How to use Costdown
Costdown is where you record a cost-reduction improvement as numbers, then compare it against other factories in the same country. The first two sections walk you through doing exactly that, start to finish. The rest explain things you cannot work out by looking at the screen — read them when you need them, not before you start. Each input field also has its own explanation in place: see The ? beside each field.
Where to start
It depends on what you came here to do. Three ways in, in no particular order:
| You want to | Go to | What you need first |
|---|---|---|
| Work out what an improvement would save, without publishing anything | Intelligence → Calculator | Nothing. It works without an account and stores nothing. |
| Record an improvement you already made, so others can see and compare it | Workspace → Write | Your cost figures before and after, and how you measured both. |
| See what others in your industry have done | Discover → Projects | Nothing. |
The Calculator and the Write page share the same formulas, so a figure you try out in the Calculator comes out the same when you turn it into a proposal.
Writing a proposal, start to finish
This follows the actual order of the fields on the Write page. Before you open it, five things are worth having in hand — you can write without any of them, but the proposal then sits at a low trust level:
- The cost BEFORE and the cost AFTER, broken down by category — not a single total.
- How you measured those two figures: hand counts, shift logs, production reports, invoices.
- Monthly volume, working days, shifts per day, hours per shift — the system needs them to bring costs down to a single part.
- The money already invested, if any. If none, tick "This proposal needs no investment".
- Machine names, material codes, specific models — the more specific, the more a reader can match it against their own shop floor.
| Step | What you do | What you have afterwards |
|---|---|---|
| 1 | Go to Workspace → Write. Type the title and pick the Industry. | The two required fields are done. Picking the industry is what makes the fields below suggest the right machines and materials. |
| 2 | Pick the Process, the main KPI type and the difficulty. | These three decide which comparison group your proposal lands in. Leave them blank and it still saves, but it stays out of the ranking. |
| 3 | Write the detailed content, then the three description fields: the problem, the mechanism, the short summary. | This is what a reader reads first. The short summary is what appears in listings, so write it as an answer, not as a headline. |
| 4 | Move to the cost table. For each category involved, click "+ Add row" and fill the Before and After columns. | The panel on the right shows the total and the savings percentage immediately, updating as you type — there is no Calculate button to press. |
| 5 | Fill in the conditions of application and the risks. | These two separate a usable proposal from a boast. A reader needs to know where your improvement does NOT apply. |
| 6 | Fill in how the result was verified and the measurement method. | This is the field that decides the trust level. Be specific: what was measured, with what, against what, who signed it off. |
| 7 | Choose hard cash or cost avoidance. If unsure, leave it blank. | The reader learns whether your figure reaches the invoice or not. |
| 8 | Enter the investment and the implementation time, or tick the no-investment box. | The system works out the months to payback — you do not have to. |
| 9 | Press Save. | The proposal appears under My proposals. From then on, every change to a savings figure is recorded: old value, new value, when. |
About the Before and After columns: Before is the current state, After is once the improvement is in place. If a category is untouched by the improvement, leave BOTH columns blank — do not copy the Before figure into After to fill the gap, because that states a category the improvement has nothing to do with.
The panel titled "What the system reads from your figures" sits in the right-hand column and runs while you type, not after you save. It points out things that are hard to notice yourself: a row contributing almost nothing, a category wildly out of line with the rest, a place where you filled one side and left the other blank. Reading it while typing is worth far more than reading it after forty rows are already in.
If you are not ready to publish anything, use Intelligence → Calculator: same formulas, no account needed, nothing stored. Try figures there first and move to Write later if you want.
The ten cost categories
Costs are split into ten categories. You do not have to fill in all ten — only the ones your improvement actually touches.
| Category | Fill it in when |
|---|---|
| Labor | Someone spends time on the product: running a machine, assembling, inspecting, setting up. |
| Machine | A machine runs: depreciation, power, maintenance, machine time occupied. |
| Material | Material that goes into the product. |
| Scrap | Material or product thrown away. |
| Consumables | Tooling, drills, grinding wheels, cutting fluid — things that wear out and get replaced. |
| Outsourcing | A step done by an outside supplier. |
| Rework | Bad parts repaired to pass, rather than scrapped. |
| Packaging | Boxes, trays, film, dunnage. |
| Logistics | Freight tied to this product. |
| Factory overhead | The share accounting has already allocated to the product. Added per month, not multiplied by volume. |
Blank does not mean zero
This is the easiest thing to get wrong, and getting it wrong produces a wrong cost figure that looks exactly like a right one.
A blank field means you have NOT STATED it, not that you are asserting the category is zero. The system distinguishes the two states and displays them differently. If you genuinely mean "this costs nothing", type 0 — that is an assertion, and it is not the same as leaving the field empty.
The reason: a proposal with half its categories blank still produces a very attractive savings figure. If blank counted as zero, nobody could tell "this improvement really saves a lot" apart from "the author has only filled in half of it".
Hard cash or cost avoidance
These two kinds of saving look identical on paper but behave differently in the accounts, so every proposal should say which one it is.
- Hard cash: after the improvement, the amount actually paid each month goes down compared with what is being paid today. You pay 50 million for electricity, replace the motor, and it becomes 42 million — next month's bill is 8 million lower.
- Cost avoidance: the amount paid does not go down; it simply fails to rise the way it would have. The supplier announces an increase from 50 to 55 million, you renegotiate and change material to hold it at 50 million — the bill is still 50 million, and accounting sees nothing go down.
If you are not sure, leave it blank. Blank means "not stated"; guessing means stating something false, and a reader who knows the trade will notice.
The four trust levels
Every proposal carries a trust level. The system RECALCULATES it from the evidence actually present — it is not a label you pick, and writing at greater length does not raise it.
| Level | What it means |
|---|---|
| Unverified | The figure was calculated by the author, with no measurement evidence. |
| Self-evidenced | The figures are complete and the author has stated how they were measured. |
| Counterpart verified | A factory applied it and agreed the real figure jointly. |
| Platform verified | An independent third party checked it and concluded it holds. |
The "How it was verified" field is the most important one when writing a proposal. Something vague like "we observed a clear reduction" leaves the proposal at the lowest level and out of the ranking, no matter how good the numbers are. Something specific — what was measured, with what, against what, who signed it off — raises the level.
The number of posts never counts towards ranking, not even as a tiebreaker. Posting more does not place you higher.
Comparing against other factories
Ranking and comparison happen WITHIN A SINGLE COUNTRY only. An improvement saving 20% in Vietnam and one saving 20% in Germany cannot be compared directly: labour, power and material prices are too far apart.
For that reason comparisons are measured in percentages and in months to payback, never in absolute money. The exchange rate is taken for real at the moment of saving and then frozen, so an old proposal does not change meaning when rates move.
Grouping with colleagues
Colleagues from the same company are grouped by a random join code, not by naming the employer. You create a code, hand it to a colleague, and whoever enters it joins the same group — which lets you share internal material codes and pool figures without saying where you work.
The code is shown ONCE at creation and is never stored in readable form. If you lose it, nobody can retrieve it, administrators included — create a new one and revoke the old.
When someone applies your improvement
A proposal can go on to become a project with two sides: the one who offered the improvement and the one who applied it. The side that applied it submits the real figures afterwards, and the other reviews and confirms them. Only when both agree is the result recorded as verified.
That is the thing a spreadsheet cannot give you: a number that someone else, at another factory, has put their name to as well.
The ? beside each field
Almost every input field has a ? beside it. It opens that field's own explanation, and each one answers four questions: what it is, when to fill it in and when not to, an example with real numbers, and what other situations the field covers beyond its own name.
The fourth is the one worth reading. A field name is the shortest name, not the fullest one — seeing "Weight (kg)" and assuming it is only for metal means skipping the whole material category when you work in plastics, textiles or wood.
Where a field has a formula, the explanation gives the formula and its source.
Why not just use Excel
That is a fair question, because Excel is faster to type into, freer in layout, and works offline. Costdown does not try to win on any of those three.
There are five things a spreadsheet cannot do, and they are the entire reason Costdown exists:
- Share one library of materials, machines and processes — instead of everyone typing their own spelling of the same thing and then being unable to add the numbers up.
- See figures from other factories in the same industry and the same country.
- Let a counterpart who actually applied your improvement confirm the real number themselves.
- Keep an edit history that cannot be deleted: change a savings figure and the system records the old value, the new value, which account changed it, and when.
- One single formula for everyone — no two spreadsheets producing two answers and an argument about which one is right.
If the work in front of you needs none of those five, Excel really is faster, and we would rather say so plainly.
When something does not match
If anything in the product behaves differently from what this page describes, that is our mistake and not yours — please tell us through the 聯絡我們. Messages are stored and given a case number immediately, so your request does not disappear.
分產業指南
針對具體產業而非泛泛而談:如何拆分成本、常見錯誤在哪裡,以及車間裡已有的數據從哪兒取。
全產業通用
- 成本改善案例要怎麼記錄,別人才會相信
在 Costdown 記錄一件降本專案的四個步驟、由證據決定的四級信任度、只在同一個國家內進行的標竿比較,以及如何沿用已驗證專案的結構。
- 省下來的錢是真的嗎——瓶頸,與數字變成現金的三個條件
為什麼機台跑快了利潤卻沒增加。從製程限制的角度,說明計算出來的節省金額在什麼條件下才會變成真正的現金,以及提案送出前怎麼自己先判斷。
- 六個最常空白的欄位,以及每一個會把成本算歪哪一邊
良率、寬放、電費、每箱裝量、每批固定費用——這些欄位最常沒填。空白和「零」是兩件事,混為一談會產生一個看起來完全正常的錯誤成本。
- 成本計算公式——兩條公式,回答兩個不同的問題
期間的會計公式和單位成本公式回答的是不同問題。會計三要素與十大項目的對照表,以及兩個數字對不起來為什麼是正常的。
機械加工・精密切削
- 機械加工成本計算——十大項目,照製程順序走一遍
把一個切削件的成本完整拆開:材料、機時、工時、刀具、報廢、重工、委外、包裝、運費、製造費用。從「同一張圖為什麼兩個人報出兩個價」開始講起。
- 廠裡沒有數據,成本計算從哪裡開始
不需要系統,也不需要花一個月調查。挑一個件,一週內量三個數字,其餘全部從廠裡已經有的單據取得,就能做出一張站得住腳的成本表。
- 刀具成本怎麼換算成單件,以及三個降法
刀具是整盒買進來的,卻是一件一件消耗掉的。怎麼換算、怎麼不花錢量刀具壽命,以及哪三個改善真的會反映到利潤上。
鈑金:雷射切割與折彎
- 鈑金加工成本計算——每一站用自己的單位量
把雷射切割加折彎的鈑金件成本用十大項目拆開。切割看切割長度、沖孔看沖次、折彎看折數——平均掉的話,真正的成本動因就看不見了。
- 排版、材料利用率,以及料骨到底值多少
鈑金成本裡材料通常是最大一項,而它大半由排版決定。怎麼量實際利用率、料骨怎麼估價,以及那個會讓每張成本表都變好看的重複扣除。
成衣製造
- 成衣成本計算——加工費裡面裝了什麼,拆成十項
布料、副料、裁剪、車縫分鐘、整燙、包裝。把報給客戶的加工費拆開,讓價格談判有依據可指,不是憑感覺。
- 標準工時、分鐘單價、稼動率——決定車縫成本的三個數字
車縫成本=標準工時 × 分鐘單價 ÷ 稼動率。三個都容易算錯,而且三個錯誤全部往同一個方向走:讓成本看起來比較便宜。
- 單件用布量,以及布到底消失在哪裡
布料是成衣成本最大的一項。怎麼從排嘜取得用布量、排嘜效率量到的範圍到哪裡,以及那些永遠不會出現在排嘜上的損耗。
射出成型・橡膠
- 射出成型成本計算——射出量、成型週期、模穴數
把成型件的成本用十大項目拆開。幾乎所有東西都回到三個數字,其中任何一個算錯,動的不是一行而是整張表。
- 模具攤提與模穴數的經濟性——模具決策背後的算術
模穴數多,模具貴、單件便宜——到某個點為止。怎麼誠實地攤提模具費用,以及怎麼用數字而不是直覺去比較兩個模穴數方案。
- 粉碎料與回收價值——唯一往下扣的那一行
流道和不良品可以粉碎後再成型。怎麼誠實地估這個回收的價值,以及那個會悄悄讓成本表變好看的重複扣除。
- 橡膠成型成本——跟射出成型幾乎一樣,除了決定性的一點
硫化橡膠件的成本計算。結構跟射出成型幾乎重合,但毛邊沒辦法粉碎回用——這一點改變了哪裡值得改善。
家具製造
- 家具製造成本計算——三種材料,三個不同的單位
把櫃體或椅子的成本用十大項目拆開。難處在於板材、封邊、五金各用不同的單位採購,混成一行會讓材料費悄悄壞掉。
- 裁板利用率——最短時間內動到最多錢的數字
板材費不隨件的面積走,而是隨你開了幾張板走。怎麼從既有記錄量實際利用率,以及為什麼不能用最佳化軟體的理論值算成本。
- 龐大貨物的運費——重量還沒滿,空間先滿了
家具運費按材積不按重量,這會改變改善的價值。單台運費怎麼算,以及為什麼平板包裝在成為設計決策之前先是成本決策。
印刷與紙器
- 印刷品成本計算——按件發生的費用,和按份發生的費用
把印刷成本用十大項目拆開。這個行業「一張單發生一次」和「一份發生一次」的混雜程度最高,兩者搞混會讓小量看起來不合理地便宜。
- 調機損紙——不隨數量變動的材料費
對色過程中報廢的紙,是穿著變動成本外衣的固定成本。它放在成本表哪一行,決定了小量的價格算不算得對。
- 紙張用量——一張拼幾個,以及裁切邊值多少
紙在印刷成本裡常是最大一項,而它大半由拼版決定。怎麼量實際用量、為什麼紙張尺寸的問題要在印刷機的問題之前問,以及損紙的真實價值。
食品加工
- 食品加工成本計算——得率、批次,以及批與批之間
把加工食品的成本用十大項目拆開。原料每一站都會少一點,而清洗、換線這些「批與批之間」發生的費用屬於批次,不屬於單件。
- 加工得率是連乘出來的,不是加出來的
每一站的損失會疊乘,所以一連串小損失遠大於它們的總和。用兩個既有數字量整條鏈,以及超填藏在哪裡。
- 清洗與換線——按批計算的費用,不是按件
批次之間的清洗費用,批大批小幾乎一樣。這才是小批量真正貴的原因。怎麼計入,以及最小批量怎麼決定。
酒類
- 酒類成本計算——時間本身要花多少錢
除了原料和加工,熟成還會產生三筆費用——被綁住的資金、天使的分享,以及被佔用好幾個月的倉儲空間。
菸草製品
- 菸草製品成本——水分讓用量數字跳動
批次成本跟食品加工相同,特有的一點是:原料重量隨水分改變,所以用量要換算到共同基準,批與批之間才比得了。
電子組裝
- 電路板組裝成本計算——零件、打件點數,以及 BOM 上沒有的東西
把組裝完成的板子成本用十大項目拆開。這個行業有兩點跟別人不同:BOM 支配了成本,而機台時間跟打件點數成正比,不是跟板數。
- 幾百行的 BOM——在不改變總額的前提下合併尾端
料表大部分是加起來也幾乎不動的便宜零件。怎麼讓表保持可讀又讓總額完全一致,以及合併會毀掉你要的東西的三種情況。
- 打件點數與換線——決定 SMT 價格的兩個數字
SMT 的組裝費由打幾點和換幾次線決定。兩者都能從產線已經產出的資料量到,而且各自指向不同的改善。
半導體:封裝與測試
- 半導體後段成本計算——報廢最貴的行業
黏晶、打線、封膠、測試的成本。最貴的材料在第一站就進到產品裡,所以越後面損失的代價跳得越高。
機械製造與產業設備
- 一台機器的成本計算——從零件往上加,不是從工時往下除
把產業機械或設備的成本用十大項目拆開。自製件、外購件、外包件的算法各不相同,而供料給外包時會發生重複計算。試車也是成本的一行。
- 一次性費用——設計、治具、首台,以及要除以幾台
數量少的時候,一次性費用攤不薄。怎麼判斷哪些算一次性、除數怎麼選,以及忽略它為什麼會讓報價看起來便宜卻賠錢。
焊接與鋼構
- 焊接結構件成本計算——用焊道長度量,不是用件數
把鋼構或焊接結構件的成本用十大項目拆開。同樣重量的兩件成本差很多,是因為推動這個行業的是焊道長度和電弧時間,不是件數。
- 焊接變形——成本表上任何一行都看不到的費用
熱會讓鋼移動,而矯正它花掉的錢分散在四個費用項目裡,沒有人看得見。怎麼把它們收成一個數字,以及數量一多為什麼治具會贏過矯正。
皮件與製鞋
- 皮革製品成本——跟成衣一樣,只是材料不均勻
鞋類或皮件的成本計算。產線生產和標準工時跟成衣完全相同,但皮革一張一張形狀和品質都不同,理論用量幾乎沒有意義。
化學品製造
- 化學品成本——跟食品加工一樣,多了賣得掉的副產物
批式化學品的成本計算。得率的連鎖和批次之間的清洗跟食品加工相同,多出來的是:當一個製程產出兩種以上可賣的東西時,費用要怎麼分。
建材
- 建材成本——運費是主要項目,不是註腳
預拌混凝土、磚、預鑄件的成本計算。原料便宜又重,所以按重量與距離計算的運費常常超過加工費,配送半徑因此成為成本表的一部分。
汽車零件
- 汽車零件成本——結構跟機械製造相同,被數量翻轉
車輛零件或總成的成本計算。展開方式跟機械製造相同,但因為量產,一次性費用沉下去,取而代之浮上來的是品質成本。
航太零件
- 航太零件成本——切削比例,以及文件的費用
航太機加工件的成本計算。製程跟機械加工相同,但胚料昂貴使得切削掉的比例成為關鍵,而且證明文件是真實的費用。
醫療器材
- 醫療器材成本——加上維持生產環境的費用
醫療器材的成本計算。組裝照電子、塑膠件照射出成型,再加上一個自己的項目——生產環境的維持是按時間發生,不是按產量。
運動用品
- 運動用品成本——一個產品,底下好幾張成本表
球拍、跑鞋、自行車把塑膠、纖維、木、皮革、金屬組合在同一個產品裡。怎麼用各自產業的方法算每一個部件,再加起來而不失去細節。
再生能源設備
- 再生能源設備成本——三個產業在一個產品裡,以及要守住的邊界
模組、變流器、支架的成本計算。各部分照不同的篇,而成本表要停在出廠——現場安裝是服務,不是產品的一部分。
不以單件計算的行業
- 產品不是「一個」的行業——先決定用什麼當單位
營建、運輸、採礦、水電、農林漁業沒有可以數的產品。怎麼選一個能用的單位,以及誠實說明這個產品還做不到的地方。