← 行銷專欄 Marketing · 行銷學習地圖 W3-5
社交平台 2026 生態
Social Platform Ecosystems 2026
八大平台的演算法邏輯、受眾輪廓、內容偏好——以及「平台選擇 = 受眾在哪 × 內容擅長 × 能持續多久」的決策框架。
Eight platforms' algorithms, audiences, and content preferences — plus the decision framework: audience × content strength × sustainability.
🎯 學完你能做到
🎯 Learning Goals
① 說明三大短影音平台演算法的關鍵差異 ② 為一支影片規劃三平台差異化發布 ③ 評估「搜尋型 vs 社交型」平台的經營差異 ④ 為自己的產品選出 1-2 個深耕平台
① Explain key algorithm differences across the video Big 3 ② Plan differentiated cross-platform publishing ③ Distinguish search-type vs social-type platforms ④ Pick 1-2 platforms to go deep on
⚠️ 演算法資訊時效敏感（研究於 2026-07）。「權重百分比」多為業界逆向分析估計；官方確認的訊號會特別標註。
⚠️ Algorithm info is time-sensitive (researched 2026-07). Percentage weights are industry estimates unless marked official.
TikTok
每支影片都是新的抽獎 · FYP 分階段派發
Every video is a fresh lottery · Staged FYP distribution
⚙️ 演算法邏輯Algorithm
- 分階段派發（staged distribution）：先投給 200-500 人測試群，表現達標才擴大
- Staged distribution: shown to a 200-500 person test group first; scales only if it performs
- 官方確認：粉絲數不是排名因子——每支影片獨立計分，新帳號也可能爆紅
- Official: follower count is NOT a ranking factor — each video scored independently
- 完播率 + watch time 權重約 40-50%（業界估計）；爆紅門檻完播率已升至 ~70%（2024 年約 50%）
- Completion + watch time ≈ 40-50% weight (industry estimate); viral threshold now ~70% completion
- Shares 與 saves > likes；追蹤 3/15/30/60 秒留存節點；rewatch/loop 率是 2026 新重點
- Shares & saves > likes; tracks 3/15/30/60s retention; rewatch/loop rate is the 2026 focus
🔍 TikTok SEO + 受眾與商業Audience & Commerce
- Caption 關鍵字權重已超過 hashtag；口說語音被轉錄索引、螢幕字卡也計入
- Caption keywords now outweigh hashtags; spoken audio is transcribed and indexed, on-screen text counts
- 搜尋語言偏口語問題式：「運動時不會掉的耳機」而非「無線藍牙耳機」
- Search language is conversational: "earbuds that don't fall out" not "wireless earbuds"
- TikTok Shop：2026 美國銷售額預估破 $20B；商品標題前段放主關鍵字
- TikTok Shop: projected $20B+ US sales in 2026
25-34 歲 35.3%
35+ 成長速度是 IG 的 3 倍
35+ growing 3x faster than IG
頻率：每週 3-5 支
Cadence: 3-5/week
誤區：以為要先養粉絲才有觸及（每支都是新抽獎）；把長片直接剪短不重做 hook；hashtag 塞滿卻不管口說關鍵字。Pitfalls: waiting to build followers first; re-cutting long videos without a new hook; stuffing hashtags while ignoring spoken keywords.
Instagram
DM 分享是最強訊號 · 版位各有邏輯
DM shares are the strongest signal · Each surface has its own logic
⚙️ 演算法（Adam Mosseri 官方確認）Algorithm (Mosseri-confirmed)
| 訊號 | Signal | 說明 | Note |
|---|---|---|---|
| 1. Watch time | 第一排名因子（2025-01 官方確認）；前 3 秒留存是門檻 | #1 factor (official, 2025-01); 3-second retention is the gate | |
| 2. Sends per reach | DM 分享權重約讚的 3-5 倍（業界估計）——觸及新受眾最強訊號 | DM shares ≈ 3-5x a like (estimate) — strongest signal for new reach | |
| 3. Likes per reach | 仍計分但權重最低 | Counts, but lowest weight |
- 版位權重不同：Reels 重 watch time+sends；Feed 重關係強度；Stories 重近期性；Explore 重興趣
- Per-surface logic: Reels = watch time + sends; Feed = relationship; Stories = recency; Explore = interest
- 格式排序：Reels（觸及新受眾）> Carousel（互動與儲存）> 單圖；Stories 養既有粉絲
- Format hierarchy: Reels (new reach) > Carousels (engagement & saves) > single image; Stories nurture existing fans
18-34 歲佔 63%
Gen Z 86%
頻率：feed 週 3-5、Stories 可每日
Cadence: 3-5 feed/wk, daily Stories
誤區：追讚不追 DM 分享；caption 放外連（不可點擊）；忽略前 3 秒 hook。Pitfalls: chasing likes over DM shares; links in captions (not clickable); ignoring the 3-second hook.
YouTube (+ Shorts)
滿意度已超越觀看時長 · Shorts 與長片演算法分離
Satisfaction now beats watch time · Shorts algorithm is separate
⚙️ 演算法重點Algorithm
- 2026 官方確認：viewer satisfaction surveys 權重已超過 raw watch time——8 分鐘看完 100%+讚，勝過 25 分鐘看 40%
- Official 2026: satisfaction surveys outweigh raw watch time — 8 min at 100% + like beats 25 min at 40%
- Shorts 與長片演算法已正式分離：Shorts 表現差不拖累長片推薦
- Shorts and long-form algorithms are now separate: weak Shorts don't hurt long-form
- Shorts 留存門檻：<30 秒約需 65% 完播、30-60 秒約需 50%（業界估計）；已進搜尋結果、支援自訂縮圖
- Shorts thresholds: <30s needs ~65% completion, 30-60s ~50% (estimates); now in search results with custom thumbnails
- SEO 改用語意主題聚類（semantic topic clusters），非單純關鍵字匹配
- SEO uses semantic topic clusters, not simple keyword matching
全年齡覆蓋最廣
Broadest age coverage
Gen Z 91% · Gen X 83%
長片週 1-2、Shorts 可高頻
1-2 long/wk, Shorts more
誤區：硬拉長影片衝 watch time（滿意度低反扣分）；把 Shorts 當獨立 app 經營、不導流長片與訂閱。Pitfalls: padding video length (low satisfaction backfires); treating Shorts as an island instead of funneling to long-form.
Threads vs X (Twitter)
互補而非替代 · 語氣差異大
Complementary, not substitutes · Very different tones
⚔️ 正面對比Head to Head
| X (Twitter) | Threads | ||||
|---|---|---|---|---|---|
| 規模 | Scale | 月活 ~6 億 | ~600M MAU | ~4 億；行動端日活已超車（1.415 億 vs 1.25 億，2026-01） | ~400M; mobile DAU now ahead (141.5M vs 125M) |
| 受眾 | Audience | 偏男性；政治/財經/加密/體育/新聞 | Male-skewed; politics/finance/crypto/sports | 偏年輕；創作者/小型企業/lifestyle | Younger; creators/SMB/lifestyle |
| 演算法性格 | Algorithm | 獎勵即時性與高聲量 | Rewards immediacy & volume | 壓制 outrage、偏好真誠對話 | Suppresses outrage, favors genuine dialogue |
| 變現 | Monetization | 廣告分潤+訂閱+Articles（2026「creator 之年」） | Ad share + subs + Articles | 建置中（背靠 Meta Creator Marketplace） | In progress (Meta Creator Marketplace) |
| 頻率 | Cadence | 每日 2-5 則 | 2-5/day | 每日 1-2 則 | 1-2/day |
誤區：同一內容原封不動雙發（X 犀利即時、Threads 對話溫和）；在 Threads 用 X 式爭議釣魚法會被降權。Pitfalls: posting identical content to both (X is sharp/immediate, Threads conversational); X-style outrage bait gets demoted on Threads.
LinkedIn
Dwell time 為王 · 外連重罰 · 個人帳號才是主力
Dwell time is king · Links punished · Personal profiles rule
⚙️ 三個關鍵機制Three Key Mechanics
- Dwell time 是首要品質指標：停留 61 秒+ 的貼文互動率 15.6% vs 0-3 秒的 1.2%（13 倍）——carousel/document 貼文 dwell time 是純文字的 2-3 倍
- Dwell time is the top quality signal: 61s+ posts get 15.6% engagement vs 1.2% for 0-3s (13x) — carousels get 2-3x the dwell time of text
- 外連重罰：帶外部連結觸及約 -60%；2026 起「連結放留言」繞法也被降權
- External links punished: ~-60% reach; the "link in comments" workaround is also demoted since 2026
- 個人 profile 佔 feed ~65%，company page 僅 ~5%——B2B 有機成長主力是員工個人帳號與高管思想領導
- Personal profiles = ~65% of feed, company pages ~5% — B2B organic growth runs on personal accounts and executive thought leadership
- 黃金 60 分鐘：發文後第一小時互動決定派發；頻率每週 3-5 則
- Golden hour: first-hour engagement determines distribution; 3-5 posts/week
誤區：用公司頁當主力發文管道；貼文塞外連；engagement bait（「同意 +1」已被明確降權）。Pitfalls: company page as main channel; links in posts; engagement bait ("comment +1 if you agree" is explicitly demoted).
小紅書 RED（RedNote）
Xiaohongshu / RED (RedNote)
生活方式搜尋引擎 · 種草文化
Lifestyle search engine · Zhongcao (seeding) culture
⚙️ 生態邏輯Ecosystem Logic
- 70%+ 用戶行為從搜尋出發——內容要當 Google 結果寫，不當社群貼文寫（標題關鍵字、封面圖、清單式乾貨）
- 70%+ of user behavior starts from search — write content like Google results, not social posts
- 種草文化：真實使用心得取代硬銷；~80% 購買決策受種草筆記影響
- Zhongcao culture: authentic reviews over hard selling; ~80% of purchase decisions influenced by seeding notes
- 演算法偏好：真實性（廣告感重會限流）、利基垂直、互動速度
- Algorithm favors: authenticity (ad-feel gets throttled), niche verticals, engagement velocity
- 商業功能：品牌號、薯條（加熱）、蒲公英平台（KOL/KOC 撮合）
- Commerce: brand accounts, paid boosting, Pugongying KOL/KOC marketplace
台灣法規：2025-12-04 起內政部對小紅書實施一年禁令（當時台灣用戶約 300 萬）。策略定位應為「華語市場（中國大陸、馬來西亞、港澳）渠道」而非台灣本地主力。使用前查證禁令現況。Taiwan ban: a one-year government ban started 2025-12-04 (~3M TW users at the time). Position RED as a Chinese-market channel (mainland China, Malaysia, HK/Macau), not a Taiwan-local one. Verify ban status before planning.
誤區：用 IG 思維經營（重人設輕搜尋）；筆記太像廣告被限流；只找頭部 KOL 忽略 KOC 素人矩陣（種草的信任基礎是「像我一樣的普通人」）。Pitfalls: IG-style persona-first operation; ad-like notes get throttled; chasing top KOLs while ignoring KOC networks (trust comes from "people like me").
Pinterest
視覺搜尋引擎，不是社群 · 手作類長尾之王
A visual search engine, not social · Handmade long-tail king
⚙️ 為什麼手作賣家不能忽略它Why Makers Can't Ignore It
- 97% 熱門搜尋是無品牌詞——最適合被「還不知道要買什麼牌子」的人發現
- 97% of top searches are unbranded — ideal for discovery before brand choice
- 購買意圖指數是社群平均 5.6 倍；平均 CPC $0.83 社群類最低；對購物網站 referral 比 Facebook 多 33%
- Purchase intent 5.6x social average; $0.83 avg CPC (lowest); 33% more shopping referrals than Facebook
- Pin 壽命以月/年計（IG 以小時計）——長尾流量資產，對 Etsy 賣家尤其被低估
- Pins live for months/years (IG posts: hours) — an undervalued long-tail asset, especially for Etsy sellers
- 受眾：女性 70.3%；核心類目＝家居、手作、食譜、穿搭、婚禮
- Audience: 70.3% female; core = home, crafts, recipes, fashion, weddings
- 頻率：每週 5-10 Pin 起步；高成效者每日 15-25，但 >50/日傷派發
- Cadence: start 5-10 pins/week; top performers 15-25/day, but >50/day hurts
誤區：當社群追互動（該追搜尋曝光與外連點擊）；不做直式 2:3 + 文字覆蓋；board 命名不關鍵字化。工具：Pinterest Trends（免費關鍵字研究）。Pitfalls: chasing engagement like a social network (chase search impressions & outbound clicks); skipping 2:3 vertical + text overlay; non-keyworded board names. Tool: Pinterest Trends (free).
Facebook
35+ 的第一平台 · 社團 = 平台內 owned audience
The #1 platform for 35+ · Groups = on-platform owned audience
⚙️ 現況與廣告生態Reality & Ads
- 品牌粉專有機觸及持續探底；演算法偏好 meaningful engagement 與朋友/社團內容
- Page organic reach keeps falling; algorithm favors meaningful engagement and friends/Groups content
- 社團（Groups）是最強有機工具：成員收到新貼文通知（粉絲沒有）＝平台內 owned audience
- Groups are the strongest organic tool: members get notifications (page fans don't) = on-platform owned audience
- 受眾：Gen X 與 Boomers 使用率皆 88%——觸及 35+ 的第一平台；FB Reels 受眾比 TikTok 老，適合 35-55
- Gen X & Boomers both at 88% usage — the #1 platform for 35+; FB Reels skews older, good for 35-55
- 廣告 2026：Advantage+（AI 自動投放）成預設；broad targeting + 強素材勝過細分受眾（素材即定向）；first-party data 是最可靠輸入
- Ads 2026: Advantage+ is the default; broad targeting + strong creative beats micro-targeting (creative IS targeting); first-party data is the most reliable input
- 熟齡投放注意：Meta 會把預算集中到 65+（便宜高轉換），要控年齡需手動設限
- Note: Meta concentrates budget on 65+ (cheap conversions); set manual age caps if needed
誤區：宣稱「FB 已死」放棄 35+ 市場；只經營粉專不建社團；在 Advantage+ 時代仍花大量時間手動細分受眾而不迭代素材。Pitfalls: declaring "FB is dead" and abandoning 35+; running only a page without a Group; micro-managing audiences instead of iterating creative in the Advantage+ era.
實作練習 + 平台選擇框架
Practice + Platform Selection Framework
深耕 1-2 平台勝過全平台淺嚐
Going deep on 1-2 beats spreading thin
平台選擇公式：受眾在哪 × 內容形式擅長什麼 × 能持續多久——三者交集才做。
Platform selection: where your audience is × what content you're good at × what you can sustain — only act on the intersection.
練習 1（W3）：三平台差異化發布
Exercise 1 (W3): Differentiated Publishing
做一支捕夢網生成過程 30 秒影片，寫出三平台各自的 caption（TikTok 口語關鍵字長 caption、Reels 100-300 字、Shorts 標題），實際發布並記錄一週數據。
Make one 30s process video; write platform-specific captions (TikTok conversational keywords, Reels 100-300 chars, Shorts title); publish and track for a week.
練習 2（W4）：同主題三種語氣
Exercise 2 (W4): One Topic, Three Voices
用同一主題寫三版：X 版（犀利短句）、Threads 版（對話式）、LinkedIn 版（carousel 大綱、正文無外連），體會平台語氣差異。
Write the same topic three ways: X (sharp), Threads (conversational), LinkedIn (carousel outline, no links in body).
練習 3（W5）：Pinterest 起步 + 平台備忘錄
Exercise 3 (W5): Pinterest Setup + Platform Memo
①用 Pinterest Trends 查捕夢網相關關鍵字，設計 5 個關鍵字化 board 名稱 + 10 個 Pin 標題。②寫一頁備忘錄：你的產品客群在哪三個平台交集最大？選出深耕的 1-2 個。
① Use Pinterest Trends for dreamcatcher keywords; design 5 keyworded boards + 10 pin titles. ② Write a one-page memo: which 1-2 platforms will you go deep on, and why?