王峰 AI × 经营 约一次诊断

总目 / 文章 / WF-0350

文章 · 存档

从“每卖一台亏一台”到“毛利转正”:我的共享CFO诊断手记

我接触这家智能制造公司时,管理层反复提到一句话:“产线在跑,订单不少,但越忙越亏。”

他们把财务报表跟我看,“财务数据是假的。”这是我看到数据的第一句话。

“你怎么知道?”

“我一看就知道,你也别问我是咋看出来的。我需要看你们的业务数据。”

于是,我把他们的真实经营情况重新做了财务口径的还原,基本上的结论就是:连续两个季度毛利率为负,更可怕的是现金流逐渐吃紧。

先不谈“战略”,只问三件小事

第一问:标准成本怎么来的?BOM是谁维护、多久校准、有没有与实际用料对齐。答案往往是“工程+采购+生产各有一套”。

第二问:节拍与良率真实多少?不看公司的月度经营报告,看产线白板、报工单、返工区。

第三问:报价口径一致吗?同一型号,不同客户、不同渠道,价格含不含运保关、售后备件、佣金返点、以旧换新折价?口径一乱,毛利就“凭感觉”。

把“毛利”为何为负翻译成一条可走的线

我用一张“单台产品毛利桥”把问题拉直:

出厂价 → 减:材料(BOM实耗)→ 加工(工时×工序成本)→ 制造间接(能耗/折旧/工装/维保/品保)→ 供应链与物流(头程、关税、最后一公里)→ 市场与渠道(佣金/返点/促销)→ 售后保修与退换 → 毛利。

再把每一块拆成“产品×工序×设备×班次×客户/渠道”的可比表。事实通常很刺眼:不是所有产品都亏,是某些工序在特定机台与特定订单组合上必亏;不是所有客户压价,是个别渠道费与隐藏条款把毛利吃没。

逐工序“解剖”:把真实成本掰成能看懂的粒度

材料(BOM):把“标准用量”改成“配方+损耗曲线”。区分首件、量产、换型时的不同损耗;把来料等级差异纳入良率影响。

工时与节拍:用“TDABC(时间驱动作业成本)”替换大锅饭工时,按节拍、换线时长、远程调机时间计入。

设备与能耗:设备折旧、刀具工装、能耗按“开机+负载曲线”计量,避免“按面积/人数摊”。

质量与返工:良率不是一个数,是首件→爬坡→稳态三段曲线;返工工时与二次检测独立核算,不允许“吞到制造费用里”。

供应链:海空运与集拼时效差异,按真实线路计价;关税税则归类、原产地证变化引起的税负差异,单独出列。

渠道与客户:佣金、平台费、跨境收单费、汇率点差、账期资金占用,都入“可变直接费用”,不再埋在“销售费用”。

把这些都放进一张“SKU-工序-渠道”的矩阵后,亏损点浮出水面:

BOM漏项导致材料成本低估4–7%;

两道关键工序的换线时间被忽略,每天损失近1.2个班次产能;

跨境渠道把售后备件与延保算在总部费用,单台少计成本2–3点;

个别订单采用“工程样机配置”却按“量产报价”签单;

汇率与运价上涨周期内,外销价未做动态基差调整;

计件制度只奖“速度”不奖“合格率”,导致快工多返修。

三张图+四个表:团队每天都能用的“看得懂”的管理底盘

三张图

“SKU毛利瀑布图”:从报价到毛利的每一步消耗,直观看差异源头。

“工序热力图”:工序×设备×班次的单位成本与良率热度,亮红处是瓶颈。

“客户/渠道毛利散点图”:横轴价格、纵轴净毛利,气泡大小代表量,异常点一眼看穿。

四个表

“BOM实耗对账表”:标准 vs 实耗 vs 供应商来料差异。

“TDABC工时表”:节拍、换线、调机、等待分档计时。

“售后费用三角表”:保修率×件均成本×时间段,反推结构性改良。

“跨境费用口径表”:运保关、收单费、平台费、佣金、返点、账期利息,统一口径,签约即固化。

AI不是口号,是几个具体的小工具

视觉质检与返修溯源:用相机+轻量模型定位关键缺陷,自动回帖到工单;返修原因闭环到工艺参数。

工艺参数与产能的“数字孪生”:把节拍、良率、换线时间喂给模型,做不同订单组合下的毛利预估,接单前就知道赚不赚钱。

Process Mining(流程挖掘):从MES/ERP日志里自动拉出等待、重工、跨部门来回的隐藏时间。

价格-成本联动器:运价、汇率、来料价实时驱动动态报价基差,避免“上半年价、下半年运”。

低代码报工:一键抓取工位开停机、首件通过、换线开始/结束,减少手工录入误差。

费用口径智能校验:签约条款与费用入账自动匹配,发现“承诺佣金未计”“渠道返点双计”等口径风险。

关键动作与顺序:先止血,再造血

一周止血:

冻结“工程样机按量产价”签约;

上线“跨境费用口径表”,立刻把渠道费、延保费计入单台成本;

对两道关键工序实行“合格率门槛+换线有价”的计件新规。

一月修复:

完成前20大SKU的BOM体检与校准;

部署TDABC,重算工序单价;

把“数字孪生”结果接到接单审批流,低于红线的单子必须有“结构性让利换量”的商业理由。

一季重构:

渠道价格体系回表:价格、返点、账期三选二;

供应商共创:把来料一致性纳入季度回扣与双赢条款;

产线IE优化:换线并单、工装快换;

售后数据反哺设计:高发缺陷进入下一版工艺与BOM。

结果是怎样来的(而不是运气)

毛利由负转正的第一推动是口径统一和BOM纠偏;

第二推动来自工序节拍与换线成本被真实计价后,订单组合被重新排程;

第三推动是渠道价格基差机制,让“量、价、费、险”回到同一张桌子谈。

数据不会拍马屁,所有变化都能在“毛利桥”里逐条对齐,在“工序热力图”里逐日变淡。

共享CFO能为老板带来什么

不是做账,是让账好用:把财务从“报表生成”升级为“经营镜像”,任何一个价格、工艺、供应、渠道的决策,都能在下单前看到毛利的走向。

让复杂变简单:三张图四个表,工厂中层也能读懂、会使用,不需要懂一堆术语。

机制重于口号:把“算得清、看得见、改得动”固化为制度与系统,使毛利改善可复制。

以小博大:AI与数字化从“几个关键点”切入,将投入控制在可承受范围内,却能为后续规模化铺路。

智能制造讲究技术,但利润从来长在口径清楚、数据可靠、机制到位的土地上。

毛利转正不是妙手回春,而是把每一笔成本、每一道工序、每一个渠道,摆到阳光下。

等你也有了属于你的“三图四表”,你会发现,赚钱这件事,其实可以被设计。

以上,是我作为共享CFO在一家智能制造企业里做过的诊断与落地路径。

愿每一家认真做产品的公司,都能把毛利折腾得清清楚楚、踏踏实实地向上走。

From “Losing Money on Every Unit” to “Gross Margin Turned Positive”: My Shared CFO Notes

When I first met this smart manufacturing company, the managers kept saying:

“The line is running, orders are plenty, but the more we work, the more we lose.”

They showed me the financial report. My first words were:

“These numbers are fake.”

“How do you know?”

“I can just tell. Don't ask me how. I need to see your business data.”

So I rebuilt their real business numbers into a finance view. The bottom line was this:

Two quarters in a row of negative gross margin. Worse, cash flow was getting tight.

Forget “Strategy,” Ask Three Small Questions

First: How do you get standard cost? Who keeps the BOM? How often is it checked? Does it match real use? The answer is often: “Engineering, buying, and production each has their own.”

Second: What's the real speed and yield? Don't read the company's monthly report. Look at line boards, job tickets, and rework areas.

Third: Are prices quoted the same way? Same model, but different clients and channels—do prices include freight, duty, spare parts, sales rebates, trade-in discounts? Once the rules are mixed, gross margin becomes “by feeling.”

Turn Negative Gross Margin Into a Clear Path

I drew a “single unit gross margin bridge” to show the story:

Selling price → minus: materials (real BOM use) → labor (hours × process cost) → factory overhead (power, wear, tools, repair, QA) → supply chain and logistics (freight, duty, last mile) → market and channel (commission, rebate, promo) → warranty and returns → gross margin.

Then I broke each part down by product × step × machine × shift × client/channel. The truth was sharp:

Not all products lose money—only some steps on some machines with some orders. Not all clients push price down—some channels hide costs that kill margin.

Break It Down by Step: Make Real Cost Clear

Materials (BOM): Replace “standard use” with “recipe + loss curve.” Separate first lot, mass run, and changeover loss. Count raw material grade in yield.

Labor and speed: Use TDABC (time-driven cost) instead of “one big pot” hours. Count line speed, changeover time, remote setup.

Machines and power: Depreciation, tools, energy all by “on-time + load curve,” not split by floor area or headcount.

Quality and rework: Yield is not one number—it's first lot → ramp up → stable. Rework hours and second checks are counted alone, not hidden in overhead.

Supply chain: Sea vs air vs group shipment priced by real route. Tariff changes from HS code or origin listed alone.

Channel and clients: Commission, platform fees, cross-border fees, FX gap, payment terms—all go to “direct variable cost,” not buried in “sales cost.”

After putting all this into a “SKU–step–channel” matrix, the loss points showed up:

BOM missing items made material cost too low by 4–7%.

Two key steps ignored changeover time—lost 1.2 shifts of output each day.

Cross-border sales hid spare parts and warranty in HQ costs—2–3 points missed per unit.

Some orders used “prototype build” but priced as “mass run.”

In cycles of higher freight and FX, export prices stayed fixed.

Piece-rate pay only rewarded speed, not yield—fast work meant more rework.

Three Charts + Four Tables: A Simple Base the Team Can Use Daily

Three charts

SKU margin waterfall: From quote to gross margin, see where the value leaks.

Step heatmap: Step × machine × shift cost and yield, red spots are bottlenecks.

Client/channel margin scatter: X-axis price, Y-axis net margin, bubble = volume. Outliers stand out.

Four tables

BOM real-use table: Standard vs real vs supplier input.

TDABC time table: Speed, changeover, setup, wait times.

Warranty triangle: Rate × cost per unit × time, push design fixes.

Cross-border fee table: Freight, duty, card fee, platform fee, commission, rebate, interest—one rule, fixed at contract.

AI Is Not a Slogan, but Small Tools

Vision QA + rework trace: Camera + small model catch defects, auto link to job ticket, close loop with process settings.

Digital twin of process: Feed speed, yield, changeover to model—predict margin before taking order.

Process mining: MES/ERP logs show hidden waits, rework, back-and-forth.

Price–cost link: Freight, FX, input prices drive live quote changes.

Low-code job log: One click to grab machine start/stop, first-pass pass, changeover start/end.

Smart fee check: Auto match contract vs cost entries, flag “promised fee not booked” or “rebate booked twice.”

Key Moves in Order: Stop Loss, Then Grow Profit

One week, stop loss:

Freeze “prototype at mass-run price.”

Start “cross-border fee table,” count all costs per unit now.

New piece-rate: yield threshold + paid changeover.

One month, fix:

Check and adjust BOM for top 20 SKUs.

Deploy TDABC, reset step cost.

Link digital twin to order flow—below-line orders need business reason.

One quarter, rebuild:

Channel pricing reset: price, rebate, terms—pick two of three.

Supplier co-work: input consistency tied to rebate and win–win terms.

Line IE improve: group changeovers, quick tool change.

After-sales data feed back to design: common defects fixed in next BOM.

Results Came from Work, Not Luck

The first push to profit was: same rules and fixed BOM.

The second push: real step speed and changeover counted, order mix reset.

The third push: channel base price tied to cost—so volume, price, cost, and risk sit at one table.

Data doesn't flatter. Every change lines up in the “margin bridge” and shows up cooler in the “step heatmap.”

What Shared CFO Brings to a Boss

It's not just making reports—it's making reports useful: finance as a mirror of business. Any price, step, supply, or channel choice can show margin before the order.

Make hard things simple: three charts and four tables—factory mid-managers can read and use, no big terms needed.

System over slogan: “clear, seen, fixable” is built into rules and systems, so margin gains repeat.

Small tools, big gain: AI and digital cut in at key points, cost is low but paves way for scale.

Smart manufacturing cares about tech, but profit grows only on clear rules, true data, and strong system.

Turning margin positive is not magic—it's putting every cost, every step, every channel in daylight.

Once you also have your own “three charts and four tables,” you'll see: making money can be designed.

This was my diagnosis and action path as a Shared CFO in a smart manufacturing firm.

May every company that builds real products make gross margin clear, and grow it steady.