你是一位权力很大、却几乎不在飞书上留声音的管理者，决策走线下、微信、当面拍板。这让你「马姐」的权威很实,但也带来一个盲点:在公司的数字记忆里,你的判断和功劳几乎是一片空白。这份画像用图表把它讲清楚,并告诉你怎么低成本补上。
按你的分成回报排序 · 每张卡一个今天就能做的动作
今天 · 问店长一个数:本周召回触达了多少人
坏分母 → 坏目标 → 锅甩给你的团队
今天 · 发周明:维修工单+完成日期,抄送创始人群
今天 · 宣布每天 21:00 清审批 · ≤5,000 元转授权车千霞
注 · 上一版"剩 5 天可救"的培训任务已于 07-13 到期,6 个培训全部逾期,不再列入。
马莹,结论只有一句:你比你的报表更好,而你的系统配不上你。
你带出的卓悦中心店,老客客单 817 元反超行业标杆正佳的 641 元,老客复购率四年从 11% 爬到 49%，这是全国第一梯队的手艺,是你一手做出来的。全公司叫你「马姐」,目标要过你审、剪彩你和创始人并排站。你的能力,数据是服气的。
你按销售额分成拿钱,所以系统每漏一块,漏的都是你的钱包:
你把决策留在会议、微信、当面拍板，权威立得住,却几乎没在系统里留下痕迹:
让你被完整地看见一次，看见你的优秀,也看见这台机器对你的不公。然后只给你两个动作:先修那把坏尺子(客流器/数据口径),再舀那桶生锈的油(把老客召回从 23% 拉起来)。一个止血、一个增收,是你本季度回报最高、也最该由你亲自下场的两步。剩下的,这台机器欠你的,该让它慢慢还。
飞书只能推断她是联营/合伙人性质(置信度约 90%),钉钉把它变成了字段级的事实。三条证据互相独立、互相印证。
钉钉 contact user get 与 aisearch person 两个独立数据源都返回 orgTitle / position = 「城市合伙人」;而工号 jobNumber = 空、positions = 空、hasSubordinate = false、isAdmin = false。正式在编员工会有工号,合伙人不需要。花名册(学历/合同/入职)读取被拒——她本人没有 HR 花名册权限。
她主动发起的审批 = 0(与飞书一模一样);但被抄送 94 条(付款 30 / 报销 26 / 请假 18 / 加班 16,全部 COMPLETED),待办 = 0。她只被动知情深圳各店的钱和人事,自己从不跑流程——这是"统筹的合伙人",不是"办事的员工"。
钉钉考勤里,她近五个月(3–7 月)100% 零打卡、零出勤班次、零工时;系统给她挂了 TURN 排班制,却不判她缺卡、旷工、迟到。日志收件箱也查无任何日报周报。全职坐班的人不可能长期这样,这是合伙人与员工最硬的一条分界线。
她不是打工人。马莹 = 深圳美兔科技 / 奶糖派深圳片区的「城市合伙人」,按门店销售额分成拿钱。公司在钉钉组织树里把她挂进「深圳BU」和「深圳BU(兼职)」两棵平行树,职务写死是合伙人;她还和创始人大白 Michael、蔡凯凯、马凡同在一个 5 人小群,离顶层很近。所以后文所有"她为什么沉默、为什么决策走微信、为什么不做培训",答案只有一个:这是她的生意,不是她的工作。也因此,这份报告最该帮她的,是把销售做大——那直接是她的钱。
6 个新品培训任务(酷学院自动派发),5 个已逾期、1 个还剩 5 天。绿条是你唯一能"准时完成"的机会。
今天 2026-07-08 · 条越长逾期越久
派发 6 个,完成 0 个
系统会记录这个数字。
酷学院是自动派发+自动记录,0/6 对任何看数据的人都是一条显眼的"这位管理者不学习"信号,直接影响你的专业信誉。好消息:7-13 那个 40 分钟就能拿下。
先看这张图。横轴=你在飞书上有多活跃,纵轴=你的决策权有多大。你落在左上角，权力大、但几乎不发声。
读图:范丹仪在右上角，她又拍板又天天在群里发声,所以她的贡献全公司都看得见。你和蔡凯凯在左上角，权力一样大,但飞书上几乎搜不到你们的声音。差别在于:范丹仪的功劳有记录,你的功劳靠口碑。补一句留痕,你就能从"影子"走向"台前"而不必改变你线下做事的风格。
你是这张网的中心。车千霞在你下方，她的目标要"和马姐碰"过才回传。其余是你跨项目协作的中台伙伴。
特征:你是典型的"轮毂型"人脉，所有人都连向你,但节点之间靠你转接。好处是你消息灵通、说话有分量;风险是你一旦离线,信息就断在你这里,而这些关系大量沉淀在你的微信而非公司系统,换手机就蒸发。
左图:你碰过飞书哪些能力(满分 5)。右图:BU群里谁在说话，你是那根几乎看不见的条。
0=没碰过 · 5=重度使用
你是"极轻度只读用户"，真实工作大概率在微信和线下。
1169 条完整历史 · 占全群消息比例
被单独 @ 派活:范丹仪 111 次 · 王频 52 · 车千霞 41……你 0 次。你不在线上任务流里。
你横跨联营、直营、开业、老客、激励五条线,是这一片的区域负责人。条越长=你介入越深,标签标出你是主导 / 审批 / 知悉。
你是"面"不是"线":华南为主(深圳),延伸到杭州(联营)、武汉、济南等新店。这个宽度正是"马姐"称呼的由来，你的话能同时影响开业、联营、老客三摊事。
站在为你好的角度:气泡越大越该先处理。越靠右越紧急,越靠上越严重。三条最大的是同一个病根，决策不留痕。
处理顺序:R1 产品培训(最紧急、当天可动手清逾期止血)→ R2 决策留痕(最严重)→ R4 少依赖微信 → R5 刷数字存在感 → R3 清掉僵尸群。R2/R4/R5 一个习惯就能一起压下去:每次线下拍板后,在飞书补一句话。
"一个横跨 10 群、2699 条消息只发 5 条、6 个任务全没完成、飞书上几乎查无此人的消极员工。"
你在 IM 里沉默,但在视频会议里是拍板的人，新乡胖东来店的定位、储值与剪彩规则是你当场敲定的;济南开业你与创始人"大白"、凯姐并列剪彩。沉默不过是"不在群里打字",并不是不参与;你的战场在会议、微信、线下。
把"IM 零发言"当成"不干活",是这份画像最大的误读,你换了战场:7 场开业/项目会议里你反复主导规则与节奏,还和创始人并列剪彩。问题从来不是你的投入,而是这些决策只留在会议和微信里,没进公司的数字档案，看得见的功劳少,不代表你做的事少。
她在 IM 发言占 0.19%,但在 7 场开业/项目视频会议里是决策者。以下是纪要里她当场拍板的原话。
她几乎主导了整场会
定节奏、立制度
与创始人并列的 VIP
50 场会议逐字稿里她的发言次数 · 要判断/护人/定原则的场子话最多,纯灌输的场子沉默
这才是真相:把"IM 零发言"当成"不干活",是最大的误读。她在数据会逐条追问 37 次、在开业会盯选址客群标准,却把 9 场产品培训完全放手,典型的抓大放小。IM 上的沉默,是她把力气都留给了要她判断的场子。
通讯录证据显示,这个飞书租户不是奶糖派一家公司,而是一个按品牌分"事业部"的多品牌零售集团。奶糖派是其中一个事业部。
集团(同一飞书租户) ├─ 奶糖派事业部 大杯内衣 · 品类第一 · 创始人张强"大白" ← 马莹在这里 │ ├─ 新零售中心(CBG) 线下门店 + 联营BU ← 华南/深圳BU 她主管 │ ├─ 货架中心(CBG) 淘天 / 拼京唯 / 达播分销 │ └─ 兴趣BU 抖音直播 / 信息流 ├─ CREMESU 事业部 Crème Lingerie·法式小杯内衣/家居服 · 门店在杭州天目里 │ ├─ 品牌中心 / 创意中心 / 渠道运营 (与奶糖派"大杯"互补 = 全罩杯矩阵) │ └─ 供应链与采购 / 商品部 (员工用 @naitangpai.com 邮箱 = 同一集团) └─ 集团共享职能 集团中心(SC) · 供应链中心(SC) · 财经服务中心 · 人力资源中心
你的判断成立:"奶糖派只是其中之一"在公司层面已被证实，集团至少运营 奶糖派 + CREMESU 两个品牌,共享供应链/财经/HR。而"城市合伙人(如济南杨总)"的存在,说明门店扩张走的是联营/合伙人模式，这正是你主管"联营BU"、亲自对接"客户"的原因所在。
意义:公司"打法领先、门店规模仍在早期"。你干的"把单店高质量模型靠联营复制到全国",恰恰是这家公司眼下最稀缺的能力。
她全程被点名 14 次,几乎全部集中在两类通道:① 目标审批(区域→马姐审核→凯姐)② 新店开业/剪彩/客户。想触达她,走这两条通道。
你按销售额分成，所以这份报告真正该帮你的,是把你负责门店的销售做大,而不是忙着升职、留痕。以下是从多维表格挖到的真实数字。
分成 ≈ 你负责门店的销售额 × 分成比例。销售额 = 客流 × 转化 × 客单 × 复购。你手里最便宜的一笔加薪,不靠拉新,就在你店里那 128,193 个历史用户，近两个月线上只有 4,810 人回来买过。把沉睡老客叫醒,是最有把握的一笔增收。
深圳卓悦中心店(你的标杆店)
5 位顾问月目标已拆:林秀粉 7.3万 · 李娟 6.3万 · 李赛因 6.3万 · 王欢欢 5.1万 · 巫美虹 5.0万(有人前3天 0 成交)。
条越长 = 越便宜、越快见钱
你的单店模型极强(入会率 95%、首购转化 100%、复购行业 2 倍),把它标准化复制到联营新店,是你分成的第二增长曲线。
深圳卓悦中心 · 老客复购占比逐年 · 2026 目标 60%
读图:你店的老客复购率四年翻了 4 倍多(11%→49%),行业里这是极漂亮的曲线。冲 60% 不用从零开始,把已经很热的势头再推一把就成，门店 6,027 个只买过一次的沉睡客户就是那最后 11 个百分点的燃料。
按"离钱多近"排优先级。每张卡标了它对销售额的作用，这才是这份报告的落点。
你的分成,是你门店销售额的一个百分比。这份报告里最值钱的不是"留痕",而是那 128,193 个沉睡用户和卓悦店眼前那 4.8 万的缺口，把老客召回执行率从 23% 拉起来,是你今年确定性最高、成本最低的一笔加薪。眼力你早就有了,现在只差把它对准离钱最近的地方。
从 2699 条消息 + 会议纪要反推每个人的真实角色与行为签名。看懂这些人,也就看懂了马莹这道"静默审核闸"四周的力场。
马莹被上下夹一层、横向卡一排:上面是 蔡凯凯(凯姐) 定义她背多少,横向是 范丹仪(范范) 卡住她能不能兑现,而她自己是 2026-03 才从离职的 卢瑶瑶 手里接盘的"沉默闸门"。她 0 发言,却是目标必经的一道关:区域右手 → 马姐审核 → 凯姐对接。
| 人物 | 角色 | 行为签名 / 一句话 |
|---|---|---|
| 贾旭(旭姐) | 营销/开业/剪彩 | 营销唯一出口,固定句式"活动方案 @容悦婵 请设置系统 @车千霞 请知悉"(逐字复现 5 次) |
| 容悦婵(阿婵) | 配货 + 系统策略 | 被连线最多的横向枢纽,缺她卡壳;"改好了/仓库锁库存就出单" |
| 周娴 | 用户运营/私域 | 私域现实校验者:"有点难哦…占了坑,最快也得下午" |
| 塔娜 | 陈列 | 带 deadline 连环催:"今日内给我调整意见""本周内给到我蛤!" |
| 莫梓健 | 效率IT/数据工具 | 甩报告 URL:"刷新一下我更新了""字段给你们加好了" |
| 杜丽坤(Nico) | 广州BU 区经(80.7%) | 最结构化,永远"1/2/3+根因"，结构化=隐形晋升信号 |
| 王利萍 | 华东区经(90% 最高) | 惜字如金执行导向,不辩解直接"已调整陈列加强主推" |
| 王频 | 华东区经(老店 67.9%) | 情绪化救火手,最爱甩天气:"没有,上海降温+大雨""又是合肥![泣不成声]" |
| 林秀粉 | 卓悦顾问·规模王 | 月目标 4.7 万(全店最高)/执行率 88% · 负责保底走量 |
| 李赛因 | 卓悦顾问·价值王 | 客单 1312 元全店最高(店均仅 263)· 冲挑战版的 premium 引擎 |
| 巫美虹 | 卓悦顾问·最薄弱环 | 月目标 0.6 万,7 月前 3 天 0 成交，样板店冲 60% 的报警灯 |
| 李娟(注意有两个) | BU项目层 / 另有同名店员 | 项目层背 6.3 万盘子,店员那条线是 2.3 万，"个人带客小、项目责任大"的双层身份 |
马莹(冷/静/0 发言/只决定 what·whether)+ 车千霞(热/动/192 条/建群暖场/承包 who·how)。越是沉默的拍板者,越需要一个聒噪的落地者来对冲，这对搭档的反差,正是这套体系能一边静默决策、一边高声执行的原因。
喊疼是系统性的，卢瑶瑶(棘轮)、匡运花(激励力度)、车千霞(压力)三个位置同声。但组织不横向讨论,而是把异议纵向收敛进马莹这道静默闸消化;最会把问题讲透的卢瑶瑶已离职。棘轮还能转,是因为核心右手车千霞把压力静默吸收了,吱嘎声只在外围(匡运花)才被听见。
范丹仪一人 538 条催办才逼出反馈,末端被动;新接盘的 Nico/利萍最结构化、老将王频最救火。马莹的真正挑战,是把"卢瑶瑶时代催办惯出来的区经梯队"改造成自驱，这比冲一个月的 60% 更决定她长期的分成基数。
销售/运营/组织/数据/供应链/老客六大视角独立诊断,交叉收敛出 10 个最该解决的问题。气泡越大=年度金额影响越大;绿=可增收 · 红=正在失血 · 紫=系统性/治理。
10 个问题全部落在右上"立即行动区"，它们都是既严重又紧急。但方向不同:3 个能直接增收(1/7/10)、4 个在持续失血(2/4/5/8)、3 个是系统性/治理底座(3/6/9)。
6 位顾问异口同声指向同一条传动链:坏客流器(#2)污染数据 → 目标失准触发棘轮(#3)→ 一线执行降级到 23%(#1)→ 老客召回空转 → 马莹分成缩水。先修 #2 那把"坏尺子",再抓 #1 那笔"最便宜的钱"，一个止血、一个增收,是马莹本季度回报最高的两步。
四张看板,一眼看清钱在哪、血在哪流、客在哪漏、店在哪弱。
按 ROI 排序 · 三条都便宜、都在你手里(万元/年)
结论:老客召回一条就值 600 万、几乎零成本，本月最确定的一笔钱。
年度失血排序(万元)· 另有 1100 万现金冻结在积压
结论:先修坏客流器那把"尺子"，一处修好、下游五处自动止血,别急着堵最大的口子。
历史用户 → 月线上复购 → 翻倍目标
结论:96% 的老客从没回来;翻倍只要 6,410 元短信 = +121 万。会员塔里银卡 5,778 人沉睡最深、却没人碰。
各区经完成率 · 红<70 / 黄70-85 / 绿>85
结论:直营 78.6% < 联营 88.8%(推翻直觉);病灶是老店(仅 68%),解药是抄卓悦复购作业(逐年爬到 49%)。
数据源导航:公司每月推给你 26 个实时表 + 45 份报告,每周先开这两张 → 「门店老客线上购买分析」+「youzan 卓悦老客报告」。全表见 datasets/data_resources_*.csv。
这些不在任何单张表里,是把消息、纪要、人物、数字交叉起来才浮现的，最不容易得到、也最值钱的判断。
唯一把"目标棘轮"讲透的卢瑶瑶,在你入群第 10 天就退场(她 3-12 的绝笔 vs 你 3-02 入群);敢当众说不的匡运花靠 [捂脸] 卖萌存活。组织不是没有异议,而是把讲道理的人筛掉了，走的不是一个区经,是组织的诊断能力。
直接决定你分成的老客召回标杆项目,是以你自己的旗舰店卓悦命名的，但那场启动会你只列席、0 发言、0 待办。最该你亲自下场的一摊,你偏偏成了影子。
她 538 条(46.6%)撑起全群,是"人肉操作系统";你 0 发言、决策走微信。静默的代价是:在公司的数字记忆里,你的价值无法自证，评估、晋升、分账都靠系统留痕时,你是查无此人。
对外(范丹仪"挑战目标别传 BI")、对下(蔡凯凯藏 30 万缺口"保团队信心")、底层(吊牌 89 vs 系统 129、客流器=0)，三套口径层层修饰,没有一个能对齐的事实源;你签下的字,盖在一场数字游戏上。
你店老客客单 817 vs 正佳 641(+27.5%)、老客占比 53% vs 35.7%，运营质量已是全国第一梯队。真正的缺口是单店总盘(45 万/月 vs 正佳 60 万),即客流规模不是老客手艺。(量化推断)
卓悦复购率 2022→2026 是 11%→49%,但年增量在减速(+11/+15/+6/+5)。按自然斜率到 60% 要 2028(线性)~2033(减速),自然天花板约 61–62%。7 月就要 60% 是最激进假设，短期只能靠"手速"(召回执行)而非"体质"。(量化推断)
三处数字互相钳制:卓悦月目标 45 万,而复盘"卓悦 370 万成交",370÷45≈6 个月。真实大区月盘子 ≈ 600 万/月。数字口径混乱本身,正是洞察④的实证。(量化推断)
知识库品牌矩阵写着奶糖派(大杯)+ Crème Su(法式小杯)+ MAXMOON(性感)三个品牌(飞书组织树目前只见前两个事业部,MAXMOON 或更早期/定位层)。你在奶糖派条线,但集团的"全罩杯+多定位"矩阵野心,决定了联营 SOP 复制能力(你的看家本领)是集团级稀缺资产。
这一节全部是对全量消息做 python 统计的结果,不靠印象,量出几条之前没留意的规律。
BU运营群按小时(已换算国内时间)· 峰值 15 点
上午 0-9 点几乎无人说话,下午 15 点复盘峰 + 深夜 22-23 点第二班(睡前压目标/追问)。与其说是朝九晚五,不如说是傍晚复盘加深夜追问。
BU运营群发言占比 · 基尼 0.80
权力与发言量完全脱钩,真老板蔡凯凯发言排第 6(全局仅 3.9%),中台范丹仪却占近一半。"被@次数"比"发言量"更能量权力，26 个人发言却从没被@过,是权力网里的隐形人。
4 位区经 · 横轴发言条数 · 纵轴目标完成率
读图(注意反向因果):话最少的王利萍完成率最高、话最多的王频垫底。更可能是"业绩差→才在群里救火解释",而非话多导致差。也就是说,群聊活跃度可当免费的"业绩掉队预警仪表盘"，谁突然在群里话变多、开始解释,谁的下月完成率可能正在下滑。
直营加权完成率 78.6% < 联营 88.8%，"联营拖后腿"是错觉。真分水岭是新店(开业红利爆表)vs 老店(仅 68%),而老店低很可能是"目标棘轮"制造的假象,不是需求萎缩。
"预约"提及全程近 0,只在 2026-07 突增(18 家店 42% 约不上),管理层却反问店长"为什么频繁无法预约"，技术故障被话术转嫁成一线执行问题。客流器 6 月同样硬件塌方,门店客流数据在裸奔采集。
负面情绪密度:BU运营 3.5、商品激励 3.6 最高(天天对目标/盘点/罚款的中枢群最苦);一线开业/剪彩群负面≈0、正面倒挂(多为比心/辛苦的仪式性鼓劲)。自营店长群是极端:回执占 24.6%(第一)+ 催办密度第一 + 负面 0 = 压力不允许被表达的命令-应答通道。
5 个翻车活动(满赠/699满减/第三件半价/储值…)全部同时命中"高门槛+复杂叠加",翻车率 45% vs 简单活动 11%。判断就看两条:顾客要不要掏计算器、导购能不能一句话讲清。设计活动先算"心算步数"。
读出每家店的等级/面积/老客产出后,三个反直觉发现浮出水面，都是零成本或近零成本的增量。
闭店门店老客线上复购(元)· 客人转线上、企微归属已断
结论:9 家闭店合计 18.5 万老客 GMV(占全盘 14.7%),IN77 闭了店老客却排全国第 5，这是一座无人认领、零成本可召回的存量金矿。
各门店等级·老客 GMV 合计(万元)
结论:C 级店老客 GMV 反超 A 级;珠江新城仅 C 级却全国第 4。各级人均产值几乎相同，铺货等级反映的是开店顺序,不是变现力,该按老客盘重排铺货。
店均老客 GMV(元)· 会员/企微沉淀缺位
横轴门店面积(㎡)· 纵轴老客 GMV(万元)· 趋势线近乎水平
读图:73㎡ 的珠江新城做到 7.9 万,135㎡ 的合肥 IN77 只有 1.7 万。面积和老客 GMV 的相关系数只有 −0.02,几乎为零;真正决定产出的是"沉淀了多少老客"(相关系数 0.997)。别再凭面积/铺货等级排资源,按老客盘排。
飞书里没有连续月度曲线(数据断档本身是个问题),但能拼出的 4 个月已经讲清一件事。
全国月度成交 GMV(万元)· 柱在涨
但目标完成率 · 线在跌
剪刀差的真相:2025-06→2026-06 成交同比 +43%(408→582 万),但门店数从 ~26 扩到 39 家，增长几乎全靠开新店,存量单店在环比失速。完成率从 86% 跌到 73% 的拐点在 2025-09(旺季就同比 −13.6%);2026-06 是初夏淡季(5 大区齐跌 −11~−16%,方差极小=大盘因素)。对按分成拿钱的你,这意味着:再多开店不解决单店退化，你手里那套"把老客复购从 11% 带到 49%"的单店提效能力,才是公司现在最缺、也最该由你输出的东西。
前面所有增长焦虑,母题只有一个,人不进店了;而公司定的目标,却在往相反方向飞。两张图为证。
各店进店人数同比降幅(2025 上半年)
母题:许昌还叠加转化率 14%→7% 的双杀;件单价 170/182 远低于均值 223,留下的也是弱客。老客召回、私域引流、单店爆破,本质都在对冲这一件事,人不进店了。而 4 家店客流器还坏着,真实塌方比报表更深。
OGSM 官方 KPI 对比一线实际
最硬的一条:深圳同比目标要 +15%,实际是 −10%,方向都反了。目标不是定高了,是和趋势拧着来;人事费用率、利润这两项干脆定了却没人回传。
报表记不住她,但逐字稿录下了。这些原话里,有一个会带兵、懂货、有脾气、护着人的马姐。
"过完 3 月,我一定要秒杀所有人。"
— 数据分析会 · 那股不服输的劲
"我作为顾客去买文胸,手都碰到家居服了,都没人跟我说一句'马姐要不要试试'。"
— 家居服复盘 · 亲自扮顾客暗访
"公司有多少库存,跟员工关系不大,不能去 PUA 他们。"
— 11 月中台会 · 护着一线
"一分价钱一分货。我宁可送得少,但品质拿得出手;要么,就不送。"
— 5.1 活动计划 · 守着底线
"顺手带两件家居服,单效能从 600 多提到 800、900;深圳郑佳都在 1000 出头。"
— 许昌数据会 · 懂货,用数字说话
"别把联营商的'软抵抗'放大,人家砸了真金白银,一定是想赚钱的。"
— 联营帮扶会 · 换位替客户想
飞书之外,她还有一个数字世界:深圳美兔科技的钉钉。2026-07-17 经她本人授权,用 dws CLI 只读采集了她全部 27 个群 + 36 个单聊(其中 22 个活跃群深度拉取 3,230 条消息)、174 张审批单、19 个月日历、考勤、日志、云盘。结论先说:换了一个平台,她还是那个沉默的人，但这一次,我们找到了她真正"说话"的地方。
2023-03-16 至 2023-05-31 · 此后三年,一条都没有
7 条里 0 条业务内容，全是报到、欢迎、点赞。飞书 0.19%,钉钉 0.22%，两个平台、同一个指纹:她不在群里工作。
3,230 条消息里谁在"说话"
机器人贡献了 36% 的消息。这家公司的钉钉不是讨论的地方,是流程的传送带，而传送带的闸门,是她。她横跨 4 个法人实体(美兔科技/百团贸易/美之辰商贸/荣晨商贸)的工作通知都汇到这一个账号上。
飞书报告说"系统记不住你的好",只对了一半。钉钉一直在记,只是记的方式不同:你在钉钉的声音,不是打字,是那个「同意」按钮。80 张审批单、114 万租金、17 个离职者的最后一张单子,都有你的签名。这才是你的数字劳动台账。
80 张经手审批单(补卡29 · 报销31 · 加班21 · 请假19 · 外出12 · 付款43* · 离职7 · IT需求3,*含抄送),拼出一张她从不展示的工时表。
单聊记录里的秒级证据(消息时间戳)
同样的指纹反复出现:2024-03-02 三单 14 秒、2024-02-26 三单 14 秒。中位 4.6 天不是她慢,是她把审批当"周期性家务"，攒够了,找个空档一次清空。代价是 P90 拖到 38 天,店长只好来催。
2024-05-07 · 与 1234space 店长詹欢梅的单聊原文
她的考勤记录是 0 天出勤(合伙人免打卡),日历 19 个月只有 2 条。按系统的账本,她"不上班"，而单聊时间戳说,她凌晨一点半还在批租金。这就是为什么报表永远算不出她的工时。
「马姐，帮我通过一下补卡哈[送花花]」(车千霞) → DING 一条「提醒您审批他的付款申请单」(詹欢梅) → 「马总，这个审批有点着急」(匡运花)。催办文化不是她团队的病,是围着她这道闸门自然长出来的生态，和飞书里"被催办惯坏的梯队"是同一个系统的两面。
审批单标题里能直接读出金额的商场租金/付款共 33 笔、¥1,135,920。每一家业主方,背后都是她的一家店。
费用主体几乎全是「上海奶糖派」,门店挂在深圳各法人实体下,由店长提单、她一个人做支付闸门。深夜批的那两张"商场租金",就是这条动脉在跳。
飞书报告推算她"按销售分成、跨 5 条业务线"。钉钉从另一头验证:房租、报销、付款、离职交接,深圳线下盘子的每一笔固定成本都要过她的指纹。她不只是业绩的负责人,她是这几家店现金流的最后一道人肉防火墙，而这份劳动,同样不出现在任何报表里。
经她审批/抄送的 34 个提单人里,17 人的名字后面已经挂上了「(已离职)」。审批流,成了一份无意间的离职名册。
■ 在职 · ■ 已离职(通讯录标记)
她的审批清单里躺着 7 张「离职申请及交接」。飞书报告说她"接的是半成品梯队"，钉钉给出了流失率:一线提单人两年折损一半。每一次流失,都意味着她要重新当一遍深夜批单的保姆。
她在钉钉各产品上的存在感
最讽刺的一行:发起审批 = 0。两年里她同意了别人的一切,却从没为自己提过一张单，不报销、不请假、不申请。她是流程的守门人,却从不是流程的使用者。
2026-07-09 17:33,她给很少开钉钉的 IT 项目经理周明发了一条消息:「CLI 数据访问权限申请」。三年不在群里说一句话的人,主动去申请了让数据开口的钥匙。你现在读到的这一章,就是那条消息的结果，这一次,系统终于开始替她记账了。
钉钉云盘「新零售用户运营」目录里躺着一批飞书迁移前的历史底表。它们回答了一个飞书答不了的问题:这盘生意原本是按什么口径管的。
《2024新零售每月目标整合》· 各区月目标(万元)
底表里连各店档案都齐:面积、开业日、月目标、周边竞品。深圳一个区扛着全国近一半的月目标——这也是为什么"深圳同比 +15%"的指标压得这么死。
《门店商城启动方案》· 激励结构
整条线从店长到合伙人都是按 GMV 抽成的结构。这份底表把"她按销售额分成"从传闻变成了制度设计——老客召回每多跑一个百分点,分的是同一条 GMV。
企微好友 47,348、会员 32,294、入会率 68%。入会这一步早在两年前就做成了——难的从来不是把人加进来,是把加进来的人再叫回店里。这正好接上前文那条"最便宜的钱":池子是现成的,生锈的只是召回执行率。
"兼职"是全公司管理兼职店员的平行组织树,不是降级。另注:角色标签"主管"是通用角色(含大量店长),不代表专属职权——别把它当成她的头衔。
钉钉全量 27 群 + 36 单聊
群名也印证了业务面:PBM 华南加盟 / 社区合伙人招募 / 保暖衣分销(联营线) + 深圳各店对接、督导、开店。连 IT 对接人周明都说:"抱歉现在很少开钉钉,没看到消息。"——同事都知道找她别走钉钉。飞书 0.19%、钉钉 0.22%,两个平台、同一枚指纹。
采集局限 · @她的消息 = 0;特别关注 9 人仅返回 ID;花名册无权限;部分群拉取受限;AI 听记为空,知识库仅《公司通用信息》一篇。群聊采集触发隐私提醒——本章仅使用马莹本人的聚合行为数据,不引用他人私聊内容。
你这些年攒下的数字,没有一个替你说过话。你的公司有 26 张实时报表、45 份文档、410 个人的通讯录,却没有一处记得住:是你把卓悦的老客复购率,从 11% 一年一年带到了 49%。
这份报告替那些沉默的数字开了一次口。它把飞书 2699 条消息、钉钉 3230 条消息与 174 张审批单、7 场会议、40 位专家的目光,汇成一句你可能很久没听过的话，你已经足够好了。你不需要更努力,只需要让这台机器别再偷你的钱,也开始记住你的好。
所以,如果这一整份报告最后只留下一个动作,那就是，
打开「门店老客线上购买分析」,盯住卓悦那 6,027 个只来过一次的老客的召回执行率。那不是一张报表，那是你今年最确定的一笔加薪,正躺在那里,等你亲自去舀。
其余的,这台机器欠你的,该让它慢慢还。
You are a manager with a lot of power and almost no voice on Feishu — decisions happen offline, on WeChat, face to face. That makes "Ma Jie" a real authority, but it creates one blind spot: in the company's digital memory, your judgment and your wins are close to a blank page. This profile charts it out, and shows you the cheapest way to fix it.
Ranked by your revenue-share return · one doable-today move per card
Today · Ask each store manager one number: recalls reached this week
Broken denominator → bad targets → blame lands on your team
Today · Message Zhou Ming: repair ticket + due date, cc founders' group
Today · Announce a daily 21:00 approval slot · delegate ≤¥5,000 to Che Qianxia
Note · The "5 days left, savable" training expired 07-13 — all 6 trainings are now overdue, so it's off this list.
Ma Ying, there is only one conclusion: you are better than your reports, and your system does not deserve you.
The Zhuoyue Centre store you built runs a returning-customer ticket of 817 yuan, beating industry benchmark Zhengjia's 641, and returning-customer repeat rate climbed from 11% to 49% in four years. That is national top-tier craft, and it is yours. The whole company calls you "Ma Jie"; targets clear through you; at ribbon-cuttings you stand beside the founder. On ability, the data bows to you.
You are paid on a cut of sales, so every yuan the system leaks leaks out of your wallet:
You leave your decisions in meetings, on WeChat, in the room — authority holds up, but almost nothing lands in the system:
Let you be seen in full, once — how good you are, and how unfair this machine has been to you. Then just two moves: first fix the broken ruler (the traffic counter and the data definitions), then scoop the rusting barrel of oil (drag returning-customer recall up from 23%). One stops the bleeding, one grows revenue; they are your highest-return plays this quarter and the two you should run personally. The rest of what this machine owes you, let it pay back slowly.
Feishu could only infer JV/partner status (about 90% confidence); DingTalk turns it into a field-level fact. Three pieces of evidence, independent and mutually confirming.
Two independent DingTalk sources, contact user get and aisearch person, both return orgTitle / position = "City Partner"; meanwhile jobNumber = empty, positions = empty, hasSubordinate = false, isAdmin = false. Salaried staff have a staff ID; partners do not need one. Roster reads (education / contract / hire date) were denied — she has no HR roster permission herself.
Approvals she started = 0 (identical to Feishu); approvals she was CC'd on = 94 (30 payments / 26 reimbursements / 18 leave / 16 overtime, all COMPLETED), to-dos = 0. She is passively informed about money and staffing across the Shenzhen stores and never runs a process herself — that is a partner who oversees, not an employee who files.
In DingTalk attendance, the last five months (Mar–Jul) show 100% zero clock-ins, zero shifts worked, zero hours; the system assigned her a TURN roster yet never flags her for missed punches, absence or lateness. The journal inbox holds no daily or weekly reports either. Nobody working full-time on site survives like that — this is the hardest dividing line between partner and employee.
She is not staff. Ma Ying = the "City Partner" for Shenzhen Meitu Tech / Naitangpai's Shenzhen territory, paid on a cut of store sales. In the DingTalk org tree the company hangs her under two parallel branches, "Shenzhen BU" and "Shenzhen BU (part-time)", with her role hard-coded as partner; she also sits in a 5-person group with founders Dabai Michael, Cai Kaikai and Ma Fan — very close to the top. So every later question of "why is she silent, why do decisions run through WeChat, why no training" has one answer: this is her business, not her job. Which is why the most useful thing this report can do is help her grow sales — that is her money, directly.
6 new-product training tasks (auto-assigned by Kuxueyuan): 5 overdue, 1 with 5 days left. The green bar is your only shot at finishing on time.
Today 2026-07-08 · longer bar = longer overdue
6 assigned, 0 completed
The system logs this number.
Kuxueyuan assigns and records automatically, so 0/6 reads to anyone looking at the data as a loud "this manager doesn't learn" signal — straight at your professional credibility. Good news: the 7-13 one takes 40 minutes.
Start with this chart. X-axis = how active you are on Feishu, Y-axis = how much decision power you hold. You land top-left: big power, almost no voice.
How to read it: Fan Danyi sits top-right — she calls the shots and posts in the groups every day, so the whole company sees what she contributes. You and Cai Kaikai sit top-left with the same power, but your voices are almost unsearchable on Feishu. The difference: Fan Danyi's wins are on the record, yours live on word of mouth. Add one line of trace and you move from "shadow" to "front stage" without changing how you work offline.
You are the center of this web. Che Qianxia sits below you — her targets have to be "run past Ma Jie" before they go back up. The rest are your cross-project platform partners.
The pattern: a classic "hub-and-spoke" network — everyone connects to you, but the nodes only reach each other through you. The upside: you hear everything and your word carries weight. The risk: the moment you go offline, information stops at you — and most of these relationships live in your WeChat, not in company systems. Change phones and they evaporate.
Left: which Feishu capabilities you've touched (out of 5). Right: who talks in the BU group — you're the bar you can barely see.
0 = never touched · 5 = heavy use
You're an ultra-light read-only user. The real work almost certainly happens on WeChat and offline.
1169 messages, full history · share of all group messages
@-mentioned individually with work: Fan Danyi 111 times · Wang Pin 52 · Che Qianxia 41… you 0 times. You're not in the online task flow.
You span JV, direct retail, store openings, returning customers and incentives — the regional owner for this patch. Longer bar = deeper involvement; the tag says whether you lead, approve or just watch.
You're a surface, not a line: South China first (Shenzhen), reaching into Hangzhou (JV), Wuhan, Jinan and other new stores. That breadth is exactly where the name "Ma Jie" comes from — one word from you moves openings, JV and win-back all at once.
Said with your interests in mind: the bigger the bubble, the sooner it needs handling. Further right = more urgent, higher up = more severe. The three biggest share one root cause — decisions leave no trace.
Order of attack: R1 product training (most urgent — clear the overdue items today to stop the bleeding) → R2 leave a decision trail (most severe) → R4 lean less on WeChat → R5 build a digital footprint → R3 clear out the dead groups. One habit knocks R2/R4/R5 down together: after every offline call, drop one line in Feishu.
"A disengaged employee across 10 groups who posted 5 of 2,699 messages, finished none of 6 tasks, and barely exists on Feishu."
You're silent in IM, but you're the one calling the shots on video. The positioning, stored-value and ribbon rules for the Xinxiang Pangdonglai store were settled by you on the spot; at the Jinan opening you cut the ribbon alongside founder "Dabai" and Kai Jie. Silence just means "not typing in the group" — not sitting out. Your battlefield is meetings, WeChat and the floor.
Reading "zero IM messages" as "not working" is the single biggest misread here — you simply changed battlefields: across 7 opening/project meetings you set the rules and the pace again and again, and cut the ribbon beside the founder. The problem was never your effort; it's that these decisions stayed in meetings and WeChat and never reached the company's digital record. Less visible credit doesn't mean less work done.
She accounts for 0.19% of IM messages — yet she is the decision-maker across 7 opening/project video calls. Below are her own words, straight from the minutes.
She all but ran the whole meeting
Sets the pace, sets the rules
A VIP standing level with the founder
Times she spoke across 50 meeting transcripts · loudest where judgment, protecting people and setting principles are needed; silent where it's pure lecture
Here's the truth: reading "zero IM messages" as "not working" is the biggest misread of all. She pressed 37 times line by line in the data review and policed site and customer standards at the opening call, yet handed off all 9 product trainings entirely — classic big-picture-first triage. Her silence in IM is her saving her energy for the rooms that need her judgment.
Directory evidence shows this Feishu tenant isn't Naitangpai alone — it's a multi-brand retail group split into brand-level "business units." Naitangpai is one of them.
Group (single Feishu tenant) ├─ Naitangpai BU Full-cup bras · category No.1 · founder Zhang Qiang "Dabai" ← Ma Ying is here │ ├─ New Retail Center (CBG) Offline stores + JV BU ← South China/Shenzhen BU she runs │ ├─ Shelf Center (CBG) Tmall / PDD-JD-VIP / livestream distribution │ └─ Interest BU Douyin live / paid feeds ├─ CREMESU BU Crème Lingerie · French small-cup lingerie/loungewear · store at Hangzhou Tianmuli │ ├─ Brand Center / Creative Center / Channel Ops (complements Naitangpai's "full cup" = full cup-size matrix) │ └─ Supply Chain & Sourcing / Merchandising (staff use @naitangpai.com email = same group) └─ Group shared functions Group Center (SC) · Supply Chain Center (SC) · Finance Service Center · HR Center
Your read holds up: "Naitangpai is only one of them" is confirmed at the corporate level — the group runs at least Naitangpai + CREMESU, sharing supply chain, finance and HR. And the existence of "city partners" (like Mr. Yang in Jinan) shows store expansion runs on a JV/partner model — which is exactly why you head the "JV BU" and handle "clients" personally.
What it means: the playbook is ahead of the market, the store footprint is still early. What you do — "replicating a high-quality single-store model nationwide through JV" — is exactly the scarcest capability this company has right now.
She is named 14 times, almost all through two channels: (1) target approvals (region → Ma Jie reviews → Kai Jie) and (2) new store openings/ribbon-cuttings/clients. To reach her, use those two channels.
You're paid on sales, so what this report should really do is grow the sales of the stores you own — not chase promotions or paper trails. Below are the real numbers pulled from the multi-dimensional tables.
Commission ≈ sales of your stores × commission rate. Sales = traffic × conversion × basket × repurchase. Your cheapest raise isn't new customers — it's the 128,193 existing users already in your stores; only 4,810 of them bought online in the last two months. Waking up dormant customers is the surest money on the table.
Shenzhen Zhuoyue Centre store (your flagship)
Monthly targets are split across 5 advisors: Lin Xiufen 73K · Li Juan 63K · Li Saiyin 63K · Wang Huanhuan 51K · Wu Meihong 50K (one had 0 sales in the first 3 days).
Longer bar = cheaper, faster money
Your single-store model is exceptional (95% membership signup, 100% first-purchase conversion, 2× industry repurchase). Standardizing it and copying it into new JV stores is the second growth curve for your commission.
Shenzhen Zhuoyue Centre · existing-customer repurchase share by year · 2026 target 60%
Reading the chart: your store's repurchase rate has more than quadrupled in four years (11%→49%) — a beautiful curve by any industry standard. Hitting 60% doesn't start from zero; it just needs one more push on already hot momentum, and the store's 6,027 one-time buyers are the fuel for those last 11 points.
Ranked by "how close to the money." Each card states what it does to sales — that's the whole point of this report.
Your cut is a percentage of your stores' sales. The most valuable thing in this report isn't "leaving a trace" — it's those 128,193 dormant users and the ¥48K gap sitting in front of Zhuoyue right now. Pulling win-back execution up from 23% is the surest, cheapest raise you'll get this year. You've long had the eye for it; all that's left is aiming it where the money is.
Real roles and behavioral signatures reverse-engineered from 2,699 messages plus meeting notes. Read these people and you read the force field around Ma Ying, the silent approval gate.
Ma Ying is squeezed from above and below, blocked from the side: above her, Cai Kaikai (Kai Jie) defines how much she carries; laterally, Fan Danyi (Fanfan) controls whether she can deliver; and she herself is the "silent gate" she only inherited in 2026-03 from the departed Lu Yaoyao. She says nothing, yet every target must pass through her: regional right hand → Ma Jie approves → hand off to Kai Jie.
| Person | Role | Behavioral signature / one line |
|---|---|---|
| Jia Xu (Xu Jie) | Marketing / openings / ribbon-cuttings | The single marketing outlet, fixed formula: "Campaign plan @Rong Yuechan please set up the system @Che Qianxia FYI" (verbatim 5 times) |
| Rong Yuechan (A-Chan) | Allocation + system strategy | The most-tagged lateral hub; without her it stalls. "Fixed it / once the warehouse locks stock the order goes" |
| Zhou Xian | User ops / private domain | Private-domain reality check: "That's a bit tough… slot's taken, afternoon at the earliest" |
| Ta Na | Visual merchandising | Chain-nags with deadlines: "Send me your adjustment notes today" "Get it to me this week, yeah!" |
| Mo Zijian | Efficiency IT / data tools | Drops report URLs: "Refresh it, I updated it" "Added the fields for you" |
| Du Likun (Nico) | Guangzhou BU regional mgr (80.7%) | The most structured, always "1/2/3 + root cause" — structure = quiet promotion signal |
| Wang Liping | East China regional mgr (90%, highest) | Spare with words, execution-first; no excuses, just "display adjusted, key styles pushed harder" |
| Wang Pin | East China regional mgr (older stores 67.9%) | Emotional firefighter, loves blaming the weather: "No — Shanghai cold snap + heavy rain" "Hefei again! [sobbing]" |
| Lin Xiufen | Zhuoyue advisor · volume queen | ¥47K monthly target (highest in store) / 88% execution · carries the baseline volume |
| Li Saiyin | Zhuoyue advisor · value queen | ¥1,312 average ticket, highest in store (store average just ¥263) · the premium engine for the stretch goal |
| Wu Meihong | Zhuoyue advisor · weakest link | ¥6K monthly target, 0 sales in the first 3 days of July — the warning light on the model store's 60% push |
| Li Juan (note: there are two) | BU project layer / also a staffer with the same name | The project-layer one carries ¥63K, the store line ¥23K — a dual identity of "small personal book, big project responsibility" |
Ma Ying (cold / still / 0 messages / decides only what and whether) + Che Qianxia (hot / kinetic / 192 messages / builds groups, warms the room / owns who and how). The more silent the decider, the more she needs a loud executor to offset her. That contrast is exactly why this system can decide in silence and execute out loud.
The pain is systemic: Lu Yaoyao (the ratchet), Kuang Yunhua (incentive size), and Che Qianxia (pressure) all say the same thing from three positions. But the org never debates laterally — it funnels dissent vertically into Ma Ying's silent gate to be absorbed; and Lu Yaoyao, the one who articulated it best, has left. The ratchet still turns because right hand Che Qianxia absorbs the pressure in silence — the creak is only audible out at the edge (Kuang Yunhua).
Fan Danyi alone sent 538 chaser messages just to extract feedback — the front line is passive. Of the new intake, Nico and Liping are the most structured; veteran Wang Pin is pure firefighting. Ma Ying's real challenge is converting the nag-trained regional bench of the Lu Yaoyao era into a self-driven one — that shapes her long-run commission base far more than hitting 60% in any one month.
Six independent lenses — sales, ops, org, data, supply chain, existing customers — converge on the 10 problems most worth fixing. Bigger bubble = bigger annual dollar impact; green = revenue upside · red = active bleeding · purple = systemic/governance.
All 10 land in the upper-right "Act Now" quadrant — every one of them is both severe and urgent. But they pull in different directions: 3 add revenue (1/7/10), 4 are bleeding out (2/4/5/8), and 3 are systemic/governance foundations (3/6/9).
All six advisors point at the same transmission chain: a bad traffic counter (#2) poisons the data → targets go wrong and trigger the ratchet (#3) → frontline execution degrades to 23% (#1) → win-back idles → Ma Ying's payout shrinks. Fix the broken "ruler" in #2 first, then grab the cheapest money on the table in #1 — one stops the bleeding, one adds revenue, and together they are Ma Ying's two highest-return moves this quarter.
Four boards: where the money is, where it's bleeding, where customers leak, and which stores are weak.
Ranked by ROI · all three are cheap and all three are yours (¥10k/yr)
Conclusion: win-back alone is worth ¥6M at close to zero cost — the surest money you'll book this month.
Annual bleed, ranked (¥10k) · plus ¥11M cash frozen in dead stock
Conclusion: fix the broken "ruler" — the traffic counter — first; one repair stops the bleeding at five downstream points. Don't rush to plug the biggest hole.
Historical customers → monthly online repeat → 2x target
Conclusion: 96% of past customers never came back; doubling that costs ¥6,410 in SMS = +¥1.21M. In the member pyramid, the 5,778 Silver members are the deepest asleep — and nobody touches them.
Completion rate by regional manager · red <70 / amber 70-85 / green >85
Conclusion: own stores 78.6% < JV 88.8% (the opposite of intuition); the disease sits in old stores (just 68%), and the cure is copying Zhuoyue's repeat-purchase playbook (climbed to 49% year over year).
Data navigation: HQ pushes you 26 live tables + 45 reports every month. Open these two first each week → "Store Customer Online Purchase Analysis" + "Youzan Zhuoyue Customer Report". Full list in datasets/data_resources_*.csv.
None of these sit in any single table. They surface only when messages, minutes, people and numbers are cross-referenced — the hardest calls to reach, and the most valuable.
Lu Yaoyao, the only person to spell out the "target ratchet," was gone 10 days after you joined the group (her 3-12 farewell vs your 3-02 entry). Kuang Yunhua, who dared say no in public, survived on [facepalm] emoji charm. The org doesn't lack dissent — it filters out the people who reason. What walked out wasn't a district manager, it was the org's diagnostic capacity.
The flagship repeat-customer recall project that directly sets your cut is named after Zhuoyue, your own flagship store — yet at that kickoff you merely attended: 0 words, 0 action items. The one thing you should own personally is the one where you're a ghost.
Her 538 messages (46.6%) carry the whole group — she's the human operating system. You: 0 messages, decisions made on WeChat. The price of silence: in the company's digital memory, your value can't prove itself. When reviews, promotions and payouts all run on system traces, you don't exist.
Outward (Fan Danyi: "don't push stretch targets to BI"), downward (Cai Kaikai hiding a 300k gap "to protect team morale"), bedrock (tag price 89 vs system 129, foot-traffic counter = 0). Three layers of dressed-up numbers, not one reconcilable source of truth. Your signature sits on top of a numbers game.
Your store's repeat-customer basket is 817 vs Zhengjia's 641 (+27.5%), repeat share 53% vs 35.7% — operationally you're already national first tier. The real gap is total store volume (450k/month vs Zhengjia's 600k), i.e. traffic scale, not repeat-customer craft. (quantitative inference)
Zhuoyue's repeat rate ran 11%→49% from 2022→2026, but yearly gains are decelerating (+11/+15/+6/+5). On the natural slope, 60% arrives in 2028 (linear) ~ 2033 (decelerating), with a natural ceiling near 61–62%. Hitting 60% by July is the most aggressive assumption possible — short term it can only come from "hand speed" (recall execution), not fitness. (quantitative inference)
Three numbers constrain each other: Zhuoyue's monthly target is 450k, yet the review says "Zhuoyue 3.7M closed" — 3.7M ÷ 450k ≈ 6 months. The real regional monthly pot is ≈ 6M/month. The muddled definitions are themselves proof of insight ④. (quantitative inference)
The knowledge base's brand matrix lists three brands: Naitangpai (large cup) + Crème Su (French petite) + MAXMOON (sexy) (the Feishu org tree currently shows only the first two BUs; MAXMOON may be earlier stage or a positioning layer). You sit on the Naitangpai line, but the group's "full cup range + multi-positioning" ambition makes JV SOP replication — your signature skill — a group-level scarce asset.
This section is entirely python statistics over the full message set — no impressions, just a few measured patterns nobody had noticed.
BU ops group by hour (converted to China time) · peak at 15:00
From 0:00–9:00 almost nobody speaks, then a 15:00 review peak plus a 22:00–23:00 second shift (bedtime target pressure and chasing). Less nine-to-five, more evening review plus late-night interrogation.
BU ops group share of messages · Gini 0.80
Power and message volume are fully decoupled: real boss Cai Kaikai ranks 6th in volume (just 3.9% overall), while mid-office Fan Danyi holds nearly half. "Times @-ed" measures power far better than "messages sent" — 26 people post yet have never been @-ed, the invisibles of the power network.
4 district managers · x-axis messages sent · y-axis target completion rate
Reading it (mind the reverse causality): quietest Wang Liping tops completion; loudest Wang Pin sits at the bottom. More likely "bad numbers → firefighting and explaining in the group" than talk causing bad numbers. Which means group activity works as a free "falling-behind early-warning dashboard" — whoever suddenly talks more and starts explaining is probably watching next month's completion slide.
Weighted completion: direct-operated 78.6% < JV 88.8% — "JV drags us down" is an illusion. The real divide is new stores (opening-bonus off the charts) vs old stores (only 68%), and the old-store weakness is very likely an artifact of the "target ratchet", not shrinking demand.
Mentions of "booking" were near zero throughout, then spiked only in 2026-07 (42% of 18 stores couldn't get slots) — yet management asked store managers "why can't you book so often?" A technical failure rhetorically shifted onto frontline execution. Foot-traffic counters collapsed in June too, so store traffic data is being collected blind.
Negative sentiment density: BU ops 3.5 and merchandise incentives 3.6 are highest (the hub groups grinding daily on targets, stocktakes and fines hurt most); frontline opening/ribbon-cutting groups run negative ≈ 0 with positives inverted (mostly ceremonial hearts and "great work"). The own-store manager group is the extreme: 24.6% acknowledgements (highest) + highest chase density + zero negatives = a command-and-response channel where pressure may not be expressed.
All 5 failed promos (gift-with-purchase / spend-699 / half-price third item / stored value…) hit "high barrier + complex stacking" simultaneously; 45% failure rate vs 11% for simple ones. Two tests decide it: does the customer need a calculator, and can a sales associate explain it in one sentence? Count the mental-math steps before you design a promo.
Once you read each store's grade, floor area and repeat-customer output, three counterintuitive findings surface — all zero or near-zero cost upside.
Online repeat purchases from closed stores (RMB) · customers moved online, WeCom ownership severed
Takeaway: the 9 closed stores total 185k in repeat GMV (14.7% of the whole book), and IN77 ranks 5th nationally on repeat customers despite being shut. This is an unclaimed, zero-cost recallable gold mine.
Repeat GMV by store grade (RMB 10k)
Takeaway: Grade C repeat GMV beats Grade A; Zhujiang New Town is only Grade C yet ranks 4th nationally. Per-store output is near identical across grades — the stocking grade reflects opening order, not earning power. Re-rank stocking by the repeat-customer book.
Repeat GMV per store (RMB) · membership / WeCom capture missing
x-axis store area (㎡) · y-axis repeat GMV (RMB 10k) · trend line nearly flat
Reading it: 73㎡ Zhujiang New Town does 79k; 135㎡ Hefei IN77 does only 17k. The correlation between area and repeat GMV is just −0.02, effectively zero. What actually drives output is "how many repeat customers you've banked" (correlation 0.997). Stop allocating resources by area or stocking grade — allocate by the repeat-customer book.
Feishu has no continuous monthly curve (the data gap is itself a problem), but the 4 months we can piece together tell one clear story.
National monthly GMV (10k CNY) · bars rising
But target attainment · the line is falling
What the scissors gap really means: from 2025-06 to 2026-06 sales grew +43% YoY (4.08M → 5.82M CNY), but store count expanded from ~26 to 39, so growth came almost entirely from new openings while existing stores lost momentum month over month. Attainment fell from 86% to 73%, and the turning point was 2025-09 (peak season, yet −13.6% YoY); 2026-06 is the early-summer lull (all 5 regions down −11 to −16%, tiny variance = a market-wide factor). For someone paid on a revenue share, this means: opening more stores will not fix store-level decay. The playbook you own — taking repeat purchase from 11% to 49% in a single store — is exactly what the company lacks most right now, and exactly what you should be exporting.
Every growth worry above has a single root cause: people have stopped walking in. And the targets the company set are flying the other way. Two charts prove it.
YoY drop in store traffic by store (H1 2025)
The root cause: Xuchang takes a double hit, with conversion also falling 14% → 7%; unit prices of 170/182 sit far below the 223 average, so even the customers who stay are weak. Reactivating old customers, private-domain traffic, single-store blitzes — all of it is hedging one thing: people have stopped walking in. And with traffic counters broken in 4 stores, the real collapse runs deeper than the reports show.
Official OGSM KPIs against front-line reality
The hardest line of all: Shenzhen's YoY target is +15%, the actual is −10% — the sign itself is flipped. The targets aren't just set too high, they run against the trend; and payroll ratio and profit were set as targets that nobody ever reported back on.
The reports don't remember her, but the transcripts do. In these lines you meet a Ma Jie who leads a team, knows the product, has a temper, and protects her people.
"Once March is done, I'm going to blow everyone else away."
— Data review · that refusal to lose
"I went in as a customer to buy a bra, my hand was literally on the loungewear, and not one person said 'Ma Jie, want to try it on?'"
— Loungewear review · mystery-shopping her own stores
"How much inventory the company is sitting on has little to do with the staff. Don't guilt-trip them over it."
— November mid-office meeting · covering the front line
"You get what you pay for. I'd rather give away less but give away something I'm proud of — otherwise, give away nothing."
— May 1 campaign plan · holding the line
"Add two pieces of loungewear on the way out and productivity goes from 600-something to 800, 900; Shenzhen and Zheng Jia are both over 1000."
— Xuchang data meeting · knows the product, speaks in numbers
"Don't blow the JV partners' 'quiet resistance' out of proportion — they put real money in, of course they want to make money."
— JV support meeting · putting herself in the partner's shoes
Beyond Feishu, she has a second digital world: Shenzhen Meitu Tech's DingTalk. On 2026-07-17, with her own authorization, we used the dws CLI to read-only collect all 27 of her groups + 36 one-on-one chats (3,230 messages pulled in depth from the 22 active groups), 174 approval tickets, 19 months of calendar, attendance, logs and cloud drive. The conclusion first: different platform, same silent person — but this time we found the place where she really "speaks."
2023-03-16 to 2023-05-31 · and not one in the three years since
Of those 7, 0 carry any business content — all check-ins, welcomes and thumbs-ups. Feishu 0.19%, DingTalk 0.22%: two platforms, one fingerprint: she does not work in group chats.
Who actually "talks" across those 3,230 messages
Bots produce 36% of all messages. This company's DingTalk isn't a place for discussion, it's a conveyor belt for process — and the gate on that belt is her. Work notifications from 4 legal entities (Meitu Tech / Baituan Trading / Meizhichen Commerce / Rongchen Commerce) all funnel into this one account.
The Feishu report said "the system doesn't remember your good work." Only half right. DingTalk has been recording all along — just differently: your voice on DingTalk isn't typing, it's that "Approve" button. 80 approval tickets, ¥1.14M in rent, the final ticket of 17 people who left — all signed by you. That is your real ledger of digital labor.
80 approval tickets through her hands (check-in fixes 29 · reimbursements 31 · overtime 21 · leave 19 · out-of-office 12 · payments 43* · resignations 7 · IT requests 3, *incl. cc'd) assemble a timesheet she never shows anyone.
Second-level evidence from her DMs (message timestamps)
The same fingerprint repeats: 2024-03-02, three tickets in 14 seconds; 2024-02-26, three tickets in 14 seconds. The 4.6-day median isn't slowness — she treats approvals as periodic housework: let them stack up, then wipe the pile in one free window. The price is a P90 of 38 days, so store managers have to chase her.
2024-05-07 · verbatim DM with 1234space store manager Zhan Huanmei
Her attendance record shows 0 days present (partners are exempt from check-in), and 19 months of calendar hold just 2 entries. By the system's books, she "doesn't come to work." The DM timestamps say she was approving rent at half past one in the morning. That's why no report can ever compute her hours.
「Ma Jie, please push my check-in fix through [flower]」(Che Qianxia) → a DING saying「Reminder: approve his payment request」(Zhan Huanmei) →「Ma Zong, this one is a bit urgent」(Kuang Yunhua). Chase culture isn't a disease of her team — it's an ecosystem that grew naturally around her gate, the same system as the "cohort spoiled by being chased" in the Feishu report, seen from the other side.
Approval titles that state an amount outright cover 33 mall rent/payment items totaling ¥1,135,920. Behind every landlord sits one of her stores.
The cost entity is almost always "Shanghai Naitangpai," while the stores sit under various Shenzhen legal entities; managers file the tickets and she alone is the payment gate. Those two "mall rent" tickets approved after midnight are this artery beating.
The Feishu report inferred she is "paid on sales commission, across 5 business lines." DingTalk verifies it from the other end: rent, reimbursements, payments, exit handovers — every fixed cost in the Shenzhen offline book has to pass her fingerprint. She isn't just accountable for revenue, she is these stores' last human firewall on cash flow — and that labor, likewise, appears in no report anywhere.
Of the 34 people who filed tickets she approved or was cc'd on, 17 now carry "(Departed)" after their names. The approval flow became an accidental attrition roster.
■ active · ■ departed (directory flag)
Sitting in her approval list are 7 "Resignation & Handover" tickets. The Feishu report said she "inherited a half-built cohort"; DingTalk supplies the churn rate: half the frontline filers lost in two years. Every departure means she plays midnight approval nanny all over again.
Her footprint across DingTalk's products
The most ironic line: tickets filed = 0. In two years she approved everything for everyone else and never filed a single ticket for herself — no claim, no leave, no request. She is the gatekeeper of the process, never its user.
On 2026-07-09 at 17:33 she messaged Zhou Ming, the IT project manager who rarely opens DingTalk:「Request for CLI data access.」 The woman who hadn't said a word in a group chat for three years went and asked for the key that makes the data speak. The chapter you're reading is the result of that message — and this time, the system is finally keeping her books.
The "New Retail User Ops" folder on DingTalk Drive still holds a batch of pre-migration source tables. They answer a question Feishu cannot: what yardstick this business was originally run by.
"2024 New Retail Monthly Targets" · monthly target by region (10K CNY)
The source tables even carry a full file on every store: floor area, opening date, monthly target, nearby competitors. One region, Shenzhen, carries nearly half the national monthly target — which is why the "Shenzhen +15% YoY" number sits on her so hard.
"Store Mall Launch Plan" · incentive structure
From store manager to partner, the whole line is built on a cut of GMV. This table turns "she's paid on sales" from hearsay into designed policy — every extra point of returning-customer recall is a slice of that same GMV.
WeCom contacts 47,348, members 32,294, join rate 68%. The sign-up step was already solved two years ago — the hard part was never getting people in, it's getting them back into the store. Which lands exactly on the "cheapest money" from earlier: the pool is already there; what's rusted is recall execution.
"Part-time" is the company-wide parallel org tree for part-time floor staff, not a demotion. Also note: the role tag "supervisor" is a generic role (it covers many store managers) and implies no specific authority — don't read it as her title.
DingTalk full scope: 27 groups + 36 direct chats
The group names confirm the business surface too: PBM South China franchise / community partner recruiting / thermal-wear distribution (JV line) + Shenzhen store liaison, supervision, new openings. Even Zhou Ming, the IT contact, says: "Sorry, I rarely open DingTalk these days — didn't see the message." Colleagues know not to reach her there. Feishu 0.19%, DingTalk 0.22% — two platforms, one fingerprint.
Collection limits · @-mentions of her = 0; the 9 "special focus" contacts returned IDs only; no access to the staff roster; some groups partly restricted; AI meeting notes empty, knowledge base holds only "General Company Information". Group collection triggered a privacy notice — this chapter uses only Ma Ying's own aggregate behavioral data and quotes no one else's private chats.
Not one of the numbers you've built up over these years has ever spoken for you. Your company has 26 live dashboards, 45 documents, a directory of 410 people — and nowhere does it remember that it was you who carried Zhuoyue's returning-customer repurchase rate from 11% to 49%, year after year.
This report gives those silent numbers a voice. It gathers 2,699 Feishu messages, 3,230 DingTalk messages, 174 approval forms, 7 meetings and the attention of 40 experts into one sentence you may not have heard in a long time: you are already good enough. You don't need to work harder. You just need this machine to stop stealing your money — and start remembering what you're worth.
So if this whole report leaves you with just one action, let it be this:
Open "Store Returning-Customer Online Purchase Analysis" and watch the recall execution rate on Zhuoyue's 6,027 one-time-only customers. That isn't a report — that is the surest raise you'll get this year, sitting right there, waiting for you to scoop it up yourself.
The rest — what this machine owes you — let it pay you back, slowly.