“We build margin into every quotation. Sales can't just quote whatever they like. But every year-end, the number on the books is way below what I calculated. Where is the money going?”
I hear this many times a year, and the owners saying it are usually not amateurs. They came up through the technical side. They know what the materials cost, how many process steps there are, and what a skilled worker costs per hour. When they quote, they really do calculate. They are not pulling numbers out of the air. The quotation spreadsheet the sales team uses has neat columns: materials, labor hours, overhead, profit. Nothing missing.
The problem is the spreadsheet itself. Every cost figure on it is an assumption. The owner filled it in from experience three years ago, five years ago, or back when the last big customer was still around. Once entered, those numbers were never checked again. How many labor hours the shop floor actually spent on this order, how many times it was redone, how long it waited for materials, how much was lost at the subcontractor: none of what actually happened ever flowed back into the quotation spreadsheet.
I'm Johnny Yang, Brightstream Technology, Taiwan. I've led implementation teams for seven years, across 252+ projects in 32+ industries. I've seen this enough times to give you the conclusion straight. If a company's year-end margin is consistently below forecast, nine times out of ten it is not an accounting error and not sales quietly discounting. It is a broken line between the cost model used for quotations and the real cost on the shop floor. And the breaks are almost always in the same few places.
Conclusion first.
Your quotations keep getting margin wrong because the costs they use only include what has paperwork. Materials come in with a purchase receipt. Subcontracted work comes with a processing invoice. Normal production has work reports. You can count all of that. But the second round of labor and materials for rework and scrap has no paperwork. Changeover and waiting time has no paperwork. Extra material sent to subcontractors and losses at the subcontractor have no paperwork. All three are quietly absorbed into “normal cost,” and the bill arrives at year-end as missing margin. To fix it, you don't need a real-time costing system first, and you don't need to demand real-time reporting from the floor right away. Pick three representative orders that are already finished. Dig up every real labor hour, material, rework and subcontracting cost, and reconcile them once against the original quotation assumptions. The difference is your quotation bias. Turn that bias into a quotation multiplier and apply it to stop the bleeding. Then narrow it down over time with shop-floor data. You can do the first round in one weekend, at zero cost.
Half the cost on your quotation is an assumption nobody has checked
Start with a fact many owners don't want to face. A quotation is a forecast by nature, and forecasts need assumptions. Nothing wrong with that, and it can't be avoided. You can't wait until an order is done to quote it.
The real problem is not that you use assumptions. It's that your assumptions have never been reconciled against results.
A healthy quoting process looks like this. You quote with assumptions. After the order is finished, you pull back the actual costs, compare them once against the assumptions, and correct the assumptions where they're off. After a few rounds, the assumptions get closer and closer to reality. That's a closed loop.
Here is what actually happens in most factories. You quote with assumptions. After the order is finished, nothing happens. The order ships, the money comes in, sales chases the next order. The quotation gets filed, and nobody looks back at what it really cost. The assumptions sit untouched in the spreadsheet. One year, three years, five years.
Once that line breaks, two things happen.
First, your assumptions drift further off over time, and they drift in one direction. Wages go up. Materials go up. Customers' quality requirements get finer. Drawing revisions get more frequent. All of these push cost up. Almost none push it down. But your assumptions stay put. So the bias isn't random, high one time and low the next. It is a steady, persistent underestimate. That's why the year-end number is always lower than expected, and almost nobody finds it higher.
Second, and worse: you won't know which kinds of orders actually make money. All the bias gets averaged out. You have big orders and small ones, standard products and custom-built ones, jobs you've done a hundred times and jobs you're doing for the first time. Their bias varies enormously, but you only see one total at year-end. So the type of order you've been fighting hardest to win may actually be losing money. The type you find annoying and want to turn down may be the one keeping the company alive. In that situation, the harder you work and the faster you run, the deeper you sink.
I often tell owners: You don't lack the ability to calculate cost. You just don't have anything to calculate with. That's different from what most people think. They assume costs are off because the method is too simple or the formula isn't detailed enough, so they want to buy something that calculates. But the method was never the bottleneck. Data is. This is the same point as how to estimate ERP implementation cost. You think you're buying an answer. What you're actually buying is a container you have to feed with your own data.
There's an even more common case. The assumptions aren't in a spreadsheet at all. They're in the head of one veteran craftsman who can glance at a job and know whether to take it and what price will make money. The problem at these companies isn't that the assumptions were never reconciled. It's that they were never written down. I wrote about how to move that judgment into a system, using print shops as the example, in Print shop quotation automation: getting the veteran's know-how into the system.
So which data doesn't come back? Three areas. And they have one thing in common, which I'll get to.
Black hole one: rework and scrap, where the second time doesn't exist on the books
The first black hole is rework and scrap.
Run this scene in your head. An order is halfway through when QC stops it, or the operator notices something's wrong. It can't ship. It has to be redone. So materials are drawn again, the job goes back on the machine, and it's processed again. It may take no less time than the first round, because the bad batch has to be removed and cleared out and the machine reset first.
Now the question: what did this leave on your books?
On the materials side, if it goes through a formal requisition, there may be one more draw record. But on many shop floors, to keep the numbers from looking bad, people make it up from leftover stock, spare materials, or the few extra pieces from the last batch. Not one entry moves on the books.
Labor hours are even worse. The rework hours usually get reported against the original work order. To the operator, he's working on this order, isn't he? Is he supposed to open a separate “I messed up” order to report it? Nobody does that. So those hours quietly become this order's “normal labor hours.”
Result: the cost of rework is 100% real, and 100% unidentified. It isn't recorded wrong. It was never treated as something to record in the first place.
This does damage on two levels.
The first level: this order's margin gets eaten, and your calculation didn't include it. That level is fairly simple. You made less this time.
The second level is the one that really kills you: the inflated hours on this order become the reference for the next quotation. Your standard hours grow out of historical records. When rework gets mixed into normal hours, your standard hours get pushed up a little, and you don't know why. So next time you quote, you're using hours that contain an unknown amount of rework. For some orders that number is too high (the customer says you're expensive and the order goes to someone else). For others it's still too low (because that order has even more rework). From then on your quotations are a gamble, and you don't even know what you're betting on.
Scrap follows the same logic, and it's even better hidden. Scrapped items simply vanish. If they don't go through a formal scrap process, they only show up as “why doesn't the inventory reconcile?” And at most factories, inventory that doesn't reconcile is treated as a stocktake problem, not a cost problem. These are two sides of the same thing, but different people are usually responsible for them, so they never get looked at together. That's also why I keep saying inventory management and ERP are two different things. An inventory system tells you something went missing. It won't tell you how it went missing, or whose time that process took and how much.
Why the floor won't report rework on its own
In fairness: when the floor doesn't report rework, it usually isn't trying to deceive you.
For frontline workers, reporting rework has only downsides. Reporting it means admitting they messed up. It means getting reviewed, a hit to their performance rating, maybe writing a root-cause analysis. If they don't report it, as long as they redo the job and hand it in, nobody will ever know. Under that incentive structure, no amount of posters or meetings will get anyone to report voluntarily.
So if you plan to solve this, step one isn't tighter auditing. It's removing the penalty for reporting rework first. Simple to say. Doing it requires the owner to change his own messaging first: rework records are for improving the process, not for catching people. You have to say it, do it, and hold to it through the first three months when the numbers look terrible. If you can't, that line will never reconnect.
Black hole two: changeovers and waiting, where the machine stops but the wages don't
The second black hole is changeovers and waiting.
The cause is very simple: at most factories, labor hours are defined as the time the machine is running.
The machine starts, the job starts. The machine stops, the job is done. The time in between counts as labor hours and goes into cost. The definition is intuitive and easy to run, so almost every shop floor does it this way.
But beyond machine run time, there's a lot of other time you're also paying for:
- Changeovers, die changes, machine setup, trial runs, first-article checks
- Waiting for materials: the previous station isn't done, the warehouse hasn't issued them, purchasing hasn't received them
- Waiting for drawings: the customer changed the drawing and the new revision isn't out yet
- Waiting for sign-off: out-of-spec work needs a supervisor's approval, a quotation change needs sales to respond, a concession needs QA release
- Waiting for the machine: the previous order is still running, and this one is queued behind it
During this time the machine isn't running, so it isn't labor hours, so it doesn't go into cost. But the people are on site, and the wages still get paid. Rent still gets paid, utilities still get paid, depreciation keeps running.
In other words, your quotation cost only includes the productive time, but it has to cover the costs of all the time. You don't avoid paying the difference. You just throw it into the big overhead bucket and spread it back with a rough percentage.
The problem is that this allocation is the same for every order. In reality, the changeover and waiting caused by different orders can vary enormously.
Take a high-volume order with simple specs that you've done many times. One changeover runs a whole day, so the changeover cost is spread thin. There's little waiting, because materials are ready and the drawings don't change.
Now take a low-volume order with unusual specs, done for the first time. Setup and trial runs alone might eat most of a day. Then you wait for the customer to approve the first article and wait for concession sign-off. Actual run time may be less than half the prep time.
If you spread overhead across these two orders at the same rate, the result is always the same: big orders subsidize small ones, familiar orders subsidize unfamiliar ones, standard products subsidize custom-built ones.
Then you'll notice something strange. The more custom orders, rush orders, and “nobody else wants it so we'll do it” orders the company takes, the better revenue looks and the thinner margins get. Owners usually put this down to “custom work is just harder,” and keep taking it. It isn't about being harder. Your quotations never counted what custom-built work actually consumes. You're using money made on standard products to feed custom orders.
Fixing this doesn't require real-time data collection
When people hear “you need to account for changeovers and waiting,” the first reaction is often: so we need sensors on every machine and real-time collection?
No. At least not at first.
What you need isn't utilization accurate to the minute. You need a sense of the order of magnitude of waiting and changeover for each type of order. You can get that the most low-tech way possible. Pick a few finished orders. Find the team lead or operator who worked on each one. Pull out the schedules, daily reports, and group chat messages from that time, and reconstruct the timeline together. Which day the line started, when it stopped, for how long, and why.
I know this sounds unscientific. But the number you have right now is zero, and any grounded estimate beats zero. You'll also find out quickly that after two or three orders, the pattern surfaces. The bottlenecks are always in the same few places.
Black hole three: subcontracting and replacement material, where the books show one processing invoice
The third black hole is subcontracting.
The most dangerous thing about subcontracting, cost-wise, is that it looks like it has the most paperwork. You send out materials. The subcontractor processes them, sends them back, and issues you a processing invoice. You pay. The whole flow is documented. Accounting can't find fault with it.
Precisely because it looks the cleanest, it gets checked the least.
But think it through. That processing invoice only records what you paid the subcontractor. What it doesn't record includes:
First, the extra material you sent out. You almost never send exactly the right amount. You send extra, because the subcontractor will have losses, trial pieces during setup, and material eaten by jigs and fixtures. How much extra? Usually a number based on experience. Whether that number is right, whether it varies each time, whether it changed when the subcontractor changed staff: hardly anyone tracks it.
Second, whether the returned quantity is right. You send out a hundred, and ninety-five come back. Where did the other five go? At many factories, in practice, nobody chases it. The amount looks small, the subcontractor is a long-time partner, and chasing it damages the relationship. But you paid for the material in those five, and it isn't on any document.
Third, material that never comes back. Offcuts and leftovers after processing are still on your books but physically sit at the subcontractor. After two or three years this piles up into dead stock you don't even know you own. Then it turns into an unexplained loss at some stocktake.
Fourth, your own people's time spent handling subcontracting. Delivering materials, picking up parts, inspection, reconciliation, chasing deliveries, resending when quality fails. Your people do all of it, but because it isn't “production,” it doesn't show up in any work order's hours.
Fifth, replacement material. This is the one most often missed. The subcontractor says they ruined a few pieces and need more material. You send it. The cost of that material, the time you spent preparing it again, and the delivery delay it causes (which may trigger overtime or air freight): none of it goes into cost. You may not even remember how many times you've sent replacements.
Add these five up and you'll see it: between the real cost of subcontracting and the number on the processing invoice, there's a gap nobody has measured. And when you quote, you use the number on the invoice.
Why these three? Because none of them has paperwork as a natural record point
By now you can probably see what these three black holes have in common.
None of them has paperwork.
Think about the costs you get right. Materials come in with a purchase receipt, because you have to pay the supplier, so there will always be a piece of paper. Subcontract processing fees come with an invoice, because you have to pay. Wages have a payroll, because you have to pay. Normal production hours have work reports, because you need output numbers and have to answer to customers.
See it? The costs that get recorded are the ones where the event itself has a reason it must be recorded, usually because you pay someone outside or have to account to someone outside. The record is a by-product of that reason, not something anyone set out to keep.
Rework, waiting, and subcontracting losses are all internal. They don't require paying anyone extra, and nobody comes asking you for a document. The money was spent long ago (wages already paid, materials already bought). It was just used somewhere else. So nowhere in the company does a piece of paper get produced naturally.
That's why these three leak so consistently, across so many industries. It isn't that someone is lazy or careless. It's a structural blind spot in the system.
Understanding this matters, because it directly determines how you fix it:
Since these three have no natural record point, you have to design one. And whether that design works depends on one thing: whether the recording action is easy for the floor.
If your fix is “from now on, fill in a separate form for rework,” I can tell you the outcome now. Three weeks of forms, then nothing. It's an action that adds burden and gives the person filling it in nothing back. I've watched this kind of thing die too many times. The survival rate is close to zero.
A good record point is tied to an action the floor already performs. They already have to start jobs, draw materials, and hand over finished work. You add one option to that action, not one more form. For example, a “this is rework” button when starting a job, instead of a separate rework form. A “replacement” option when drawing materials, instead of filing a requisition afterward. The difference looks small, but the survival rates are worlds apart. This is the same principle as the limits you hit managing with Excel. Excel can record things fine. The problem is that recording is separate from doing the work, so it breaks the moment things get busy.
How to fix it: reconstruct three orders after the fact, don't start with real-time costing
OK. That's the diagnosis. Now the fix.
First let me block the most common wrong path: don't start by trying to do real-time costing.
I know it's tempting. Every labor hour goes into the system instantly, every piece of material is deducted instantly, and any time you open it you see what this order has cost so far. Every vendor's slide deck has that page, and it looks great.
But real-time costing has a prerequisite: shop-floor data that is real-time and accurate. Your situation right now is that the shop-floor data doesn't exist. You're building a house that runs on data, on a foundation with no data. What happens? The system goes live, the costs it calculates are just as wrong, and you conclude “this system is terrible.”
The system isn't terrible. What you're feeding it never existed in the first place. I've said this in a lot of places, often enough that some people in my industry don't much like me: when ERP costs come out wrong, most of the time it isn't the system's fault. It's the data's fault. A system is just an honest mirror. Feed it garbage and it hands the garbage back to you, fast, in real time, with charts. This is also the most common way ERP implementations fail. Not missing features, but a data source that was never connected.
So the right order is to start with after-the-fact reconstruction.
Step one: pick three representative finished orders
Not any three. They need to be representative. I suggest picking like this:
- One of your core orders: highest volume, most familiar, largest share of revenue.
- One you think “should be profitable”: the margin looked great at quotation time, and you think a few more of these would be ideal.
- One you think was “exhausting”: custom, rush, unusual specs, something went wrong along the way.
Why these three? Because you don't just want “how far off on average.” You want how the bias differs across order types. If the three orders are off by about the same amount, one overall multiplier will do. If they differ a lot, and they usually do, then your problem isn't that your quotations are too low. It's that the structure of your quotations is wrong, and you need different methods for different order types.
Pick orders that are completely closed: materials returned, payment collected, complaint period over. You want final numbers, not numbers that can still change.
Step two: dig up every cost that actually happened
This step takes the most time, and it's the only step that has real value. You dig out everything these three orders actually consumed, item by item:
Materials: not the quantity calculated at quotation, but how many times material was drawn, how much, whether there were extra draws, whether material was borrowed from elsewhere, and how much was returned. Go through requisition records, warehouse in/out records, and the operators' memories.
Labor hours: not just the numbers in the work reports. Ask “during that time, how many people were actually busy with this order?” That includes whoever prepared materials, set up machines, and did inspection, plus the sales or engineering staff going back and forth with the customer. Count all of their time.
Rework and scrap: ask the team leads and operators directly. “Was this order ever redone? How many times? Roughly how long?” They remember. Nobody has ever asked them. This question usually turns up answers that make you uncomfortable. That's normal. It means you asked the right question.
Changeovers and waiting: pull out the schedules, daily reports, even the LINE group messages from that time, and reconstruct the timeline. How many days from start to finish, how many of those were actual work, where it got stuck, and for how long.
Subcontracting: how much material went out, how much came back, how many replacements were sent, the processing fee, how many trips your own people made, and how much was rejected at inspection.
Other: any overtime for this order, any expensive freight to make the deadline, any compensation paid to the customer over quality issues.
Step three: reconcile once against the original quotation assumptions
Take everything you dug up and check it against every cell of the original quotation spreadsheet. Line by line. Don't just compare totals.
You'll see three situations:
One: the assumption was right, and so was reality. Good. Leave that line alone.
Two: the assumption was too low. This is the main finding. Note how far off it is.
Three, less common but worth attention: the assumption was too high. Some owners pad certain costs heavily to be safe, and that line ends up clearly above market, so they keep losing orders. Bring that line down, or you'll keep losing orders you should have won.
After reconciling these three orders, the difference is your quotation bias. That number is yours. It's real, and no consultant or vendor can give it to you.
Step four: turn the bias into a quotation multiplier and apply it to stop the bleeding
Once you have the bias, don't rush to build a perfect costing system. Stop the bleeding first.
The method is crude but effective: add a multiplier to the quotation spreadsheet. The multiplier is the bias you just calculated. If the bias varies a lot by order type, set a different multiplier for each order type.
At first the multiplier will be rough, a judgment call, based on only three orders. That's fine. Your alternative right now is “no adjustment at all,” and no adjustment leaves 100% of the bias in place. A rough correction in the right direction beats a precise mistake by a wide margin.
Then, and many people forget this step, write down the multiplier: where it came from, which three orders it's based on, and when it was calculated. Six months from now you will forget. Once you forget, the multiplier becomes the next generation of “assumptions nobody checked,” and you're back where you started.
Step five: start collecting shop-floor data and narrow it down
Only after you've stopped the bleeding is it time to design record points.
And the order matters: your reconstruction has already told you which area leaks the most. Add record points for that area first. Don't roll out all three at once.
If the reconstruction shows rework eats the most, the first thing to do is let the floor mark rework easily. Note: mark, not fill in a form. If waiting hurts most, start with the crudest possible stoppage-reason checkbox, five options at most. Don't design twenty reason codes on day one.
Collect data for two or three months and you can use real data to correct the multiplier. Another six months and it gets closer and closer to reality. Only then do you actually have what you need for real-time costing, and by then you'll find most of what you need is already in hand. Do it in reverse order and you spend a lot of money on a system with no data to feed it. This is the same thing I keep saying in the first step of digital transformation: first give data a place to be created, then talk about letting a system calculate it.
This article is about the three black holes at quotation time. If you want to go one level back, to a cost structure that is wrong as a whole, that's a different topic. I broke it down in Your costs may be wrong: cost blind spots in small and mid-sized manufacturers. And if the bias comes from material usage not matching, it usually points to the BOM: Companies with messy BOMs usually get the same thing wrong.
Comparison table: quotation assumptions vs. what actually happens vs. the record point to add
You can print this table, sit down with your plant manager, accountant, and sales manager, and go row by row asking “which one are we?”
| Cost item | Assumption at quotation | What actually happens on the floor | Record point to add |
|---|---|---|---|
| Direct materials | Calculated from BOM quantities plus a fixed loss rate | Loss rate varies by batch, operator, and material lot; extra draws often come from leftover stock and the books don't move | At requisition, distinguish “first draw” from “extra draw”; extra draws require a reason |
| Direct labor | Standard hours times labor rate | Standard hours already contain rework and waiting, so they're inflated | When reporting work, flag rework separately; don't merge it into the original process step's hours |
| Rework | Not included at all; assumed it “won't happen” | Rework hours often reported against the original work order; materials often made up from spare stock | Add a “rework” option when starting a job, and record the general reason category |
| Scrap | Glossed over with a rough defect rate | Doesn't go through a formal scrap process; ends up as a stocktake discrepancy | Record scrap when it happens, tied to the source work order |
| Changeovers and setup | Usually not included, or a small fixed value | For low-volume custom orders, changeover and setup time can approach or even exceed processing time | Treat changeover as its own process step with its own start and end |
| Waiting (materials, drawings, sign-off) | Not included at all | People on site, wages paid, but the time goes into no work order | Check a reason when work stops; keep it to five options or fewer |
| Subcontract processing fees | Calculated from the subcontractor's unit price | Beyond the unit price: extra material sent, leftovers never returned, replacement material | Record subcontract material out and in separately; discrepancies must be closed out |
| Subcontract management hours | Not included at all | Delivery and pickup, inspection, reconciliation, and chasing deliveries all take your own people's time | Make subcontracting work a reportable labor item |
| Rush work and overtime | Not included; treated as occasional | Custom and rush orders are far more likely to need overtime than standard orders | Overtime hours must be traceable to a specific order |
| Sales and engineering back-and-forth | Usually lumped into overhead allocation | Custom orders can need several times as many confirmation rounds as standard orders | Record “number of spec changes” at the order level as a proxy metric |
When you read this table, the point isn't to do every row. The point is to find the two or three rows where you leak the worst, and do those first. I've seen what happens when people do all ten rows at once. None of them sticks.
Honestly: a system can help, but not with the part you think
Finally, something that doesn't work in my favor.
If your conclusion after reading this is “so I should buy a system that calculates cost,” you'll most likely be disappointed again.
What a system can do is, when the data exists, calculate it correctly, quickly, and without missing anything. People can't do that. Thirty orders, each with a dozen or more process steps, plus subcontracting and rework. Doing that in Excel will drive you crazy, and you won't trust the result. For that part, a system genuinely can't be replaced.
What a system can't do is make the floor willing to admit rework. That's management, not software. A system can make “mark rework” a single tap and reduce friction to a minimum. But if the owner's first reaction to a rework record is to find someone to blame, nobody will press the button, no matter how easy it is.
So to owners who want to fix their quotation margins, I usually say this:
If you haven't even done the after-the-fact reconstruction on three orders, don't spend money yet. Do that first. It's free. One weekend plus a few conversations with the floor gets you through the first round, and it tells you something critical: whether your problem is “the data can't be calculated” or “the data doesn't exist.” The fixes are completely different. The latter means fixing management first. Only the former calls for buying a tool.
If you've done the reconstruction, you know where the leaks are, you've tried setting up record points on the floor, and you're stuck because “there's too much data to handle by hand”— that's when a system should come in. When you then look at what AI ERP can do, your questions will be very specific, and you'll be able to tell whether a vendor's answer is real. You'll already know which fields you need, which action they attach to, and which quotation the calculated numbers need to be compared against.
People who do it in reverse end up buying an expensive dashboard that reminds them every day that the data isn't coming in.
Frequently asked questions
My quotation margins are wrong. Should I switch to an ERP that calculates cost first?
In most cases, no. First figure out which kind of problem you have. Either the data exists but you can't process it (too much to organize by hand, too many complex process steps), or the data doesn't exist at all (nobody records rework, nobody records waiting, nobody tracks subcontracting losses). In the first case, a new system has an immediate effect. In the second, a new system calculates just as badly, because a system can only process what it's fed. The test is simple: pick a finished order and see whether you can piece together its real cost. If you can, but it's a lot of work, that's a tool problem. If you can't, because there are no records, that's a management problem. Solve a management problem with a tool and the money is wasted, and you'll wrongly conclude the system is bad.
I don't have a work-reporting system. Can I still do the reconstruction?
Yes. Many of the most successful first rounds are done entirely by hand. Reconstruction doesn't need a system. It needs three things: the original quotation spreadsheet for a finished order, the requisition and warehouse in/out records from that time, and the memory of the team lead or operator who worked on it. The third is the most important and the most overlooked. People on the floor remember far more detail than you'd expect. Nobody has ever formally asked them. What you do is bring them in and reconstruct the timeline with a no-blame attitude. The process itself will give you the answer. After two or three rounds, when re-interviewing people every time gets tiring, then think about a system. By then your requirements will be very clear.
The floor says rework is rare, but the numbers feel wrong to me. What do I do?
Don't question people first. Look at indirect evidence. Rework usually leaves traces not in work reports but elsewhere: an order's material usage clearly exceeds the BOM, overtime spikes during a period while output doesn't go up, an order's delivery slips for no reason visible on the schedule, a certain part keeps coming up short at stocktake. These are all shadows of rework. Point to those shadows, then go back and ask the floor “was something going on those days?” Usually it comes out. Also note that when the floor says “rare,” sometimes they mean it. In their view, redoing something that went wrong is just part of the job, not a special event. That gap gets closed with a clear definition, not with more questioning.
How should changeover and waiting time be allocated to orders fairly?
At first, don't chase fair. Chase having it at all. The crudest approach is to treat changeover as its own process step, with its own start and end times, attached directly to the work order that triggered it. For waiting, use the simplest stoppage-reason checkbox, with five options or fewer: for example, waiting for materials, waiting for drawings, waiting for sign-off, machine problem, other. With rough data on these two, you can see which types of orders generate the most changeover and waiting, then add a separate multiplier for those order types at quotation time. More refined allocation logic, by time share, by capacity used, or by batch size, is something to worry about once you have six months or more of data. Dig into it too early and you'll just get stuck.
How do I calculate subcontracting cost without missing anything?
Record at least four things, not just a processing invoice. First, the actual quantity sent out and the actual quantity received back, recorded separately, with someone responsible for closing out any discrepancy. It can't just be left hanging. Second, the number and quantity of replacement material sends. Every one needs a record, because you pay for that material. Third, leftover material sitting at the subcontractor that never came back. Reconcile it periodically, or it piles up into dead stock. Fourth, your own people's time spent on subcontracting: delivery and pickup, inspection, reconciliation, chasing deliveries. That time needs to be reportable as labor. Once you add these four, you'll find a gap between the real cost of subcontracting and the number on the invoice. That gap is what your quotations have been missing.
If this is where you're stuck
Nothing in this article requires spending money first.
Pick three finished orders, dig up the real costs, and reconcile them once against the original quotations. You can start this weekend. All it takes is you, your plant manager, and a veteran craftsman with a good memory sitting down for two or three hours. Afterward you'll know two things: how far off your quotations are, and where. Those two things alone are enough to change how you look at every order you're taking right now.
Maybe after the reconstruction the bias is big enough to keep you up at night, or the three orders differ so much you don't know how to categorize them, and you want someone to work through the logic with you. Our free tools and case library are public, and a lot of our Podcast episodes cover the implementation details of exactly this. Use them first. If they help, use them. If not, no hard feelings.
The right time to spend money is when you already know what you need and you clearly know it can't be done by hand. Until then, getting the data out matters more.




