一笔订单的完整旅程，与利润的真相 · AI 与电商产业 · 2026-07One order's complete journey, and the truth about profit · AI & e-commerce · Jul 2026

AI 没有整体替代电商，而是重构了利润分配、流量入口与人力结构——它吃掉了美工客服投放这些表层劳动，却攻不下履约、账期、退货这五块硬骨头 AI didn't replace e-commerce wholesale — it reallocated profit, migrated the traffic gateway, restructured labor. It ate the surface labor of design, service and ads, yet can't crack the five hard bones of fulfillment, credit terms and returns

诚实层＝平台税的数学：商家净利 = GMV × 毛利率 − 平台佣金 − 营销投流 − 履约 − 退货逆向。单品综合抽成常达售价 35–45%、抖音自播间投流占 GMV 常态 50%+、双十一综合退货率 61.5%——绝大多数卖家已沦为给平台算法与物流打工的「数字佃农」。「开店致富」在数学上站不住脚。The honesty layer = the math of the platform tax: net profit = GMV × margin − commission − ad spend − fulfillment − returns. Per-item take often hits 35–45% of price, Douyin self-broadcast ad spend is routinely 50%+ of GMV, and Singles' Day returns hit 61.5% — most sellers are now 'digital sharecroppers' working for the platform's algorithm and logistics. 'Get rich by opening a store' doesn't add up.

范式层＝流量入口第三次迁移：搜索货架（人找货）→ 内容推荐直播（货找人）→ AI 导购 / 智能体代客决策。买方 Agent 与卖方 Agent 后台比价谈判，信息差利润不可逆消亡，回归绝对供应链性价比；针对人类心理设计的营销文案，在冷静理性的 AI 面前彻底失效。但 OpenAI 在 2026-03 收缩原生结账，证明它仍是「进行时」。The paradigm layer = the third migration of the traffic gateway: search shelf (people find goods) → content feed (goods find people) → AI shopping agents deciding for you. Buyer and seller agents haggle in the backend, information-asymmetry profit dies irreversibly, and price returns to raw supply-chain value; copy designed for human psychology fails before a coldly rational AI. But OpenAI pulled back native checkout in Mar 2026 — proof it's still 'in progress.'

主脊是一笔订单的八节点旅程（选品→采购→内容上架→流量→转化客服→履约→退货逆向→复购），每节点标注「已完成时」（AI 已吃掉）vs 「进行时」（尚未跑通）。另含平台税瀑布、AI 吃掉的表层 vs 攻不下的硬骨头、五块硬骨头、龙头分栏、五种电商模式。这是一张批判性行业解剖，不是开店教程。相邻议题见姊妹图：B2B 销售漏斗 / AI SDR / CRM→sell、达人 KOL 内容变现→creator、虚拟模特 / 颜值经济→beauty、平台商业模式 / take rate→money。 The spine is one order's eight-node journey (selection → sourcing → listing → traffic → conversion/service → fulfillment → returns → repurchase), each tagged 'done' (AI already ate it) vs 'in progress' (not yet working). Plus the platform-tax waterfall, the eaten surface vs the uncrackable bones, five hard bones, leader columns, and five commerce models. A critical dissection, not a how-to-open-a-store guide. Adjacent topics on siblings: B2B sales funnel / AI SDR / CRM → sell, KOL content monetization → creator, virtual models / the looks economy → beauty, platform business models / take rate → money.

传统节点Traditional

AI 已吃掉 · 表层劳动（已完成时）AI ate it · surface labor (done)

AI 导购 · Agent 第三次迁移（进行时）AI shopping · agent migration (in progress)

平台税 · 退货黑洞 · 危机Platform tax · returns · crisis

硬骨头 · 供应链上游 · 真正杠杆Hard bones · upstream · real leverage

61.5%

2025 双十一综合退货率，服装直播领域超 80%（女装大促极端达 90%）——一件 100 元女装退货商家亏 50–80 元，前端卖 3 单才抵 1 单退货2025 Singles' Day blended return rate, over 80% for live-stream apparel (extreme 90% for womenswear) — a ¥100 dress returned loses the seller ¥50–80, so 3 sales are needed to offset 1 return

35–45%

单品平台综合抽成（佣金+FBA+广告，$25 商品，Marketplace Pulse）；亚马逊综合成本从 2016 约 1/3 涨到 2022 超 50%——利润大头流向平台，卖家＝数字佃农Per-item blended platform take (referral+FBA+ads, a $25 item, Marketplace Pulse); Amazon's blended cost rose from ~1/3 in 2016 to over 50% in 2022 — profit flows to the platform, sellers are digital sharecroppers

+693%

2025 假日季生成式 AI 导流零售网站同比暴增（Adobe，超 1 万亿次访问），AI 来源访客转化率高 31%——流量入口正第三次迁移向 AI 导购Generative-AI referrals to retail sites surged YoY in the 2025 holidays (Adobe, 1T+ visits), with AI-sourced visitors converting 31% higher — the traffic gateway is migrating a third time toward AI shopping

45.22%

2025 上半年商家投诉中「任意仅退款」占比第一（网经社）；2025 年 3–4 月主流平台密集取消强制仅退款——平台规则的政治性是攻不下的硬骨头'Refund-only at will' was the #1 merchant complaint in H1 2025 (100EC); in Mar–Apr 2025 major platforms scrapped forced refund-only — the politics of platform rules is an uncrackable bone

口径警告：本页是批判性行业分析，非开店教程 / 投资建议，并置厂商自述与独立数据。货币化率 / 平台抽成有多口径冲突（拼多多 4.55% vs 7.6% 因分母 GMV 口径不同；亚马逊 22–29% 占平台收入 vs 单品 35–45% vs 占卖家收入 1/3→50%，统计对象不同，已逐条标注）。退货率 61.5%（双十一综合）/ 女装 80–90% / 美国线上 19–20% 因类目地区口径不同。平台自报的 AI 效果（CTR +25% / Listing +40% / 90% 几乎不改 / Rufus 120 亿增量）多为厂商口径、未经独立审计；市场规模（McKinsey 3–5 万亿）为 2030 预测。agentic commerce 归「进行时」（OpenAI 2026-03 收缩结账）。每张卡片右上角 A/B/C/D＝证据强度。 Basis warning: a critical industry analysis, not a store guide / investment advice, placing vendor claims beside independent data. Take rate / platform cut carries conflicting bases (PDD 4.55% vs 7.6% differ by GMV denominator; Amazon 22–29% of platform revenue vs 35–45% per item vs 1/3→50% of seller revenue — different objects, flagged individually). Return rates 61.5% (SD blended) / 80–90% womenswear / 19–20% US online differ by category and region. Platform-reported AI effects (CTR +25% / listing +40% / 90% barely edited / Rufus $12B incremental) are largely vendor figures, unaudited; market sizes (McKinsey $3–5T) are a 2030 forecast. Agentic commerce is filed as 'in progress' (OpenAI pulled back checkout Mar 2026). Each card's top-right A/B/C/D = evidence strength.

◆ 诚实层 · 平台税的数学The honesty layer · the math of the platform tax

100 元 GMV，流向了哪里？Where does ¥100 of GMV actually go?

判断任何电商技术变革，必须先剥离平台公关的虚假繁荣，直视利润流向。下面是一件商品售价的示意拆解——层层扣掉货品成本、平台佣金、营销投流、履约、退货逆向后，商家净利常只剩一条细缝。AI 不主要把 GMV 做大，而在哪项成本被压缩、哪项被抬高、哪项从显性变隐性。To judge any e-commerce tech shift, first strip the platform-PR gloss and look at where profit flows. Below is an illustrative breakdown of one item's price — after goods cost, commission, ad spend, fulfillment and returns are peeled off layer by layer, the seller's net is often a sliver. AI doesn't mainly grow GMV; it changes which cost is compressed, which is raised, which goes from visible to hidden.

▚ 一件商品售价的去向（示意 · 非精确，锚定行业区间）Where one item's price goes (illustrative, anchored to industry ranges)

综合抽成 35–45% + 退货损耗 + 履约 + 货本，挤出「数字佃农」的真实薄利A 35–45% blended take + returns + fulfillment + goods cost squeeze out the 'digital sharecropper's' true thin margin

38%货品成本goods

16%平台佣金commission

20%营销投流ad spend

11%履约fulfillment

11%退货逆向returns

货品成本（供应链）Goods (supply chain)

平台佣金 / 抽成Platform commission

营销投流（抖音自播 50%+）Ad spend (Douyin 50%+)

履约仓配Fulfillment

退货逆向损耗Returns loss

商家净利（细缝）Seller net (the sliver)

货币化率对照：亚马逊 16% / eBay 10.82% vs 国内起步期阿里 3.66% / 京东 5.5% / 拼多多 2.17%——但低税率时代已终结：拼多多货币化率约 7.6%（2024）几乎是淘天 3.9% 的两倍。巨头防御性天量营销：阿里 2025 财年营销费用激增 77.67% 至 1916 亿元、京东增 75.07% 至 839.5 亿、拼多多 1253 亿——为争夺存量燃烧的费用，最终通过更贵的竞价广告系统性转嫁到商家获客成本。年销破亿的女装商家，扣近千万退货费后净利仅两三百万。Take-rate comparison: Amazon 16% / eBay 10.82% vs China's early days Alibaba 3.66% / JD 5.5% / PDD 2.17% — but the low-tax era is over: PDD's take is ~7.6% (2024), nearly double Taotian's 3.9%. Giants' defensive ad blitz: Alibaba's FY2025 marketing spend jumped 77.67% to ¥191.6B, JD +75.07% to ¥83.95B, PDD ¥125.3B — money burned to fight for existing users is systematically passed to sellers' acquisition cost via pricier auction ads. A womenswear seller with ¥100M+ sales nets only ¥2–3M after ~¥10M in return costs.

▸ 退货黑洞 · 诚实指标The returns black hole · the honest metric

卖得越多，亏得越快The more you sell, the faster you bleed

直播+内容推荐缩短交易链路、刺激冲动消费，却严重透支售后：女装日常退货 50–60%、大促叠加主播激情+夸张滤镜飙至 80–90%。运费险从可选增值变成事实标配，彻底摧毁试错成本边界。一旦退货，双向运费+运费险+仓储分拣+清洗除味+残次折旧全是隐性成本——财报高 GMV，现金流枯竭。美国线上服饰退货率也达 22–26%（线下仅 6.2%）。Live + content feeds shorten the path and spur impulse buys, but wreck after-sales: womenswear returns run 50–60% daily, spiking to 80–90% in promos with hype and heavy filters. Return-shipping insurance went from optional to de facto standard, erasing the cost of trying things on. On a return, two-way shipping + insurance + re-sorting + cleaning + defect write-down are all hidden costs — high GMV on paper, cash flow bone-dry. US online apparel returns also hit 22–26% (vs 6.2% in-store).

▸ 仅退款 · 平台规则的政治性Refund-only · the politics of platform rules

从「合法劫掠」到集体纠偏From 'legal looting' to a collective reversal

拼多多 2021 在生鲜首推「仅退款」，本为惩罚劣质商家；却向全品类蔓延、被淘宝京东抖音效仿，异化为羊毛党 / 黑产「合法劫掠」商家的工具。2025 上半年商家投诉中「任意仅退款」占 45.22%（第一）、「任意罚款」11.94%、「扣押保证金」8.64%。直到 2025 年 3–4 月，监管+商家承压双重因素下，主流平台密集取消强制仅退款——标志中国电商从「讨好买家、流量优先」向 「公平商业博弈」回归。PDD launched 'refund-only' for fresh food in 2021 to punish bad sellers; it spread to all categories, was copied by Taobao/JD/Douyin, and mutated into a tool for scalpers and fraud rings to 'legally loot' merchants. In H1 2025 'refund-only at will' was 45.22% of complaints (#1), 'arbitrary fines' 11.94%, 'seized deposits' 8.64%. Only in Mar–Apr 2025, under regulatory and merchant pressure, did major platforms scrap forced refund-only — marking China's shift from 'please the buyer, traffic first' back to 'fair commercial contest.'

◆ 判断层杠杆排序 · 结论卡The leverage stack · the conclusion

选品与供应链（上游）Selection & supply chain ＞运营技巧operations ＞视觉优化（已被 AI 商品化）visuals (AI-commoditized)

退货率与复购（诚实指标）Returns & repurchase ＞ GMV（可投流买出的虚荣指标）GMV (a vanity metric you can buy)

核心洞察：真正的杠杆从来不在被 AI 吃掉的环节，而在 AI 尚难渗透的选品与供应链上游。再逼真的 AI 模特 / 数字人 / GEO 也救不了质量低劣、成本倒挂的商品；淘宝家居 AI 图能把点击率 +25%，但线上服饰 24.4% 的退货率足以把再漂亮详情页的利润全吞掉。Core insight: the real leverage is never in what AI eats, but upstream in selection and supply chain, where AI barely reaches. No AI model / digital human / GEO saves a low-quality, cost-inverted product; Taobao's AI home images lift CTR +25%, but a 24.4% apparel return rate swallows the profit of even the prettiest page.

从这张图看到的五条规律Five patterns this map makes visible