零售产业 · 一次拿起放下 × AI · 2026-07The Retail Industry · The Pick-Up Moment × AI · Jul 2026

零售的原子是一次「拿起放下」——AI 重构了这次接触的前与后:选品、补货、价签、选址、导购、复购;却始终碰不到接触本身:触感、即时拥有、逛的乐趣、人情,和租金 Retail’s atom is one pick-up-and-put-down — AI has rebuilt everything before and after that touch: assortment, replenishment, price tags, siting, guidance, repeat; yet it still cannot touch the touch itself: texture, instant possession, the joy of browsing, human warmth — and the rent

渗透的坡度(编者按源报告三层评分锚定):后台运营 8/10 > 前台转化 5–7/10 > 无人化 2–4/10。AI 在零售 ROI 最扎实的应用是补货预测,没有之一;最响亮的失败是无人店。激进版死,朴素版活——这是全页的试金石。The gradient of penetration (anchored on one source’s three-layer score): back office 8/10 > front-of-store 5–7/10 > unmanned 2–4/10. Retail AI’s surest ROI is replenishment forecasting, bar none; its loudest failure, the unmanned store. The radical version dies; the modest version lives — this page’s litmus test.
店的再定价:门店=最便宜的流量入口+最贵的信任背书+现成的履约前置仓。「黄金交叉」正在发生——线上获客成本涨过线下租金;但它是筛选器,不是涨潮:只对好商圈×好店型×好坪效的窄集合成立。The store, repriced: the cheapest traffic gate + the dearest trust endorsement + a ready fulfilment node. The «golden cross» is real — online acquisition costs have climbed past offline rent; but it is a filter, not a rising tide: it holds only for the narrow set of good location × good format × good yield.

双轨主脊:商品线九节点(选品采购→定价促销→物流补货→陈列动线→进店流量→拿起试用→导购转化→收银防损→会员复购)+门店线六节点平行轨(选址→装修→爬坡→稳态→衰退→关店,挂死法图鉴)。版图 C 按业态横切五带(含无人零售墓地带);外加化妆品深挖、产品指南(2C/2B)与立场地图。姊妹分工:shop 讲把货卖到屏幕里,retail 讲把人拉回店里——线上生意→shop,品类纵深→dine/fashion,即时履约的骑手侧→work Twin spines: the merchandise line’s nine nodes (assortment → pricing → replenishment → planogram → footfall → pick-up → guidance → checkout → repeat) plus the store line’s six-node parallel rail (siting → fit-out → ramp → steady state → decline → closure, carrying the death catalog). Map C cuts five format bands (including the unmanned-retail graveyard); plus the cosmetics deep-dive, a product guide (2C/2B) and the stance map. Sibling split: shop covers selling into the screen; retail covers pulling people back into the store — online business → shop, category depth → dine/fashion, the courier side → work.

传统节点Traditional
后台 · AI 主战场Back office · AI’s field
前台 · 人情与体验Front · warmth & experience
地产 · 空间权力Property · spatial power
墓地带 · 无人零售The graveyard
57.4%
2024 中国美妆线上渠道占比(2019 年 40.8%;2021 年线上首超线下;KPMG,B)。展厅化压力最大的品类,也因此成了线下 AI 化最激进的品类——专柜数字化、AR 试妆、O+O 全在这里先跑China’s beauty online share in 2024 (40.8% in 2019; online passed offline in 2021; KPMG, B). The category most showroomed — and therefore the most aggressively AI-fitted offline: counter digitisation, AR try-on and O+O all ran here first
2,300
Walmart 电子价签铺设的美国门店数(2025-07 官方,B)。ESL 是「朴素自动化」的旗舰:不性感、不上头条,把「省 10%」做实——与无人店的命运构成全页对照组US Walmart stores fitted with electronic shelf labels (Jul 2025, official, B). The ESL is modest automation’s flagship: unsexy, unheadlined, making the 10% saving real — the control group to the unmanned store’s fate
7,905
名创优品集团门店数(2025-06-30 财报,A;MINISO 7,612+TOP TOY 293),2025H1 收入 +21.1%,近 12 个月新增门店 74.5% 在海外;2024 年成立 AI 创新部——IP×快反×全球配货的算法化样本MINISO group store count (30 Jun 2025 filings, A; 7,612 MINISO + 293 TOP TOY), H1 revenue +21.1%, 74.5% of new stores overseas in 12 months; an AI innovation unit since 2024 — the algorithmic exhibit of IP × fast reaction × global allocation
2.69
2025 年中国淘汰的化妆品品牌数(监管新规下的行业出清,C)。监管抬高安评与功效门槛——「成分党 vs 品牌溢价」之战的裁判提前进了场:营销滤镜白牌先死Chinese cosmetics brands culled in 2025 (the clean-out under new regulation, C). Safety and efficacy bars raised — the referee entered the «ingredient-hackers vs brand premium» fight early: filter-built white labels die first
口径警告:本页综合四份深度研究交叉整理。四源基线:一份带可核 URL(数字最可信);一份禁网写作、自标「需二次核验」但判断框架最锐(其数字一律 ≤C;2026-07 已重跑为联网深研、94 条可核引文——存量 C 级孤例未逐条升级,新增数字按新源分级);一份引用多为厂商/行业软文(硬数字一律 D);一份正文中途截断、引文列表丢失(同批重跑已补全收束、引文恢复)。厂商宣称的提升数字(预测准确率 92%+/缺货率 −76%/复购 +28%)全部挂 D 角标,只作「厂商口径」展示。⚠️两处定性打架已并存呈现:便利蜂(动态定价单项有效 vs 整体收缩)与 Just Walk Out(技术可行 vs 单位经济学失败,转封闭场景存活——不是「技术失败」)。化妆品市场两口径勿混:KPMG 零售额(2024 线上 5,365 亿)≠ 全渠道交易额(2025 破 1.1 万亿)。本图未覆盖:零售劳动力大盘、行业 AI 采用率调查、美国生物识别法/EU AI Act 对店内 CV 的限制——引用时请注意这些缺口。 Basis warning: cross-compiled from four deep-research reports. Source baseline: one carries verifiable URLs (most trustworthy figures); one wrote offline, self-flagged for re-checking, yet sharpest in framework (its figures capped at C; re-run online in Jul 2026 with 94 verifiable citations — legacy C-graded orphans stand, new figures graded on the new source); one cites mostly vendor/trade-PR material (hard numbers graded D); one truncated mid-text with references lost (completed in the same re-run, citations restored). Vendor uplift claims (92%+ forecast accuracy / −76% stockouts / +28% repeat) all wear D badges and appear only as vendor claims. ⚠️Two clashes are shown side by side: Bianlifeng (dynamic pricing worked; the company shrank) and Just Walk Out (technically viable; unit economics failed; survives in closed venues — not a «tech failure»). Two cosmetics bases must not mix: KPMG retail value (online ¥536.5B, 2024) ≠ all-channel transaction value (¥1.1T+, 2025). Not covered by this map: the retail labour force at large, adoption surveys, and US biometric law / EU AI Act limits on in-store CV — mind these gaps when citing.
中心悖论 · 黄金交叉之辩The central paradox · the golden cross
线上越贵,线下越值钱——但只对好店成立The dearer online gets, the more the store is worth — but only the good store
2024–2025 年,一个反直觉的算术开始流传:线上获客成本涨过线下租金的折算成本,品牌重回线下「从情怀问题变成数学问题」。这是真的吗?两边的账本都摆在下面——读完再下结论。Through 2024–2025 a counter-intuitive arithmetic spread: online customer-acquisition costs climbed past the amortised cost of offline rent, and returning to the street «turned from sentiment into math». Is it true? Both ledgers are laid out below — conclude after reading.
交叉为真:门店的三重再定价The cross is real: the store repriced three ways
店=流量入口+信任背书+前置仓Gate + endorsement + fulfilment node
线上 CAC 指数级攀升,平台扣点+退货+达人费用层层堆高;线下租金进入存量时代、可谈可压。门店的新账法:最便宜的获客渠道(路过即曝光)、最贵的信任背书(实体即担保)、现成的履约前置点(即时零售的库存网络)。名创入股永辉、屈臣氏宣布再开 500 店、品牌回街——方向是真的。Online CAC compounds upward — platform takes, returns and creator fees stack; offline rents entered the stock era, negotiable and compressible. The store’s new arithmetic: the cheapest acquisition channel (passing is exposure), the dearest trust endorsement (bricks as bond), a ready fulfilment node (instant retail’s inventory web). MINISO buying into Yonghui, Watsons announcing 500 new stores, brands returning to the street — the direction is real.
交叉限定版:CBRE 的冷水The qualified cross: CBRE’s cold water
租金便宜,恰因客流在衰退Rent is cheap precisely because footfall fell
CBRE 数据:中国购物中心空置率与首层租金仍承压——「便宜的铺位」很多是衰退的铺位。丝芙兰中国收缩、屈臣氏同时在关低效店:再好的数字化也扛不住渠道结构变迁。交叉只发生在好商圈×好店型×好坪效的窄集合;而且交叉本身会推高好铺位租金,自我削弱CBRE’s data: Chinese mall vacancy and ground-floor rents remain under pressure — many «cheap» sites are declining sites. Sephora China is shrinking while Watsons culls weak stores: no digitisation outruns channel-structure change. The cross holds only for the narrow set of good location × good format × good yield — and by holding, it bids good sites back up, self-attenuating.
判词Verdict黄金交叉是筛选器,不是涨潮:低效卖货店照关,强体验/强履约/强会员的店重新划算。配套试金石一并记住:朴素自动化 > 激进颠覆——把「省 10%」做实的活下来了(ESL/自助收银/防损),「重新发明零售」的进了墓地带(版图 C 第五带)。The golden cross is a filter, not a tide: weak sales-only stores keep closing while experience-heavy, fulfilment-heavy, membership-heavy stores pencil out again. Keep the companion litmus: modest automation > radical reinvention — those who made the 10% real survived (ESL / self-checkout / loss prevention); those who «reinvented retail» went to the graveyard (Map C, band five).
Reading the MapReading the Map

从这张图带走的五条规律Five patterns to take away

立场声明:本页是批判性、祛魅的行业结构分析,用 A–D 角标区分财报/官方与厂商软文,口径打架处标 ⚠️ 两边并存(便利蜂/JWO 两案即样本)。产品指南是市场地图,不构成经营、投资或消费建议、不构成推荐背书。核心判断一句话:AI 重构了「拿起放下」的前与后,碰不到接触本身——经验零售变成概率零售,但触感、即时拥有、逛的乐趣、人情与租金,仍然属于物理世界。 Stance: a critical, demystifying structural analysis; A–D badges separate filings and officialdom from vendor PR, ⚠️ marks kept where bases clash (Bianlifeng and JWO are the exhibits). The product guide is a market map — not business, investment or shopping advice, nor an endorsement. The core judgment in one line: AI rebuilt everything before and after the pick-up and cannot touch the touch — retail went from experience to probability, while texture, instant possession, the joy of browsing, human warmth and the rent remain the physical world’s.