Cash flow forecasting for garment OEM: the 4-week model成衣OEM的現金流預測：4週模型成衣OEM的现金流预测：4周模型

Garment OEM is a business where you routinely spend other people's money before they give it to you. Fabric must be bought, trims committed, and factory deposits paid weeks or months before the customer's balance arrives. A company can be profitable on paper and still miss a payroll, because profit is a season-level concept and cash is a Tuesday-level concept. The tool that bridges the two is a rolling short-horizon cash forecast — and for most OEM and trading operations, four weeks is the horizon that matters.

成衣OEM是一門「例行性地先花別人的錢，之後才收到那筆錢」的生意。面料要先買、輔料要先訂、工廠訂金要先付，客戶的尾款卻在幾週甚至幾個月後才到。一家公司可以帳面獲利卻發不出薪水，因為獲利是「一季」層級的概念，現金是「這個星期二」層級的概念。銜接兩者的工具，是一份滾動的短期現金預測——對多數OEM與貿易營運而言，四週就是那個關鍵視野。

成衣OEM是一门「例行性地先花别人的钱，之后才收到那笔钱」的生意。面料要先买、辅料要先订、工厂订金要先付，客户的尾款却在几周甚至几个月后才到。一家公司可以账面盈利却发不出工资，因为盈利是「一季」层级的概念，现金是「这个星期二」层级的概念。衔接两者的工具，是一份滚动的短期现金预测——对多数OEM与贸易运营而言，四周就是那个关键视野。

Why four weeks為什麼是四週为什么是四周

Longer forecasts are strategy; the four-week window is operations. Within four weeks, almost every cash event is already determined by a document that exists today: an invoice with terms, a vendor payment milestone, a deposit due on a signed subcontract, a shipment that will trigger billing. Four weeks is near enough to be accurate and far enough to act — to accelerate a receivable, negotiate a payment date, or delay a discretionary purchase before the squeeze arrives rather than during it.

更長的預測是戰略；四週視窗是營運。在四週之內，幾乎每一筆現金事件都已經被今天就存在的某份單據決定了：一張帶付款條件的發票、一個供應商付款里程碑、一筆已簽外包合約的到期訂金、一票即將觸發請款的出貨。四週近到足以準確，又遠到來得及行動——在資金吃緊來臨之前（而不是之中）催收一筆應收、談一個付款日、或推遲一筆非必要採購。

更长的预测是战略；四周窗口是运营。在四周之内，几乎每一笔现金事件都已经被今天就存在的某份单据决定了：一张带付款条件的发票、一个供应商付款里程碑、一笔已签外包合约的到期订金、一票即将触发请款的出货。四周近到足以准确，又远到来得及行动——在资金吃紧来临之前（而不是之中）催收一笔应收、谈一个付款日、或推迟一笔非必要采购。

The model itself模型本身模型本身

The structure is deliberately simple: one row per week, four weeks out, rebuilt every Monday. Expected receipts: open customer invoices placed in the week their terms say they should pay — adjusted by each customer's actual paying behavior, which your AR aging already reveals — plus deposits due on newly confirmed orders. Committed payments: vendor invoices by due date, factory payment milestones (deposit, against-shipment, balance), subcontract obligations, and the recurring block of payroll, rent, and freight. The difference, week by week, laid against your opening balance, is the forecast. No heroics, no macros — the sophistication is not in the arithmetic.

結構刻意簡單：一週一列，往後四週，每週一重建。預期收款：未結客戶發票按付款條件應付的那一週入列——再依每位客戶實際的付款習性調整（您的應收帳齡早就把這件事告訴您了）——加上新確認訂單的到期訂金。已承諾付款：供應商發票按到期日、工廠付款里程碑（訂金、見提單、尾款）、外包義務，以及薪資、租金、運費這塊固定支出。逐週的差額對上期初餘額，就是預測。不需要絕技，不需要巨集——這個模型的深度不在算術。

结构刻意简单：一周一行，往后四周，每周一重建。预期收款：未结客户发票按付款条件应付的那一周入列——再依每位客户实际的付款习性调整（您的应收账龄早就把这件事告诉您了）——加上新确认订单的到期订金。已承诺付款：供应商发票按到期日、工厂付款里程碑（订金、见提单、尾款）、外包义务，以及工资、租金、运费这块固定支出。逐周的差额对上期初余额，就是预测。不需要绝技，不需要宏——这个模型的深度不在算术。

The three inputs that decide everything決定一切的三個輸入决定一切的三个输入

Aging you can trust. If the AR aging report mixes settled invoices with open ones, or misses the deposit that was already applied, the receipts side of the model is fiction. This is the argument for running receivables on a proper ledger — every invoice, receipt, deposit, and offset recorded as a permanent movement — so 'what is actually open, by customer, by week' is a query, not a reconstruction.

可信任的帳齡。如果應收帳齡把已結清的發票和未結的混在一起，或漏掉已經沖抵的訂金，模型的收款端就是虛構。這正是應收該跑在真正帳本上的理由——每張發票、每筆收款、訂金與沖抵都記為不可改的動作——這樣「以客戶、以週來看，實際還有多少未收」是一句查詢，而不是一場重建。

可信任的账龄。如果应收账龄把已结清的发票和未结的混在一起，或漏掉已经冲抵的订金，模型的收款端就是虚构。这正是应收该跑在真正账本上的理由——每张发票、每笔收款、订金与冲抵都记为不可改的动作——这样「以客户、以周来看，实际还有多少未收」是一句查询，而不是一场重建。

Payables with milestones, not just totals. A factory contract worth $80,000 is not one cash event; it is a deposit this month, an against-shipment payment next month, and a balance after inspection. If your AP only knows the total, the forecast will be right on average and wrong every single week. Payment milestones must live on the contract and flow into the model by their own dates.

帶里程碑的應付，而不只是總額。一張八萬美元的工廠合約不是一筆現金事件，而是這個月的訂金、下個月的見提單付款、驗貨後的尾款。如果您的應付只知道總額，預測就會「平均起來對、每一週都錯」。付款里程碑必須掛在合約上，各自按日期流進模型。

带里程碑的应付，而不只是总额。一张八万美元的工厂合约不是一笔现金事件，而是这个月的订金、下个月的见提单付款、验货后的尾款。如果您的应付只知道总额，预测就会「平均起来对、每一周都错」。付款里程碑必须挂在合约上，各自按日期流进模型。

FX at the rates that actually apply. Most OEM books have a structural mismatch: revenue in USD, costs split across TWD, RMB, VND, or others. A four-week model in one blended currency hides exactly the exposure it should reveal. Run the receipts and payments per currency, converted at the rate that was in force when each document was booked — and treat the gain or loss at settlement as its own line, not as noise smeared into margins. In a year when tariff decisions and trade rulings have been moving faster than shipping schedules, currency and duty assumptions deserve their own visible row, reviewed weekly.

用真正適用的匯率看外幣。多數OEM帳本有一個結構性錯位：收入是美元，成本散在台幣、人民幣、越南盾或其他幣別。用單一混合幣別做四週模型，恰好把它最該揭露的曝險藏起來。收付款按幣別分開跑，各自以單據入帳當時生效的匯率換算——結算時的匯兌損益單獨成列，不要抹進毛利當雜訊。在一個關稅決定與貿易裁決比船期變得還快的年份，匯率與關稅假設值得擁有自己看得見的一列，每週檢視。

用真正适用的汇率看外币。多数OEM账本有一个结构性错位：收入是美元，成本散在台币、人民币、越南盾或其他币别。用单一混合币别做四周模型，恰好把它最该揭露的敞口藏起来。收付款按币别分开跑，各自以单据入账当时生效的汇率换算——结算时的汇兑损益单独成行，不要抹进毛利当杂讯。在一个关税决定与贸易裁决比船期变得还快的年份，汇率与关税假设值得拥有自己看得见的一行，每周检视。

Keeping the numbers clean enough to forecast讓數字乾淨到能拿來預測让数字干净到能拿来预测

Every finance team eventually learns that forecast quality is a byproduct of transaction discipline. Two habits matter most. First, corrections should be new reversing entries, never silent edits — an append-only record means last Monday's forecast can always be explained against this Monday's. Second, sensitive money actions — vendor payments, credit notes — should require a second approver before they post. Maker-checker is usually framed as fraud control; its quieter benefit is that the numbers entering your forecast have already been looked at twice.

每個財務團隊遲早會學到：預測品質是交易紀律的副產品。兩個習慣最重要。第一，更正應該用新的沖銷分錄，永遠不要無聲修改——只增不改的記錄，代表上週一的預測永遠能對著這週一的解釋清楚。第二，敏感金流動作——付款給供應商、開折讓——過帳前應該經第二人核准。雙人審批常被說成防弊；它更安靜的好處是：進入預測的每個數字，都已經被兩雙眼睛看過。

每个财务团队迟早会学到：预测质量是交易纪律的副产品。两个习惯最重要。第一，更正应该用新的冲销分录，永远不要无声修改——只增不改的记录，代表上周一的预测永远能对着这周一的解释清楚。第二，敏感资金动作——付款给供应商、开折让——过账前应该经第二人核准。双人审批常被说成防弊；它更安静的好处是：进入预测的每个数字，都已经被两双眼睛看过。

None of this requires replacing your accountant or your accounting software. The operational system carries the orders, invoices, milestones, and FX events as they happen; the model reads them; and the accounting system receives clean exports. The forecast lives where the operation lives, because that is where the cash events are born.

這一切都不需要換掉您的會計師或會計軟體。營運系統即時承載訂單、發票、里程碑與匯率事件；模型讀取它們；會計系統收到乾淨的匯出。預測活在營運所在之處——因為現金事件就是在那裡誕生的。

这一切都不需要换掉您的会计师或会计软件。运营系统实时承载订单、发票、里程碑与汇率事件；模型读取它们；会计系统收到干净的导出。预测活在运营所在之处——因为现金事件就是在那里诞生的。

Start next Monday: pull your open AR by expected week, your AP and milestones by due date, split by currency, and write the four rows. The first version will expose gaps in your data before it exposes gaps in your cash — and closing those data gaps is, quietly, the whole game.

就從下週一開始：把未收帳款按預期收款週拉出來、應付與里程碑按到期日排好、按幣別分列，寫下那四列。第一版預測會先暴露資料的缺口，然後才輪到現金的缺口——而補上那些資料缺口，安靜地說，就是這整場賽局。

就从下周一开始：把未收账款按预期收款周拉出来、应付与里程碑按到期日排好、按币别分列，写下那四行。第一版预测会先暴露数据的缺口，然后才轮到现金的缺口——而补上那些数据缺口，安静地说，就是这整场比赛。