4 ← also near peak |
| 199  | 78%  | 155.2 |
| 149  | 87%  | 129.6 |

**Revenue-maximizing price**: NT$249–NT$299 range. NT$299 is defensible given IRON LAW — it maintains premium positioning while staying near peak revenue.

### Price Elasticity from Gabor-Granger

```
Point Elasticity = (ΔQ / Q) / (ΔP / P)
```

Between NT$249 and NT$299:

```
ΔQ = (52% − 64%) / 64% = −18.75%
ΔP = (299 − 249) / 249 = +20.08%

Elasticity = −18.75% / 20.08% = −0.93
```

Elasticity of −0.93 means **inelastic** in this range — a 20% price increase reduces demand only 19%. This supports pricing at NT$299 over NT$249.

Between NT$399 and NT$499:

```
ΔQ = (18% − 31%) / 31% = −41.9%
ΔP = (499 − 399) / 399 = +25.1%

Elasticity = −41.9% / 25.1% = −1.67
```

Elasticity of −1.67 means **elastic** above NT$399 — demand drops faster than price rises. Avoid pricing above NT$399.

### Gabor-Granger Limitations

- **Hypothetical bias**: stated intent overestimates actual purchase 30–50% (rule of thumb: treat "Probably would" as only 50% likely)
- **No quality signal**: unlike PSM, does not capture "too cheap" concerns
- **Sequential anchoring**: in descending sequences, respondents who refuse NT$499 are primed to see NT$299 as reasonable

---

## A/B Price Testing (Revealed Preference)

### When to Use

Use when you have live traffic and can randomize price exposure. This is the only method measuring **actual willingness to pay** (not stated intent).

### Minimum Viable Setup

```
Control:   Price = NT$249, shown to 50% of eligible visitors
Treatment: Price = NT$299, shown to 50% of eligible visitors

Measure: Conversion rate (sign-ups / unique visitors)
Run duration: Until statistical significance (see below)
```

### Required Sample Size

For a two-sample proportion test:

```
n = 2 × (Z_α/2 + Z_β)² × p̄(1−p̄) / (p1−p2)²

Where:
  Z_α/2 = 1.96  (α = 0.05, two-tailed)
  Z_β   = 0.84  (power = 80%)
  p̄     = average conversion rate across both groups
  p1, p2 = expected conversion rates per group
```

**Example**: Baseline conversion 5%, expect NT$299 reduces it to 4%

```
p̄ = (5% + 4%) / 2 = 4.5%
(p1 − p2)² = (0.05 − 0.04)² = 0.0001

n = 2 × (1.96 + 0.84)² × 0.045 × 0.955 / 0.0001
n = 2 × 7.84 × 0.04298 / 0.0001
n ≈ 6,739 per group
```

You need ~6,700 visitors per variant — ~13,400 total — before reading results.

### Revenue per Visitor Comparison

Do not optimize for conversion rate alone. Compare **revenue per visitor**:

```
RPV = Price × Conversion Rate

Control:   NT$249 × 5.0% = NT$12.45 / visitor
Treatment: NT$299 × 4.2% = NT$12.56 / visitor  ← +0.9% RPV despite lower conversion
```

Even though conversion dropped, revenue per visitor is higher at NT$299. **This is the correct metric.**

### A/B Testing Pitfalls

- **Peeking**: checking results before reaching required n inflates false positive rate. Commit to the sample size before starting.
- **Segment leakage**: if the same customer sees both prices (different sessions, shared devices), results are contaminated. Use user-ID randomization, not session-ID.
- **Temporal confounds**: do not run over a promotional period, major holiday, or competitor campaign.
- **Legal / ethical exposure**: in some markets (EU consumer law, Taiwan Consumer Protection Act), showing different prices to different users for the same product requires disclosure. Confirm with legal before running.

---

## Choosing a Method

| Condition | Use |
|-----------|-----|
| Pre-launch, no live traffic | Van Westendorp + Gabor-Granger |
| Want demand curve + elasticity estimates | Gabor-Granger |
| Want acceptable price range and "too cheap" signal | Van Westendorp |
| Live traffic, can randomize | A/B test |
| Need to compare segments (SMB vs enterprise) | Gabor-Granger with split samples |
| High-stakes decision, budget available | All three in sequence |

**Sequencing when budget allows:**

1. **PSM** → find acceptable range and OPP (early concept stage)
2. **Gabor-Granger** → build demand curve within that range (pre-launch)
3. **A/B test** → validate with real transactions (post-launch)

Each method anchors the next: PSM narrows the range, Gabor-Granger identifies the revenue-optimal point, A/B test confirms with revealed behavior.

---

## Quick Reference: Formulas

```
Price Elasticity of Demand:
  E = (% change in quantity) / (% change in price)
  E < −1  → elastic   (price-sensitive; revenue falls when price rises)
  E > −1  → inelastic (price-insensitive; revenue rises when price rises)
  E = −1  → unit elastic (revenue unchanged)

Revenue Index (Gabor-Granger):
  RI = Price × Purchase Intent %

Revenue per Visitor (A/B):
  RPV = Price × Conversion Rate

Sample Size (A/B, two-proportion):
  n per group = 2(Z_α/2 + Z_β)² × p̄(1−p̄) / (p1−p2)²
  Shortcut: use n≈16,000 / (p1−p2)² for α=0.05, power=80%, p̄≈0.05
```

---

## Supporting file: references/saas-pricing.md

# SaaS Pricing Models

SaaS pricing differs from one-time product pricing in one fundamental way: the customer pays repeatedly, so **the price must justify renewal at every billing cycle**, not just at the moment of purchase. This shifts the pricing question from "what is this worth?" to "what is this worth *every month*?"

---

## The Five SaaS Pricing Models

| Model | Structure | Best Fit | Revenue Predictability |
|-------|-----------|----------|------------------------|
| **Per-seat (per-user)** | Fixed price × number of users | Collaboration tools, productivity software | High |
| **Flat-rate** | One price for everything | Simple product, uniform customer base | High |
| **Usage-based (consumption)** | Pay per API call, GB, transaction | Infrastructure, APIs, data pipelines | Low–Medium |
| **Tiered** | Fixed price per tier (feature-gated) | Products with clear usage segments | High |
| **Freemium** | Free base + paid upgrade | High viral coefficient, self-serve growth | Low (depends on conversion) |

These are not mutually exclusive. Most mature SaaS products combine them — e.g., tiered plans where each tier has a per-seat price up to a seat cap, then usage-based overages.

---

## Model Selection Decision Framework

```
Is your product's value tied to a countable resource (API calls, storage, messages)?
  YES → Usage-based is honest and aligns incentives. Consider hybrid (flat base + usage overage).
  NO ↓

Do your customers have very different usage levels (1-seat freelancer vs 200-seat enterprise)?
  YES → Per-seat or tiered. Flat-rate will either overprice small customers or underprice large ones.
  NO ↓

Is viral/word-of-mouth adoption important to your growth model?
  YES → Freemium. Calculate whether you can afford the free tier (see Freemium Math below).
  NO → Tiered with a clear feature ladder.
```

### Per-Seat vs Usage-Based: The Core Tradeoff

**Per-seat** aligns with how buyers budget (headcount is a known quantity). It creates a ceiling on expansion revenue per account — you can only earn more by adding seats.

**Usage-based** aligns revenue with value delivery. It expands naturally as customers grow. But it creates **revenue volatility**: a customer who reduces usage next month costs you MRR without churning. Forecasting is harder.

Rule of thumb: if your product's value is primarily about *enabling* a person (communication, project management, writing), go per-seat. If it's primarily about *processing* something (data, transactions, API calls), go usage-based.

---

## Tier Architecture: Designing the Feature Ladder

The goal of a tier structure is to route each customer segment to the tier that maximizes their perceived value while capturing a fair share of it.

### The Three-Tier Template

Most SaaS products use three named tiers. The names themselves communicate positioning:

```
Starter / Basic / Free     ←  Acquisition tier (low or zero cost)
Pro / Growth / Standard    ←  Primary revenue tier (target 60–70% of paying customers)
Business / Team / Plus     ←  Expansion tier (upsell target)
Enterprise                 ←  Optional fourth tier, sold not bought
```

### Feature Gating Principles

**Gate by outcome, not by feature count.** Customers shouldn't feel punished — they should feel that the higher tier solves a bigger problem.

| Effective Gates | Ineffective Gates |
|----------------|-------------------|
| Number of team members / workspaces | Number of templates (feels arbitrary) |
| Automation / API access | Color themes |
| Advanced analytics / reporting | Export to CSV |
| SSO / SAML / audit logs (for Enterprise) | Priority support (often hollow) |
| Data retention period | Storage (unless storage is the product) |

**The limit gate pattern**: Starter gets X of something; Pro gets unlimited. Customers feel the limit naturally as they grow.

Example — project management SaaS:
```
Starter  NT$99/user/month   → 5 active projects, no automation
Pro      NT$299/user/month  → Unlimited projects, automation, reporting
Business NT$499/user/month  → Everything in Pro + SSO, admin controls, audit log
```

### Applying the Decoy Effect (from parent SKILL.md)

The decoy tier is usually **Starter** — priced close enough to Pro that upgrading feels obvious, but limited enough that power users feel genuinely constrained.

The Business tier serves as a **price anchor** for Pro: NT$299 feels reasonable when NT$499 exists. This is why Enterprise pricing is often not shown publicly — showing an extreme anchor on the pricing page cheapens lower tiers.

**Decoy test**: Ask "would a reasonable new customer look at Starter, intend to stay on Starter, but then feel nudged to upgrade within 30-90 days?" If yes, the decoy is working. If Starter is good enough for most customers forever, you've mis-tiered.

---

## Freemium Math

Freemium is not "free pricing" — it's a **customer acquisition channel**. The economics work only if the cost to serve free users is covered by the conversion revenue.

### Freemium Unit Economics Formula

```
Freemium CAC = (Cost to serve N free users for 1 month) / (N × conversion rate)
```

Compare this to your blended CAC from paid channels.

**Worked example:**
- Monthly server + support cost per free user: NT$15
- Free users in cohort: 1,000
- Conversion rate (free → paid within 90 days): 4%
- Average paid plan: NT$299/month

```
Monthly cost to serve 1,000 free users = NT$15,000
Converted users = 1,000 × 4% = 40
Freemium CAC = NT$15,000 / 40 = NT$375 per converted customer
```

If your paid acquisition CAC (Google Ads, sales) is NT$2,000, freemium at NT$375 CAC is excellent. If it's NT$200 (strong SEO, viral loop), freemium may not be worth the infrastructure cost.

### Conversion Rate Benchmarks

Industry averages are noisy and product-specific, but rough guidance:

| Product type | Typical free→paid conversion |
|-------------|------------------------------|
| Developer tools / technical SaaS | 5–15% |
| B2B collaboration / productivity | 3–8% |
| Consumer SaaS | 1–5% |
| Infrastructure / API | 10–25% (trial model) |

**Warning**: A 1% conversion rate is not a failure if free users are also a referral engine. Model the referral-adjusted CAC.

### The Freemium Trap

If free tier is too generous, customers have no reason to upgrade. Common signs:
- Free users stay on free for 12+ months without hitting limits
- Support volume from free users exceeds revenue contribution
- Paid features are "nice to have" but free tier covers 80% of use cases

Fix: tighten the free tier limits or deepen the paid feature differentiation — do not lower the paid price as the first response.

---

## Expansion Revenue: The SaaS Pricing Multiplier

One-time product pricing ends at the sale. SaaS pricing compounds through **expansion revenue** — additional revenue from existing customers via upsell, cross-sell, and seat growth.

### Net Revenue Retention (NRR)

NRR measures whether your revenue base is growing or shrinking from existing customers alone, independent of new customer acquisition.

```
NRR = (MRR start + Expansion MRR − Contraction MRR − Churned MRR) / MRR start × 100%
```

| NRR | Interpretation |
|-----|---------------|
| > 120% | Excellent — existing customers are funding growth |
| 100–120% | Good — customer base self-sustaining |
| 90–100% | Acceptable — some churn/contraction, offset by expansion |
| < 90% | Problem — pricing or product-market fit issue |

**Why this matters for pricing design**: a pricing model that enables NRR > 100% means the company can grow revenue even with zero new customers. This is only possible if the pricing architecture **expands with customer success** — per-seat grows as teams hire; usage-based grows as customers process more; tiered grows as customers hit limits and upgrade.

### Designing for Expansion

Build expansion triggers into the pricing structure before launch:

| Expansion lever | How to design it in |
|----------------|---------------------|
| Seat growth | Per-seat or per-user pricing |
| Usage growth | Usage-based overages or step-up tiers |
| Feature adoption | Feature-gated upgrades with in-product prompts |
| Department expansion | Multi-workspace / multi-team pricing |
| Compliance / security needs | Enterprise tier with SSO, audit, SLA |

The worst outcome is a customer who grows significantly (more users, more usage, more revenue for them) but pays you the same flat fee. Value-based pricing (from parent SKILL.md) is the solution: tie the price to the metric that scales with customer success.

---

## SaaS Price Sensitivity: Willingness to Pay by Segment

B2B SaaS buyers think in ROI, not sticker price. The question they're asking is not "is NT$299 cheap?" but "does this save us more than NT$299/month per user?"

### The Value Metric Frame

For each customer segment, identify the **value metric**: the unit of outcome your product delivers.

| SaaS category | Value metric | Pricing implication |
|---------------|-------------|---------------------|
| Project management | Hours saved per user | Per-user pricing justified by productivity ROI |
| Email marketing | Revenue per send / open rate | % of revenue or per-subscriber |
| Customer support | Tickets resolved / agent | Per-agent seat |
| Analytics | Decisions improved / queries run | Usage-based or per-seat (analyst) |
| E-signature | Contracts signed | Per-document or per-seat |

When you can quantify the value metric, you can calculate the **value-to-price ratio** (V/P). Target V/P ratio of 10:1 for B2B SaaS: if the product saves NT$3,000/user/month, NT$299/user/month is credible. NT$1,500/user/month may stall sales even if technically justified.

### SMB vs Enterprise Pricing

Do not use the same pricing page for both. SMBs self-serve and need transparent pricing with instant credit card checkout. Enterprise needs custom quotes, procurement processes, and negotiation room.

Typical structure:
```
SMB tiers (NT$X–NT$Y/user/month):  published, self-serve
Enterprise:                         "Contact sales" — custom contract, annual commitment
```

Enterprise deals typically carry a 30–60% premium over the equivalent per-seat cost on the public pricing page, justified by SLA, onboarding, custom integrations, and security review.

**Do not publish Enterprise pricing.** The moment you anchor Enterprise at NT$999/user, you've created a ceiling expectation. Sales needs room to price NT$1,500–NT$2,000/user for large, complex accounts.

---

## Annual vs Monthly Billing: The Cash Flow Decision

Always offer annual billing with a discount. The standard discount is 15–20% (equivalent to 2 free months).

```
Annual plan value = Monthly × 12 × (1 − discount)
Example: NT$299/month × 12 × 0.83 = NT$2,978/year (vs NT$3,588 monthly)
```

Why this matters for pricing design:
1. **Cash flow**: annual billing provides capital upfront
2. **Churn reduction**: annual customers churn at 1/4 the rate of monthly customers (they have to make an active renewal decision once, not twelve times)
3. **Commit signal**: annual customers are more invested; they onboard more thoroughly

Make annual the **default selection** on the pricing page, with monthly as the alternative. The psychological default drives a meaningful % of customers to annual without reading the fine print.

---

## Worked Example: Pricing a B2B SaaS for Taiwan SMBs

**Product**: AI meeting notes tool for Taiwanese SMBs  
**Market**: Companies with 5–100 employees, Chinese-speaking users

### Step 1: Three Anchors

```
Cost floor: NT$80/user/month
  (AI inference cost + storage + customer support allocation)

Competitor reference:
  Otter.ai: ~NT$450/user/month (USD pricing, converted)
  Fireflies.ai: ~NT$380/user/month
  Local alternative: none at feature parity

Customer ceiling: NT$600/user/month
  (Based on 25 SMB interviews: 1 meeting/day × 20 min saved × NT$300/hr wage = NT$1,800/user/month value)
  (Target V/P ratio 3:1 for SMB — they're more price-sensitive than enterprise)
```

### Step 2: Tier Design

```
Starter   Free
  → 3 meetings/month, 30-min limit
  → Freemium acquisition channel

Pro       NT$249/user/month (annual: NT$2,490/year)
  → Unlimited meetings, full transcripts, action item extraction
  → Primary revenue tier

Team      NT$399/user/month (annual: NT$3,990/year)
  → Pro + team analytics, shared workspace, admin controls
  → Upsell for managers / ops teams
```

**Decoy check**: Starter is genuinely limited (3 meetings/month is enough for a trial, not a workflow). Pro at NT$249 is well below the NT$380–450 competitor range, creating a clear value signal. Team at NT$399 looks affordable next to competitors' base pricing.

### Step 3: Freemium Economics Check

```
Cost to serve 1 free user/month: NT$80 (at cost floor)
Target conversion rate: 6% (meeting tool category, product-led growth)
Conversion timeline: 90 days

For a cohort of 500 free users:
  Monthly cost = 500 × NT$80 = NT$40,000
  90-day total cost = NT$120,000
  Converted = 500 × 6% = 30 users
  Freemium CAC = NT$120,000 / 30 = NT$4,000

Acceptable if LTV > NT$4,000 × 3 = NT$12,000
  Pro plan annual = NT$2,490; need ~5 year retention to justify
  → Freemium may not work at NT$80/user cost floor unless conversion exceeds 10%
  → Consider limiting free tier to reduce serving cost (e.g., 1 meeting/week, lower quality transcript)
```

This math is the reason many SaaS products degrade free tier over time — not anticompetitive behavior, but unit economics forcing a rebalance.

### Step 4: NRR Design

```
Expansion levers built into pricing:
  - Adding team members → more seats → linear MRR growth
  - Team plan upgrade when manager joins → seat × price increase
  - Annual upgrade (15% discount but 12× upfront commitment)

Target NRR: 105%
  Achieved if: average account grows from 2.5 seats to 2.8 seats within 12 months
  (This is a modest assumption — 0.3 additional users per account over a year)
```

---

## Common SaaS Pricing Mistakes

**Setting per-seat price below cost floor after applying team discounts.** If you offer "50% off for teams of 10+", model the new effective per-seat price against your cost structure. A NT$299 plan discounted 50% for 10 seats = NT$1,495/month. At NT$80/user cost floor, cost is NT$800/month. Margin holds. But if cost floor is NT$160/user (larger AI inference), cost = NT$1,600 > revenue.

**Annual discount too deep.** 30–40% annual discount is common in hyper-competitive markets but destroys ARR predictability. 15–20% is standard; 25% maximum unless you have evidence it materially improves conversion rate.

**Free tier that cannibalizes paid.** If free users can export data, integrate with third-party tools, or invite unlimited collaborators, there's no forcing function to upgrade. Gate these specifically on paid tiers.

**Charging per seat for a product used by one person on behalf of many.** A social media scheduling tool used by one marketing manager for a 50-person company should not be priced per employee — it should be priced per workspace or per connected social account (usage metric).

**Ignoring price-tier cognitive load.** More than 4 public tiers creates analysis paralysis. If you need more segments, hide them behind "Compare all plans" or use an interactive pricing calculator rather than a static table.