ecom-conversational

asgard-ai-platform ｜ 商業與營運 ｜ 原文 ｜ MIT

設計 LINE、IG 等訊息平台上的對話式購物流程

---
name: "ecom-conversational"
description: "設計跨訊息平台的對話式商務體驗，包含聊天機器人流程、商品卡片與轉換策略。當使用者需要透過 LINE、WhatsApp、Instagram DM 或其他訊息管道銷售、設計聊天機器人購物流程，或把訊息整合進銷售漏斗時使用。就算使用者只說「用 LINE 賣東西」、「做一個購物聊天機器人」、「客人都在私訊問要買」或「對話式銷售策略」也要用。"
metadata:
  category: "WP-01 電商"
  tags: ["e-commerce", "conversational-commerce", "chatbot", "messaging"]
---

# 對話式商務

## 框架

```
IRON LAW: Conversation First, Commerce Second

Conversational commerce works because it meets customers WHERE THEY
ALREADY ARE (messaging apps). Forcing a sales pitch into a chat channel
kills trust. The conversation must provide genuine value (answering questions,
solving problems) BEFORE introducing products or purchases.

The sequence: Help → Trust → Recommend → Convert
```

### 平台比較（以亞太地區為主）

| 平台 | 使用者數（台灣） | 商務功能 | 最適合 |
|----------|---------------|------------------|----------|
| **LINE** | 2,100 萬以上（滲透率 95%） | LINE 購物、Rich Menu、LIFF、支付 | 台灣、日本、泰國的主要管道 |
| **Instagram DM** | 約 1,000 萬 | 商品標籤、快速回覆、商品貼圖 | 視覺型商品、較年輕的族群 |
| **Facebook Messenger** | 約 1,800 萬 | 商店整合、自動回覆 | 觸及廣、較年長的族群 |
| **WhatsApp** | 台灣使用有限 | 型錄、購物車、支付（部分市場） | 東南亞、印度、巴西 |
| **WeChat** | 約 100 萬（台灣） | 小程式、微信支付 | 與中國有往來的企業 |

### 對話流程設計

**1. 進入點**：客戶怎麼開始對話
- 門市的 QR code、廣告的點擊後傳訊息、網站聊天視窗、社群連結

**2. 歡迎流程**：前 3 則訊息
- 親切問候、設定期待、提供導覽選項
- 用 Rich Menu（LINE）或快速回覆按鈕，一開始不要問開放式問題

**3. 商品探索**：幫他們找到需要的東西
- 引導式提問：「你在找什麼？」→ 類別 → 商品
- 附圖片、價格和「購買」按鈕的商品卡片
- 依對話脈絡做 AI 推薦

**4. 購買流程**：把阻力降到最低
- 盡量在聊天室內完成結帳（LINE Pay、App 內支付）
- 如果要導到網站，要 deep link 到特定商品（不是首頁）
- 附追蹤資訊的訂單確認訊息

**5. 購後**：留住客戶並追加銷售
- 用訊息通知出貨進度
- 後續關心：「商品用得怎麼樣？」（到貨後 7 天）
- 依購買紀錄做個人化推薦

### 機器人 vs 轉真人

| 情境 | 由機器人處理 | 轉給真人 |
|----------|----------------|------------------|
| FAQ（營業時間、運送、退貨） | ✓ | — |
| 商品推薦（簡單的） | ✓ | — |
| 複雜的商品問題 | — | ✓ |
| 客訴 / 問題 | — | ✓（立即） |
| 高單價購買 | 機器人協助 → 真人成交 | ✓ |

**關鍵指標**：機器人自助解決率（不需真人就解決的比例），成熟的機器人目標是 60-70%。

## 輸出格式

```markdown
# Conversational Commerce Plan: {Business}

## Channel Selection
- Primary: {platform} — rationale: {why}
- Entry points: {QR / ad / social / website}

## Conversation Flow
1. Welcome: {message template}
2. Discovery: {question flow}
3. Product Card: {template}
4. Checkout: {in-chat / redirect}
5. Post-purchase: {follow-up sequence}

## Bot vs Human Split
| Scenario | Handler | SLA |
|----------|---------|-----|
| {scenario} | Bot/Human | {response time} |

## KPIs
| Metric | Target |
|--------|--------|
| Response time | < {X} seconds (bot) / < {X} minutes (human) |
| Containment rate | > 60% |
| Conversation-to-purchase rate | > {X%} |
| Customer satisfaction (CSAT) | > 4.0/5 |
```

## 常見陷阱

- **LINE 官方帳號的方案等級很重要**：免費方案限制每月訊息數，超過就會被限流。客戶數超過 500 就要編預算升級付費方案。
- **不要過度自動化**：機器人聽不懂問題，一直循環「我不太明白，請從選單中選擇」，會比沒有機器人更快毀掉信任。一定要提供轉真人的路徑。
- **訊息是非同步的**：跟電話不同，客戶預期傳了訊息之後晚點再回來。流程要能承受中斷：保存購物車狀態、記住上下文。
- **訊息中的隱私**：聊天紀錄是個人的。沒有取得同意，不要在內部分享對話，也要對資料用途保持透明。
- **社群電商在東南亞/台灣快速爆發**：LINE 購物、Instagram 購物、TikTok Shop 讓社群和商務的界線變得模糊。把它們當成主要管道，不是附加選項。

## 參考資料

- LINE 官方帳號的設定，見 `references/line-oa-setup.md`
- 聊天機器人 NLU 設計模式，見 `references/chatbot-design.md`

---
name: "ecom-conversational"
description: "Design conversational commerce experiences across messaging platforms including chatbot flows, product cards, and conversion strategies. Use this skill when the user needs to sell through LINE, WhatsApp, Instagram DM, or other messaging channels, design chatbot purchasing flows, or integrate messaging into their sales funnel — even if they say 'sell through LINE', 'build a shopping chatbot', 'customers ask to buy in our DMs', or 'conversational sales strategy'."
metadata:
  category: "WP-01 電商"
  tags: ["e-commerce", "conversational-commerce", "chatbot", "messaging"]
---

# Conversational Commerce

## Framework

```
IRON LAW: Conversation First, Commerce Second

Conversational commerce works because it meets customers WHERE THEY
ALREADY ARE (messaging apps). Forcing a sales pitch into a chat channel
kills trust. The conversation must provide genuine value (answering questions,
solving problems) BEFORE introducing products or purchases.

The sequence: Help → Trust → Recommend → Convert
```

### Platform Comparison (Asia-Pacific Focus)

| Platform | Users (Taiwan) | Commerce Features | Best For |
|----------|---------------|------------------|----------|
| **LINE** | 21M+ (95% penetration) | LINE Shopping, Rich Menu, LIFF, payment | Taiwan, Japan, Thailand primary channel |
| **Instagram DM** | ~10M | Shop tags, quick replies, product stickers | Visual products, younger demographic |
| **Facebook Messenger** | ~18M | Shops integration, automated responses | Broad reach, older demographic |
| **WhatsApp** | Limited in Taiwan | Catalog, cart, payment (select markets) | SEA, India, Brazil |
| **WeChat** | ~1M (Taiwan) | Mini Programs, WeChat Pay | China-connected businesses |

### Conversation Flow Design

**1. Entry Points** — How customers start the conversation
- QR code in store, ad click-to-message, website chat widget, social media link

**2. Welcome Flow** — First 3 messages
- Greet warmly, set expectations, offer navigation options
- Rich menu (LINE) or quick reply buttons — don't ask open-ended questions early

**3. Product Discovery** — Help them find what they need
- Guided questions: "What are you looking for?" → category → product
- Product cards with image, price, and "Buy" button
- AI-powered recommendations based on conversation context

**4. Purchase Flow** — Minimize friction
- In-chat checkout where possible (LINE Pay, in-app payment)
- If redirecting to website, deep-link to the specific product (not homepage)
- Order confirmation message with tracking

**5. Post-Purchase** — Retain and upsell
- Shipping updates via message
- Follow-up: "How's the product?" (7 days after delivery)
- Personalized recommendations based on purchase history

### Chatbot vs Human Handoff

| Scenario | Handle with Bot | Hand off to Human |
|----------|----------------|------------------|
| FAQ (hours, shipping, returns) | ✓ | — |
| Product recommendations (simple) | ✓ | — |
| Complex product questions | — | ✓ |
| Complaints / issues | — | ✓ (immediately) |
| High-value purchases | Bot assists → human closes | ✓ |

**Key metric**: Bot containment rate (% resolved without human) — target 60-70% for mature bots.

## Output Format

```markdown
# Conversational Commerce Plan: {Business}

## Channel Selection
- Primary: {platform} — rationale: {why}
- Entry points: {QR / ad / social / website}

## Conversation Flow
1. Welcome: {message template}
2. Discovery: {question flow}
3. Product Card: {template}
4. Checkout: {in-chat / redirect}
5. Post-purchase: {follow-up sequence}

## Bot vs Human Split
| Scenario | Handler | SLA |
|----------|---------|-----|
| {scenario} | Bot/Human | {response time} |

## KPIs
| Metric | Target |
|--------|--------|
| Response time | < {X} seconds (bot) / < {X} minutes (human) |
| Containment rate | > 60% |
| Conversation-to-purchase rate | > {X%} |
| Customer satisfaction (CSAT) | > 4.0/5 |
```

## Gotchas

- **LINE Official Account tiers matter**: Free tier limits monthly messages. If you exceed, messages are throttled. Budget for premium tier if customer base > 500.
- **Don't over-automate**: A bot that can't understand the question and loops "I didn't understand, please choose from the menu" destroys trust faster than no bot at all. Always offer a human escalation path.
- **Messaging is asynchronous**: Unlike phone calls, customers expect to message and come back later. Design flows that work with interruptions — save cart state, remember context.
- **Privacy in messaging**: Chat history is personal. Don't share conversations internally without consent, and be transparent about data usage.
- **Social commerce is exploding in SEA/Taiwan**: LINE Shopping, Instagram Shopping, TikTok Shop — these blur the line between social and commerce. Treat them as primary channels, not add-ons.

## References

- For LINE Official Account setup, see `references/line-oa-setup.md`
- For chatbot NLU design patterns, see `references/chatbot-design.md`