The Denominator Problem

Eighth article in this silo, and the first with no regulator in it at all. The discipline here comes from arithmetic rather than from a statute.

The Verdict, Stated First

Five claims, in descending order of confidence.

One. The costing formula is not in dispute. Four independent sources give the same structure: setup divided by quantity, plus cycle time times rate, plus material, plus tooling.

Two. The billable hours denominator is where the rate goes wrong, and on our own arithmetic a plausible error there understates the rate by more than half.

Three. One publisher gives two different CAM-to-actual multipliers on two of its own pages, which we document.

Four. On our own arithmetic the setup curve is brutal below about 25 pieces and effectively flat above 100.

Five. And on our own arithmetic three ordinary errors compound multiplicatively, not additively, which is why the result is so much worse than any one of them.

The fifth is the reason to read this, ours. Each error alone is survivable and the three together are not.

Our Grades For These Claims

Applying the scheme this publication uses throughout. Like our trucking article, this rests on commercial sources rather than research, and the grades reflect that.

Grade A for our own arithmetic, which depends on no source and is reproducible from the formula.

Grade A for the internal contradiction we document, since we obtained both pages and quote them.

Grade C for the costing formula, which four commercial sources give consistently, and consistency across vendors is weaker evidence than a single authority.

Grade D for every dollar and percentage figure attributed to a source. All come from machining vendors, quoting platforms and estimating software companies selling into this market.

Grade D for the Canadian cost stack specifically, which comes from one machinery dealer's guide and which we could not corroborate.

A Note On Method

Everything here is checked to 29 August 2026.

We obtained eight commercial sources, of which one addresses Canadian operating costs specifically[1].

We found no Statistics Canada, industry association or academic source for machine shop rates or utilisation, and we searched for one.

Several sources are selling estimating software or instant quoting, which gives them an interest in the proposition that manual quoting is unreliable. We flag it and note that their formula is nonetheless the standard one.

One source quotes rates in euros[5] and others in unspecified or United States dollars, and we have not converted any of them.

All arithmetic is ours. The shop, its cost base, its setup and cycle times and its material cost are invented to demonstrate a structure.

This article discusses cost estimation and is not accounting or pricing advice. A shop should compute its rate from its own ledger.

The Formula Everyone Agrees On

Unusually for this silo, the sources do not conflict on the structure.

One gives "Part Cost = (Setup ÷ Qty) + (Cycle Time × Rate) + Material + Tooling"[2]. Another gives "Part Cost = Material + (Setup ÷ Batch) + (Rate × Cycle Time) + Tool Wear", with material as stock weight times dollars per pound, setup as programmer plus fixture setup time, rate as machine hourly depreciation plus overhead, and tool wear as cutting tool cost divided by tool life[4].

A third states that setup cost is fixed: programming, fixturing, tool selection and machine preparation occur once per job and are then amortized across the production quantity[7]. A fourth puts it as "Setup is a fixed cost paid per batch, not per part"[5].

Four observations, ours.

Four vendors, one formula. That agreement is worth something, though consistency among sources selling into the same market is weaker evidence than a single authority would be.

The formula is also trivially simple, which is the problem: it looks like something a shop could not get wrong.

Every term except one is a measurement. Material, cycle, setup and tooling are all observable, and the rate is a construction.

And that is where this article goes, ours. The only manufactured number in the formula is the one that multiplies the largest term.

The published rates themselves span a range too wide to be a check on anything, ours. One source gives $50 to $150 an hour[4] and another $50 to $200[8], so a shop cannot test its own rate against the literature even approximately.

The Divisor

The construction of the rate, and the assumption inside it.

A source gives it directly: "Shop Rate = (Fixed Costs + Variable Costs) / Billable Hours"[2].

The same source warns against the obvious error, stating that at 65 percent overall equipment effectiveness a shop has 1,300 billable hours rather than 2,000, and that overestimating hours underestimates the true rate[2]. A companion page repeats the same warning in almost identical terms[3]. We paraphrase rather than quote both passages, because each uses punctuation this publication does not set.

Four observations, ours.

Overall equipment effectiveness is not a shop's idea of how busy it is. It is the product of availability, performance and quality, and it strips out changeover, breakdown, slow running and scrap.

A shop with a full order book can still have low OEE, because being busy and having spindles cutting saleable parts are different states.

The number is also rarely measured in small shops, so the divisor tends to be a calendar calculation rather than an operational one.

And the sources do not tell us what OEE is typical, ours. The 65 percent figure is used illustratively in both places we found it, not asserted as a benchmark, and we treat it the same way.

Which leaves a reader with the one instruction that actually works, ours. Do not adopt 65 percent or any other published figure; count last year's cutting hours from your own records, because that number exists in every shop whether or not anyone has ever added it up.

What That Does To The Rate

Our own arithmetic on an invented shop. Every figure is ours.

Take an annual cost base of $260,000 for one machine and its share of overhead.

Divided by 2,080 hours, being 52 weeks at 40 hours, the rate is $125.00. By 2,000, a round number many shops use: $130.00. By 1,700, at 85 percent OEE: $152.94. By 1,300, at 65 percent: $200.00. By 1,000, at 50 percent: $260.00.

Four observations.

A shop quoting off 2,000 hours that actually delivers 1,300 has set its rate 53.8 percent too low, and loses $70 on every hour it sells.

The error is proportional to the whole cost base, so it does not shrink with volume or improve with experience.

It is also invisible in the gross margin on any single job, because the job was quoted and costed at the same wrong rate.

And note where the sensitivity is worst, ours. The gap between 2,080 and 1,700 hours is $28 an hour; the gap between 1,700 and 1,300 is $47. The rate accelerates as utilisation falls, so the shops least able to absorb the error are the ones making it largest.

Why It Goes Unexamined

Our own reasoning about a durable practice.

Four observations.

The rate is set once and then inherited. A shop rate tends to be established when the business is founded or when a machine is bought, and thereafter adjusted rather than rebuilt.

Adjustment moves the numerator. Shops revisit their rate when costs rise, which is a numerator event, and the denominator is not implicated by anything a cost increase does.

The denominator also has no invoice attached to it, so nothing arrives in the mail to prompt a review.

And the one thing that would prompt it is the thing least likely to be measured, ours. A shop tracking OEE would notice; a shop that is not has no signal at all.

One Publisher, Two Multipliers

A conflict of the kind this silo keeps finding, and this time it is internal.

One page states: "CAM doesn't include load/unload, deburring, inspection, or the 80% efficiency factor. Real time is 1.5-2× CAM time."[2]

A companion page from the same publisher lists using CAM time as actual time among its common estimating mistakes, and states that real floor-to-floor time is 1.3 to 1.5 times CAM time[3]. Paraphrased for the same reason.

Four observations, ours.

The ranges overlap at a single point. One says 1.5 to 2.0 and the other 1.3 to 1.5, meeting only at 1.5.

A shop taking the low end of one page quotes 1.3; a shop taking the high end of the other quotes 2.0. That is a 54 percent difference in estimated machine time for the same part.

Some of the gap may be definitional. One page mentions an 80 percent efficiency factor inside its multiplier and the other refers to floor-to-floor time, and those may not be the same measurement.

And neither page says which it means, ours, which is the actual problem. A multiplier without a stated definition cannot be applied confidently at either end.

What The CAM Gap Contains

What sits between simulated cycle time and the clock, from the sources.

One source lists load and unload, deburring and inspection as excluded from CAM output[2]. Another notes that secondary operations including deburr, quality control and packaging add 10 to 30 percent to cost[3].

Four observations, ours.

These are all human operations attached to a machine estimate, which is why a machine simulation misses them.

They also scale with piece count rather than with batch, so unlike setup they do not amortise away at volume.

That makes them the opposite of setup in a useful way. Setup punishes small batches and secondary operations punish large ones, and a shop that models only the first will misprice long runs.

And the two sources give different treatments of the same costs, ours. One folds them into a time multiplier and the other adds them as a cost percentage, and a shop applying both would double-count.

Setup And The Batch Curve

Our own arithmetic on an invented job, costed at the true $200 rate derived above.

A 45-minute setup and a 15-minute cycle, with $12.60 of material and $0.10 of tooling per part. Setup costs $150.00, charged once per batch. Everything else comes to $62.70 per part.

At a batch of 1, setup per part is $150.00 and the total is $212.70. At 5: $30.00 and $92.70. At 10: $15.00 and $77.70. At 25: $6.00 and $68.70. At 50: $3.00 and $65.70. At 100: $1.50 and $64.20. At 500: $0.30 and $63.00.

Four observations.

A single part costs 3.4 times what the same part costs at 500. Nothing about the part changed.

The setup duration is within a sourced range. One source gives 15 to 45 minutes of skilled labour for fixturing, tool loading, zeroing and first-article verification, charged once regardless of whether the order is 1 piece or 100[5].

A second source works the same arithmetic at different inputs: a 45-minute setup on 5 parts adds $12.75 per part at $85 an hour[3], and a 30-minute setup on a 5-part run adds $85 to the job[2].

And a third gives a cleaner illustration than ours, ours to acknowledge. One hour of setup at $75 is $7.50 a part on ten parts and $0.15 a part on five hundred[6].

Our Curve Against Their Premium

A check on our own arithmetic against an independently published figure. Ours.

A source states that "for prototype quantities of one to 10 parts, expect to pay three to five times the per-unit cost you'd see at production volumes", and that this is normal and unavoidable[6].

Our curve gives 3.38 times at a batch of one and 1.23 times at ten, against a production reference of 500.

Four observations.

Our single-piece figure sits inside their three-to-five range, at the bottom of it, which is a reasonable corroboration of the structure.

Our figure at ten pieces is well below their range, and we think the difference is our inputs rather than an error in either. A longer setup or a shorter cycle would raise our multiple sharply.

The comparison is worth doing precisely because it is imperfect, ours. It tells a reader the shape is right and the level depends entirely on the ratio of setup time to cycle time, which is shop-specific and part-specific.

And that ratio is the single input a shop should know for its own work. A shop whose setups are long relative to its cycles has a much steeper curve than ours, and a much bigger problem quoting small batches.

Where The Curve Goes Flat

The other end, which matters for a different decision. Ours.

Four observations.

On our figures the total moves from $68.70 at 25 pieces to $63.00 at 500, a difference of 8.3 percent across a twentyfold increase in quantity.

So volume discounting above about 100 pieces has almost no cost basis in setup terms, and a shop offering one is giving away margin rather than passing on a saving.

A source describes the same shape: per-unit cost drops steeply from one to about 100 units and then the curve flattens[6].

And it identifies where the real break comes, ours. Beyond 500 to 1,000 parts, machining begins competing with injection moulding or die casting[6], which is a process decision rather than a pricing one.

The Three Errors Compound

Our own arithmetic, and the reason this article exists.

Take a shop that does three ordinary things. It divides costs by 2,000 hours rather than 1,300. It quotes CAM time rather than floor-to-floor. And it omits setup on a small run because the job is only ten pieces.

The quote: cycle of 0.25 hours at $130 is $32.50, plus material and tooling of $12.70, plus setup of nil. Total $45.20.

The reality: cycle of 0.25 hours at a 1.5 multiplier is 0.375 hours, at $200 is $75.00, plus $12.70, plus $150 of setup across ten pieces at $15.00. Total $102.70.

Four observations.

The quote recovers 44.0 percent of cost. The loss is $57.50 a part, or $575 on the ten-piece run, before any margin.

The errors multiply rather than add, because the rate error and the time error both act on the same term before setup is added on top.

Each is individually defensible, ours. 2,000 hours is a plausible year, CAM time is a real number from a real tool, and setup on ten pieces feels like rounding.

And that is precisely why they survive. No single one of the three would fail a sanity check, and nobody performs the check on all three together.

Why The Shop Cannot See It

Our own reasoning, and it is the same shape as findings elsewhere in this silo.

Four observations.

The job ran fine. Nothing on the floor announced a costing error, because the parts were made and shipped.

The customer paid. There is no dispute, no credit note and no exception report.

And the job costing report agrees with the quote, because it values the hours at the same wrong rate the quote used. A shop comparing quoted to actual finds them close and concludes the estimate was good.

So the loss surfaces only in the annual result, ours, where it is indistinguishable from soft pricing, a bad mix or a slow year, all of which are more comfortable explanations and none of which is fixable.

The Canadian Cost Stack

The one source we found addressing Canadian operations, with its limitations flagged.

A machinery dealer's guide gives, for Canadian manufacturers: depreciation typically over 7 to 10 years representing $5 to $25 per hour; financing costs adding approximately 2 to 5 percent of machine price annually; facility overhead allocation of $5 to $15 per hour in most Canadian manufacturing regions; consumable tooling at $5 to $20 per hour; maintenance supplies averaging $2 to $5 per operating hour; and energy consumption of 20 to 40 kWh during active cutting for high-powered machining centres[1].

On labour it gives skilled CNC operators at $25 to $45 per hour, setup technicians at a premium of $30 to $50, programmers at $35 to $60, and maintenance technicians at $40 to $75[1].

Four observations, ours.

A machinery dealer, not a research body, flagged, and we could not corroborate any of these figures.

The ranges are also wide enough to be nearly unusable as benchmarks, with tooling spanning fourfold and depreciation fivefold.

What is useful is the list of categories rather than the amounts, since a shop building its own rate needs to know what to include and can price each line itself.

And one item on that list is frequently missing from shop rates entirely, ours. Financing cost or the opportunity cost of capital is a real charge against a machine and rarely appears in an hourly rate.

Fewer Setups Beats A Cheaper Machine

A counterintuitive point the sources make well.

One states that a higher five-axis rate "is offset when a part would otherwise require three or four separate 3-axis setups", each adding fixturing, alignment and re-zeroing time plus tolerance stack-up between faces, and that "a part requiring 2 setups instead of 4 costs roughly 30-40% less per part"[5].

Another puts it as: accepting a higher hourly rate may reduce overall cost if it eliminates setups and consolidates operations into fewer toolpath and tooling changes, reduces cycle time through better orientation and access, or avoids rework[7]. It adds that "hourly rates are best treated as context, not a pricing shortcut."[7]

Four observations, ours.

This inverts the usual purchasing instinct, which is to route work to the cheapest capable machine.

It follows directly from the setup arithmetic above. If setup is a fixed cost per batch, halving the number of setups halves that whole term, and at low volume that term dominates.

The tolerance point is separate and also financial. Stack-up between faces on multiple setups produces scrap, and scrap is not in the quote at all.

And the effect runs the other way at volume, ours. At 1,000 pieces the setup term is negligible and the hourly rate is nearly the whole cost, so the cheaper machine wins after all. The right answer changes with batch size.

The Costs That Arrive Later

Items the sources say are routinely omitted.

One estimates that hidden costs add roughly 25 to 40 percent to most traditional quotes, through fixture fees, waste markup, inspection surcharges and shipping premiums[5]. Another notes fixturing specifically: a simple vise setup is essentially free, a custom soft-jaw set might add $100 to $300, and a dedicated fixture for a complex part can run $500 to $2,000 or more, amortised over the run[6].

On tooling, one source warns that titanium and Inconel can burn through $50 in inserts per 100 parts and advises tracking cost per edge[2], and gives the amortisation as a $15 insert cutting 200 parts being $0.075 a part plus two to five cents for holder amortisation[2].

Four observations, ours.

The 25 to 40 percent figure comes from a company selling instant quoting, flagged, and has an obvious interest in the size of the number.

Fixturing is the item most worth pricing explicitly, because a $2,000 fixture on a 10-piece run is $200 a part and would dominate our entire worked example.

The insert arithmetic is trivially small per part and matters only on difficult materials, which is why the exotic-alloy warning is separated from the general case.

And the general lesson is about the shape of the omission, ours. Every hidden cost listed is a fixed or step cost, which means all of them behave like setup and all of them punish small batches.

Which suggests a single test rather than a checklist, ours. Ask of any cost whether it is incurred once per batch or once per part, and price the first group as setup regardless of what it is called.

Material Is Not The Material Price

A small point with a consistent direction of error.

One source gives material as stock weight times dollars per pound, with stock form and utilisation as the optimisation levers[4]. Another works it as a part machined from $12 bar stock with 40 percent material removal at $12 times 1.05 for kerf and scrap, giving $12.60[2].

Four observations, ours.

The cost is the stock, not the part. Material removed is paid for and thrown away, subject to any scrap recovery.

So a design change that removes mass does not reduce material cost unless it changes the stock size required, which is a distinction customers rarely make when asking for a lighter part.

The kerf and scrap factor in that example is 5 percent, which is a modest allowance and would be low for small parts cut from bar.

And material is the one term where the error usually runs in the shop's favour, ours, because stock is bought and counted while the other terms are estimated.

That asymmetry is worth naming explicitly, ours. The measured terms in a quote are close to right and the estimated terms carry all the risk, which is why material rarely explains a job that lost money.

What To Fix First

In order of return, ours.

Four steps.

Rebuild the denominator. Count the hours the spindle actually cut saleable parts last year and divide the cost base by that, not by the calendar.

Measure one job floor-to-floor against its CAM estimate. One measurement gives a shop its own multiplier and settles the 1.3-to-2.0 question for its own work.

Put setup on every quote as a separate line, even when it is small, because a line that exists can be argued about and a line that is absent cannot.

And price fixturing explicitly, since it is the largest of the step costs and the one most likely to exceed the entire machining charge on a short run.

The order matters more than it looks, ours. The first step is free and corrects every quote the shop will ever issue; the others correct individual jobs.

If You Advise A Shop

For our own profession. Ours.

Four points.

Ask what the shop rate is divided by. It is the one number in the calculation with no source document behind it and the one most likely to be wrong.

Do not accept a favourable quoted-to-actual comparison as evidence. Both sides of that comparison use the same rate, so it cannot detect a rate error.

Look for financing or capital cost in the hourly rate, which one source lists as a category and which is frequently absent.

And test the small-batch jobs specifically. On our own arithmetic that is where setup omission, fixturing and the rate error all land hardest at once.

One caution on how to raise it, ours. A shop owner hearing that their rate is half of what it should be will hear an accusation, and the finding lands better framed as a divisor nobody was ever asked to check.

What To Do

Divide by billable hours, not calendar hours. On our own figures the difference between 2,000 and 1,300 hours is $70 an hour on a $260,000 cost base.

Establish your own CAM-to-actual multiplier, because the published figures range from 1.3 to 2.0 and one publisher gives both.

Amortise setup across the batch on every quote, and show it as a line.

Expect a single piece to cost several times a production piece. Our arithmetic gives 3.4 times against a 500-piece reference and a source gives three to five times for one to ten pieces.

Stop discounting above about 100 pieces on setup grounds. On our figures the curve is flat there and the saving does not exist.

Price fixturing as its own line, since a dedicated fixture can run into the thousands and amortises like setup.

Consider fewer setups on a dearer machine for low volume, and the cheaper machine at high volume, because the right answer changes with batch size.

And check whether the three errors are compounding. Individually they look like rounding; together, on our own arithmetic, they produced a quote recovering 44 percent of cost.

The Limits Of This Analysis

Several caveats matter. This article discusses cost estimation and is not accounting, tax or pricing advice; a shop should build its rate from its own ledger and validate any estimate against its own routing history. This is a weakly sourced article by design of the available material. Every dollar and percentage figure attributed to a source comes from a commercial vendor: machining suppliers, instant-quoting platforms, estimating software companies and a machinery dealer. We found no Statistics Canada, industry association or academic source for machine shop rates, utilisation or setup durations, and we searched for one. Several sources sell estimating or quoting software and therefore have an interest in the proposition that manual quoting is unreliable, which is the proposition this article partly advances; we flag that rather than resolve it. Currencies are mixed and unconverted: one source quotes euros, others quote unspecified or United States dollars, and we have made no adjustment. The Canadian cost stack comes from a single machinery dealer's guide which we could not corroborate, and its ranges are wide enough to be of limited use as benchmarks. One source calls its own default values placeholders for early estimating rather than authoritative market benchmarks[3], and we would extend that caution to every figure in this article. We did not establish what a typical OEE is for a Canadian machine shop; the 65 percent figure is used illustratively by its source and by us. All arithmetic is ours: the $260,000 cost base, the 45-minute setup, the 15-minute cycle, the $12.60 material and every derived figure are invented to demonstrate a structure. Our comparison of our own batch curve against a published three-to-five-times prototype premium is a check on shape rather than on level, and our figure at ten pieces falls well below that range, which we attribute to our inputs rather than to an error in either. And our characterisation of the three errors as ordinary is our own judgement, not a finding that shops commonly make them.

Related Reading

Niche Industry Finance Per Loaded Mile, Not Per Mile The same denominator problem in a different trade, and the same absent primary source. Niche Industry Operational Finance A Policy Number, Not A Price Ontario dairy quota is capped at $24,000 per kilogram of butterfat. In March 2026, 1,908 producers bid to buy it and 18 offered to sell. That is not a... Niche Industry Finance The Lesser Of Another business where a price set once quietly governs revenue nobody re-examines.

Frequently Asked Questions

How is a shop rate calculated?

Fixed plus variable costs divided by billable hours. The numerator is documented and the denominator is usually assumed, which is why the denominator is where the error lives.

What difference does the hours assumption make?

On our own arithmetic for an invented $260,000 cost base, dividing by 2,000 hours gives $130 an hour and dividing by 1,300 gives $200. A shop using the first while achieving the second has set its rate 53.8 percent too low and loses $70 on every hour it sells.

How much longer is real machining time than CAM time?

The published figures disagree, and one publisher gives two different answers on two of its own pages: 1.5 to 2 times on one and 1.3 to 1.5 times on the other. Neither states its definition precisely. A shop should measure one job floor-to-floor and establish its own multiplier.

How much more does a single prototype part cost?

On our own arithmetic, 3.4 times a production part at a 500-piece reference, driven entirely by setup amortisation. One source gives three to five times for quantities of one to ten. The level depends on the ratio of setup time to cycle time, which is specific to the shop and the part.

Should we discount for large volumes?

Not on setup grounds above about 100 pieces. On our own figures the total moves only 8.3 percent between 25 and 500 pieces, so a volume discount there is margin given away rather than a saving passed on.

Is a more expensive machine ever cheaper?

At low volume, yes, when it eliminates setups. One source states that a part requiring two setups instead of four costs roughly 30 to 40 percent less per part. At high volume the setup term becomes negligible and the cheaper hourly rate wins, so the right answer changes with batch size.

Why does a quoting error like this survive?

Because nothing signals it. The job ran, the customer paid, and the job costing report values the hours at the same wrong rate the quote used, so quoted and actual agree. The loss appears only in the annual result, where softer pricing or a bad mix are more comfortable explanations.

References

1. Machinery dealer's guide to CNC machine operating costs for Canadian manufacturers. States that depreciation is typically calculated over 7 to 10 years for CNC equipment, representing $5 to $25 per hour depending on initial machine price and expected useful life; that financing costs add approximately 2 to 5 percent of machine price annually as interest expense or opportunity cost of capital; that facility overhead allocation for space, utilities and support systems ranges from $5 to $15 per hour in most Canadian manufacturing regions; that high-powered CNC machining centres potentially consume 20 to 40 kWh during active cutting; that consumable tooling represents $5 to $20 per hour depending on material and tool quality; that maintenance supplies average $2 to $5 per operating hour; that programming costs range from $40 to $100 per hour of programming time; and that in the Canadian context skilled CNC operators typically command $25 to $45 per hour, setup technicians $30 to $50, programmers $35 to $60 and maintenance personnel $40 to $75. Note: a machinery dealer's marketing guide, NOT a research or statistical source, flagged. The ONLY source we found addressing Canadian operating costs specifically, and we could not corroborate any of its figures. mikonmachinery.com
2. Estimating software company's guide to CNC machining cost estimation, shop rate and quoting, February 2026. Gives Shop Rate as fixed costs plus variable costs divided by billable hours, and Part Cost as setup divided by quantity plus cycle time times rate plus material plus tooling; states that typical fully loaded shop rates include operator, overhead and approximately 15 to 20 percent profit margin, with aerospace and medical shops charging 20 to 40 percent premiums, and that its rate figures come from industry surveys and job shop owner reports; that a part machined from $12 bar stock with 40 percent material removal is costed at $12 times 1.05 for kerf and scrap, giving $12.60 per part; that a $15 insert cutting 200 parts is $0.075 per part plus $0.02 to $0.05 for holder amortisation; that a 30-minute setup on a 5-part run adds $85 to the job while on a 500-part run it is negligible; that CAM does not include load and unload, deburring, inspection or the 80 percent efficiency factor, and real time is 1.5 to 2 times CAM time; that titanium and Inconel can burn through $50 in inserts per 100 parts; and that if OEE is 65 percent a shop has 1,300 billable hours, not 2,000, and overestimating hours underestimates the true rate. Note: a commercial estimating software company, NOT a research source, flagged, with an interest in the unreliability of manual quoting. Our source for the shop rate formula and the OEE warning. NOTE THE CONTRADICTION with ref 3 on the CAM multiplier, documented in this article. cncoptimization.com
3. Same publisher's CNC cost estimator calculator page, listing common estimating mistakes: not counting setup time, a 45-minute setup on 5 parts adding $12.75 per part at $85 per hour; using CAM time as actual time, when real floor-to-floor is 1.3 to 1.5 times CAM time; ignoring tooling wear, titanium and Inconel burning $50 or more in inserts per 100 parts; overestimating billable hours, at 65 percent OEE having 1,300 hours not 2,000; and forgetting secondary operations, deburr, quality control and packaging adding 10 to 30 percent to cost. The page states that its default values are placeholders for early estimating rather than authoritative 2026 market benchmarks, and that users should replace them with their own shop ledger, payroll burden, material quotes and routing history before sending a customer quote, with machine utilisation calibrated from their own data. Note: the SAME PUBLISHER as ref 2, giving a DIFFERENT CAM multiplier of 1.3 to 1.5 against ref 2's 1.5 to 2. This internal contradiction is documented in the article. Its own disclaimer about placeholder values is the most candid statement in any source we obtained. cncoptimization.com
4. Design for manufacture and assembly software company's machining cost estimator resource, March 2026, giving Part Cost as material plus setup divided by batch plus rate times cycle time plus tool wear, where material is stock weight times dollars per pound, setup is programmer plus fixture setup time, rate is machine hourly depreciation plus overhead, cycle time is roughing plus finishing passes, and tool wear is cutting tool cost divided by tool life; and stating that machine rate includes depreciation, energy, maintenance, floor space and planning, typically $50 to $150 per hour, that setup cost reduces per part as batch size increases and is fixed by design choice, and that material cost is optimised through stock form and utilisation. Note: a commercial software company, NOT a research source, flagged. Cited for the costing formula, which agrees with ref 2. dfma.com
5. Manufacturing marketplace's buyer's guide to CNC manufacturing cost calculation, June 2026. States that billing rates typically sit at 70 to 120 euros per hour for 3-axis milling, 75 to 125 for CNC turning and 150 to 250 for 5-axis machining; that the higher 5-axis rate is offset when a part would otherwise require three or four separate 3-axis setups, each adding fixturing, alignment and re-zeroing time plus tolerance stack-up between faces; that setup is a fixed cost paid per batch, not per part, taking 15 to 45 minutes of skilled labour for fixturing, tool loading, zeroing and first-article verification, charged once regardless of whether the order is 1 piece or 100; that a part requiring 2 setups instead of 4 costs roughly 30 to 40 percent less per part; that four levers move price, being material machinability, tolerance band, geometric complexity and batch size; that hidden costs add roughly 25 to 40 percent to most traditional quotes through fixture fees, waste markup, inspection surcharges and shipping premiums; and that requesting three quotes routinely produces price spreads of 2 to 3 times. Note: a commercial instant-quoting marketplace, NOT a research source, flagged, with a direct interest in the unreliability of traditional quoting. RATES ARE QUOTED IN EUROS and we have not converted them. makerverse.com
6. Machining supplier's guide to CNC machining cost, price factors and hourly rates, May 2026. States that if setup takes one hour at $75 per hour that cost spreads across the batch, being $7.50 per part on ten parts and $0.15 per part on five hundred; that for low-volume orders setup can account for 30 to 50 percent of per-unit cost; that a simple vise setup is essentially free, a custom soft-jaw set might add $100 to $300, and a dedicated fixture for a complex part can run $500 to $2,000 or more, amortised over the run; that anodising might add $0.50 to $5 per part; that per-unit cost drops steeply from one to about 100 units and then the curve flattens, stabilising at 50 to 100 parts as setup becomes a negligible fraction; that beyond 500 to 1,000 parts machining begins competing with injection moulding or die casting; and that for prototype quantities of one to 10 parts one should expect to pay three to five times the per-unit cost seen at production volumes, which is normal and unavoidable. Note: a machining supplier's marketing content, NOT a research source, flagged. Our source for the batch curve shape and the prototype premium against which we checked our own arithmetic. xtjcnc.com
7. Manufacturing platform's article on custom CNC machining costs, June 2026, stating that CNC quoting is extracted from a structured combination of fixed and variable cost drivers layered with risk-assessment based overhead; that setup cost is fixed, programming, fixturing, tool selection and machine preparation occurring once per job and then amortised across the production quantity, which is why unit price drops sharply as volume increases; that cycle time is highly variable and a direct result of part complexity, with geometry driving toolpath length and tool changes, elevated surface finish requirements increasing tool passes and reducing feed rates, and material characteristics affecting removal rates, chip load, heat and tool wear; that accepting a higher hourly rate may reduce overall cost if it eliminates setups and consolidates operations, reduces cycle time through improved orientation and access, or avoids rework; and that hourly rates are best treated as context, not a pricing shortcut. Note: a commercial manufacturing platform, NOT a research source, flagged. jiga.io
8. Precision machining company's article on CNC machining cost per hour, June 2026, stating that hourly charges for CNC machining typically range between $50 and $200 depending on machine type, part complexity, material machinability and local operating costs; that the hourly rate captures the total cost of operating a CNC machine tool for an hour; that equipment specifications drive the rate base, with average three-axis machine tools and high-end five-axis multi-process machining centres differing significantly; and that savings are possible through part design optimisation, material selection and planning production batches ahead of time. The article states that its advice comes from the company's own experience. Note: a machining company's marketing content, NOT a research source, flagged. Currency unspecified in part of the article and given as USD elsewhere. Cited only for the breadth of the published rate range. lsrpf.com

This article discusses cost estimation and is not accounting, tax or pricing advice. Every figure attributed to a source comes from a commercial vendor; no statistical, association or academic source was found. Several sources sell estimating or quoting software and have an interest in the unreliability of manual quoting. Currencies are mixed and unconverted. All arithmetic is the authors' own and every shop figure is invented to demonstrate a structure.