Precision Stamping Tolerances: What ±0.05 mm Really Buys You

Every stamped part drawing carries one number that decides more about the final price than the material callout, the finish, or the annual volume: the tolerance block. Buyers who understand what those numbers cost, and what they buy, routinely save 20-30 percent on precision programs without giving up a single functional dimension. Buyers who copy tolerances from an old drawing pay for capability they never use, or worse, discover at first article that the part cannot be made at the number on the print. This guide explains the tolerance tiers that show up on real RFQs, what each one demands from the press, the die, the measurement system, and the SPC program, and how to specify precision that your supplier can actually hold for the whole run rather than just on the first article.

It is written from the floor of a precision stamping plant that runs 21 presses from 25 to 110 tons with positioning accuracy of ±0.005 mm, stamping material from 0.05 to 3.0 mm thick on strip up to 650 mm wide, and verifying the output with CMM, optical measurement, and in-line vision under IATF 16949:2016. The numbers in the tables below are either published standards, published material data, or capability levels this plant and its tool room hold every day. Use them to sanity-check the next quote on your desk.

The First Number on Your Drawing Decides Your Margin

Specify a tolerance tighter than the part needs and you pay for capability you never use. Loosen it past what the mating feature can swallow and the assembly fails on the customer's line, not yours - which is worse, because the failure shows up three months and eight thousand kilometres away. And somewhere between those two mistakes sits the quiet killer: measurement error. A ±0.05 mm tolerance backed by a ±0.04 mm measurement uncertainty is not a specification, it is a coin flip dressed up as engineering. After thirty years of quoting precision stampings, I can tell you the tolerance block is where most of the cost and most of the risk are actually born. This article is about what the numbers mean, what they cost, and what ±0.05 mm really buys you on a production press.

Three Tiers That Show Up on Every RFQ

When buyers talk about "precision stamping," they usually mean one of three bands. The numbers stay remarkably consistent across the industry because they track the physical limits of the process, not the marketing department.

• General tolerance, ±0.1 mm. The default for non-critical features: holes that clear a fastener, tabs that locate loosely, edges that only need to look straight. Held comfortably on a well-maintained progressive die running mild steel or stainless.
• Precision tolerance, ±0.05 mm. The workhorse band for functional dimensions on connector terminals, sensor brackets, and medical device clips. This is where statistical process control stops being optional and starts earning its keep.
• Ultra-precision tolerance, ±0.02 mm. The territory of carbide tooling, climate-controlled cells, and in-process gauging. Phosphor bronze contacts, optical housings, parts that mate with molded plastic at sub-tenth levels.

To put those against a published standard, ISO 2768 fine class (f) assigns ±0.05 mm to linear dimensions between 0.5 mm and 3 mm, and ±0.1 mm between 3 mm and 6 mm. So a ±0.05 mm callout is roughly the boundary where "general tolerance" stops and "you must actively control this" begins. Push to ±0.02 mm and you have left ISO 2768 behind entirely - that number lives in the IT grade world of ISO 286, around IT7 to IT8 for small bores, where you need a calibrated gage and a measured process, not a caliper and hope.

Tolerance Tier	Typical Callout	Die Construction	Measurement Method	Typical Parts
General	±0.1 mm	Standard tool steel dies	Caliper, fixture gages, go/no-go	brackets, covers, spacers, clearance holes
Precision	±0.05 mm	Hardened tool steel, tight clearances, pilot registration	CMM, optical comparator, digital mics with SPC	connector terminals, sensor clips, locating holes
Ultra-precision	±0.02 mm	Carbide inserts, restrike stations, in-die gauging	Sub-micron CMM, vision systems, in-process probes	contact beams, optical housings, lead frame features

One more distinction belongs here. A tolerance on a blanked edge, where the punch cuts through the strip, is a different animal from a tolerance on a formed surface, where springback is in play, and both differ from a tolerance on a hole position, which depends on strip feed and pilot registration. The table in the tolerance-by-feature section below breaks down what each feature type actually costs to hold.

What ±0.05 mm Buys You in Capability: Cpk

Tolerance is a number on paper. Capability is what your press actually delivers, hour after hour, across every operator and every shift. The bridge between them is Cpk - the process capability index. The formula is simple: Cpk = min(USL − μ, μ − LSL) / (3σ), where USL and LSL are the upper and lower spec limits, μ is the process mean, and σ is the standard deviation of the parts coming off the line.

Here is the part buyers miss. A tolerance does not become "capable" until the spread of real parts fits inside it with room to spare. The benchmarks that matter on the floor:

• Cpk ≥ 1.33 is the automotive and medical floor. It is a 4-sigma process. The expected defect rate is about 63 parts per million - roughly one bad part in every 16,000.
• Cpk ≥ 1.67 is the target for critical safety and high-reliability programs. That is a 5-sigma process, around 0.57 parts per million, or one in 1.7 million.

Now do the arithmetic on a ±0.05 mm feature. The spec window is 0.10 mm wide. For Cpk 1.33 you need the natural process spread (±4σ, or 8σ total) to fit inside that window, so your real-world σ must be at or below 0.0125 mm. For Cpk 1.67 the same window demands σ ≤ 0.010 mm. That is the trade nobody explains on the quote: a ±0.05 mm tolerance is not five hundredths of a millimetre of slack, it is a demand that your process variation stay under roughly a hundredth of a millimetre, on every shift, in summer and winter. Hold that and the tolerance is cheap. Miss it and you are scraping parts or sorting them by hand at the end of the line.

Two more indices deserve a sentence because they show up on capability studies. Cp measures spread only - it ignores where the process is centered - while Cpk punishes the process for drifting off target. A process can look wonderful on Cp and still ship out-of-spec parts if the mean has walked toward one limit. When you review a supplier's study, look at Cpk, and look at the gap between Cp and Cpk: a large gap is the signature of a process that is running off-center and will need frequent adjustment.

Cpk Target	Sigma Level	Defects per Million	Required σ for ±0.05 mm	When to Use It
1.00	3σ	~2,700	0.0167 mm	non-critical commercial features
1.33	4σ	~63	0.0125 mm	automotive and medical floor
1.67	5σ	~0.57	0.0100 mm	safety-critical, high-reliability features

The Hidden Enemy: Measurement Uncertainty and GR&R

You cannot improve what you cannot measure, and you cannot trust a measurement whose error is a third of the tolerance. This is where GR&R - Gage Repeatability and Reproducibility - comes in. It is a study that splits your total measurement variation into what the gage itself contributes (repeatability) and what different operators contribute (reproducibility).

The AIAG Measurement Systems Analysis guideline is the one auditors cite:

• %GR&R below 10% - the measurement system is acceptable.
• %GR&R between 10% and 30% - conditionally acceptable, depending on the importance of the characteristic and the cost of improving the gage.
• %GR&R above 30% - unacceptable. The gage is telling you more about itself than about the part.

Pair that with the measurement uncertainty rule of thumb: your gage resolution and uncertainty should be no more than about one-tenth of the tolerance you are trying to hold. On a ±0.05 mm feature that means you want a measurement uncertainty budget under roughly ±0.005 mm, which puts you into sub-micron digital metrology, optical comparators, or pneumatic gaging - not a hand caliper reading 0.01 mm gradations. I have watched two qualified inspectors disagree by more than the tolerance on a ±0.02 mm bore using the wrong gage. The part was fine. The measurement was not. That is the coin flip I opened with, and it is why first-article approval and running capability have to use the same instrument.

Feature Tolerance	Required Uncertainty Budget	Instruments That Qualify	Instruments That Do Not
±0.1 mm	≤ 0.01 mm	digital calipers, micrometers, fixture gages	steel rules, tape measures
±0.05 mm	≤ 0.005 mm	CMM, optical comparator, digital indicators	standard calipers at the edge of their resolution
±0.02 mm	≤ 0.002 mm	sub-micron CMM, vision systems, air gages	hand calipers, non-temperature-controlled measurements

There is also a temperature rule that most buyers never think about. A 100 mm steel part changes length by about 0.0012 mm for every degree Celsius of temperature change; aluminum moves about twice as fast. If the inspection room runs at 25°C and the press shop at 35°C, a 100 mm aluminum dimension can legitimately differ by 0.02 mm between the two locations with no defect at all. Precision programs should state the inspection temperature - 20°C is the ISO standard - and the gage and the part should soak to the same temperature before a tight dimension is judged. This is not a theoretical nicety; it is why automotive PPAP labs insist on climate-controlled inspection rooms for ±0.02 mm features.

Why Precision Drifts: Die Wear Compensation

A die does not stamp the same part on hit number one and hit number one million. Tool steel wears. Carbide wears slower, but it still wears. As the punch and die radii and clearances change, the cut-edge geometry shifts and dimensions creep - usually the hole gets smaller as the punch shrinks, and the blank grows as the die wears open.

The numbers are not small. A typical progressive die in tool steel runs a few hundred thousand to a couple of million hits before a major refurbish; carbide tooling pushes past a million hits more reliably and holds its edge far longer. Over that life the dimensional drift can be several times the per-part tolerance if nobody is watching. The fix is die wear compensation, and there are three ways shops actually do it:

• Predictive shimming. Build the die with adjustable shut height or removable shims so the engineer can recover lost dimension as the tool beds in, rather than discovering the drift at first article inspection.
• Compensated punch and die sets. Buy the punch a few microns oversize and the die a few microns undersize at the start, banking against the known wear direction so the part rides the center of the tolerance band for most of the run.
• In-process gauging. Put a probe or vision system at the exit of the press and feed corrections back to the die settings automatically. This is what makes Cpk 1.67 sustainable on a long automotive run, because it catches the drift before it becomes scrap instead of after.

The buyer's takeaway: ask how the supplier plans to hold the tolerance at the end of the run, not just at first article. A ±0.02 mm part that is ±0.02 mm on part one and ±0.06 mm on part 500,000 has failed you even though the certificate looked perfect on the day it shipped.

Die maintenance discipline is the companion to wear compensation. Punches and dies get sharpened on a schedule tracked by stroke count, not by feel: a burr-height limit is set at tool qualification, burr is checked at defined intervals, and the die is pulled for maintenance when the limit is reached. Ask the supplier for the die maintenance schedule and the burr-height acceptance limit, and check whether the pressroom logs strokes per die. These are the quiet details that separate a process from a run of luck.

Why the Press Lies: Material Springback

Stamping is not just cutting. Forming bends metal past its yield point and the material does not stay where you put it - it springs back. For a simple bend in mild steel the springback angle might be one or two degrees; in high-strength steel or precipitation-hardened alloys it can run five degrees or more, and the straightening of a formed wall can throw a location dimension off by a percentage of the material thickness that no die drawing predicted.

Springback compensation is the craft of bending the part past where you want it so that it returns to the target. Two techniques carry most of the load:

• Over-bend. Set the forming station a few degrees past the design angle, tuned against the actual material's springback curve measured on the floor, not in a textbook. The number that matters is the one off your own press with your own coil, not a table value.
• Restrike. Add a second station that re-forms the feature against a hardened pad after the initial spring, locking the dimension that the first hit could not hold.

The material matters as much as the die. Phosphor bronze and stainless spring back harder than cold-rolled steel; thicker stock springs back more than thin. So a ±0.05 mm location on a formed feature in 0.8 mm stainless is a different engineering problem than the same number on a flat blanked edge, and the quote should reflect that difference. Springback is the reason a tolerance on a bent part is a material specification as much as a geometric one - change the temper of the coil and you change the number you can hold.

Material	Typical Springback, 90° Bend	Compensation Approach
Mild steel (SPCC/DC01)	0.5-1°	small over-bend, punch bottoming
HSLA / DP steels	1-2°	over-bend plus restrike
Stainless 301/304 full-hard	3-8°	over-bend tuned to coil, restrike stations
Phosphor bronze C5191/C5210	2-5°	over-bend plus bottoming; temper locked in purchase spec
Beryllium copper C17200	1-3° after aging	age before forming or compensate per heat lot
Aluminum 5052	2-5°	over-bend plus coining; expect more die tryout

Coil temper variation is the quiet variable under all of this. A spring part stamped from a coil at the soft end of a temper band behaves differently from one stamped from a coil at the hard end, even though both carry the same mill certificate. Precision programs lock the temper band into the purchase spec and verify hardness on incoming coils. This is the difference between a die that holds ±0.05 mm for two years and one that drifts every time the coil changes.

Tolerance by Feature Type: What the Process Actually Holds

The same ±0.05 mm callout means very different things on different features. The table below is the one our quoting engineers work from: it separates what the strip, the die, and the material can deliver on each feature class, and it is the honest answer to "how tight can you stamp this?"

Feature Class	Commercial	Precision	What Moves the Number
Punched hole position (hole to hole)	±0.10 mm	±0.05 mm	pilot registration, strip feed accuracy, die alignment
Pierced hole diameter	±0.10 mm	±0.03 mm	punch clearance, punch sharpness, material hardness
Formed (bent) dimensions	±0.15 mm	±0.05 mm	springback compensation, temper variation, restrike
Overall length of blanked profile	±0.20 mm	±0.05 mm	die wear, strip width tolerance, temperature
Critical contact features	±0.05 mm	±0.02 mm	carbide inserts, restrike stations, SPC
Flatness	0.5 mm / 100 mm	0.1 mm / 100 mm	strip stress, die balance, leveling or coining after forming
Burr height	≤ 10% of thickness	≤ 5% of thickness	punch sharpness, die clearance, sharpening interval

Read the table the way a die designer would. Hole position is bought with tooling accuracy and strip control - it is the cheapest precision on the list. Formed dimensions are bought with springback engineering - they cost tryout time. Flatness is bought with secondary stations - coining or leveling. Burr is bought with maintenance discipline. When a buyer tightens a callout, the die builder immediately knows which of those four budgets just grew.

What Precision Costs: Tooling, Inspection, and Speed

Precision is not a flat surcharge; it compounds. Moving a feature from ±0.1 mm to ±0.05 mm changes how the die is built, how fast it can run, how often it is inspected, and how much scrap the program tolerates. The table below is the cost ladder we see on real programs.

Cost Driver	±0.1 mm	±0.05 mm	±0.02 mm
Die construction	standard tool steel	hardened steel, tighter clearances	carbide inserts, restrike stations, precision guides
Relative tooling cost	1.0×	1.3-1.5×	1.8-2.5×
Press speed	full rated SPM	80-90% of rated SPM	often derated, sometimes special cells
Inspection	sampled, fixture gages	SPC sampling, CMM on criticals	in-process gauging, 100% vision on criticals
Typical piece-price uplift	baseline	+5-15%	+15-30%

None of this says precision is bad. It says precision is a resource to spend where the function demands it and to withhold where it does not. The single most profitable act in a stamping RFQ review is deleting tight tolerances from features that nothing mates with. Every ±0.05 mm callout you can honestly relax to ±0.1 mm is a permanent reduction in the piece price, the inspection cost, and the scrap rate - not a one-time saving.

Material Thickness and Temper: The Hidden Variables

Two variables inside the coil decide more about achievable tolerance than anything on the drawing: thickness tolerance and temper. A coil ordered as 0.8 mm ships with a mill tolerance - for precision programs the strip is specified at half-standard thickness tolerance, because every micron of incoming thickness variation lands directly on the formed dimension and on the bend force the press must deliver. Spring force, as any spring designer knows, scales with the cube of thickness: a 0.05 mm swing on a 0.4 mm spring beam changes its force by double-digit percentages even though the dimension looks trivial on a caliper.

The same logic applies to hole piercing. Thicker stock pushes the minimum hole diameter up (punch fragility) and widens the clearance window the punch sees. Material that varies in thickness across the width of the strip feeds the die differently and distorts the blanked edge. For parts with ±0.02 mm features, the material spec should name the thickness tolerance, the temper band, and the surface condition - and the incoming inspection should verify them before the coil ever reaches the press.

Thermal Environment: The Tolerance You Cannot See

Presses generate heat. A high-speed progressive die running at 200-300 strokes per minute heats the tooling, the strip, and the lubricant, and the die grows. A die that was set at 25°C in the morning can be running 10-15°C hotter by mid-shift, and the thermal growth of the tool steel alone moves critical dimensions. Shops that hold ±0.02 mm features run temperature control on the die - cooling channels, thermal compensation in the die design, or simply slower, steadier press speeds - and they measure the part at a defined temperature. If a supplier tells you they hold ±0.02 mm and you visit to find no temperature control anywhere in the cell, you have learned something important about their capability claims.

For the buyer the lesson is simpler: specify the inspection temperature, ship and store precision parts in a way that does not distort them, and remember that a dimensional argument between two plants 500 km apart may be a temperature difference, not a defect. A part measured at 35°C in a tropical warehouse and 20°C in a northern lab can differ by more than a ±0.02 mm tolerance band with no one at fault.

Tolerance Stack-Up: When Five Loose Features Beat One Tight One

The most expensive mistake in tolerance specification is tightening one dimension to compensate for a stack-up that a datum scheme would solve for free. Consider a bracket with three holes that must line up with three studs. Buyers often put ±0.05 mm on each hole position, hoping the pattern lands. The cheaper and better answer is one datum scheme: locate the pattern from a single datum feature, hold the datum feature to ±0.05 mm, and let the individual holes run ±0.1 mm relative to the pattern. The assembly outcome is identical, and the tooling and inspection cost collapse.

This is what GD&T exists for, and it is why a drawing with a proper datum reference frame is cheaper to stamp than a drawing that tolerances every feature independently from the paper edge. If your drawing lacks datums, ask a stamping engineer to review it before tooling - the review costs a phone call and routinely saves five figures on die cost. Our ten DFM rules before tooling walk through the same discipline from the design side.

SPC and the Long Run: Capability Is a Process, Not a Certificate

A Cpk number on a first-article report is a photograph; a process is a movie. What matters on a program running 500,000 parts a year is whether the capability holds across shifts, across coil lots, and across seasons. That takes statistical process control: measured features plotted on control charts, rules that trigger action when the process drifts before it produces scrap, and a change-control discipline that re-qualifies the die after maintenance, after coil changes, and after any die revision.

Ask the supplier three questions. First: which features are on the control plan, and how are they sampled? Second: what happens when a control chart signals - is there a written reaction plan, or does someone hope the next sample is fine? Third: how often is the die re-qualified, and what triggers a full capability study? The supplier that answers all three with documents, not anecdotes, is the one whose ±0.05 mm will still be ±0.05 mm on part 400,000. The quality system behind that discipline is described on our quality page, and it is the same system the IATF 16949:2016 audit reviews. For the broader picture of what the certification means for your program, see our IATF 16949 guide.

Supplier Audit: Seven Questions That Separate Real Precision from Marketing

Anyone can quote ±0.05 mm. Far fewer can hold it for a year of production. These are the questions we would ask a supplier before committing a precision program - and the answers you should demand:

• Show the press positioning spec. A stated ±0.005 mm positioning accuracy tells you the machine can register the strip; ask for it in writing, not as a brochure claim.
• Show the GR&R study for the gages you will inspect my parts with. If the gage cannot measure the tolerance, the certificate is fiction.
• Show the capability study on a part similar to mine. The number on the first article is a snapshot; the study over multiple lots is the movie.
• Describe the die wear plan. Shims, compensated punches, in-process gauging - which of these exists, and who owns the decision to pull the die?
• Describe the coil control. Thickness tolerance, temper band, incoming verification - precision dies fail quietly on bad coils.
• Show the temperature control. For ±0.02 mm features, climate control and defined inspection temperature are not optional.
• Ask for the defect history on similar programs. Scrap rate, customer returns, and the root causes behind them - this is the number that predicts your program's future.

FAQ

What is the cheapest tolerance grade for a stamped part? Whatever the mating features will actually forgive. For most non-locating features, ±0.1 mm on blanked dimensions and ±0.1 mm on hole position costs the least to hold and inspect. Tighten only where function demands it.

Can a progressive die really hold ±0.05 mm on hole position? Yes, with pilot registration and a well-maintained die - this is routine on connector and terminal programs. The same number on a formed dimension is harder, because springback is in play.

Why did my supplier quote more for the same part after I tightened one hole? Because one tight feature changes the die construction, the inspection plan, and often the press speed. Tolerance cost is not linear and not additive per feature - it compounds on the critical features.

What is a realistic burr height? Under 10% of material thickness is the commercial norm; under 5% is the precision grade. Burr is controlled by punch sharpness and die clearance, which makes it a maintenance metric more than a design one.

How much does measurement uncertainty matter in practice? It is the difference between a tolerance and a gamble. If your gage uncertainty is more than about a tenth of the tolerance, the inspection result is noise. This is why ±0.02 mm features require sub-micron metrology.

Specify for Function, Not for Fear

The most expensive tolerance on your drawing is the one you copied from a sibling part because you were not sure. Every step tighter than the assembly needs adds cost: better tooling, slower press speed, more inspection, climate control, and scrap. The cheapest part to make is the one toleranced exactly to what the next component will forgive.

My rule for an RFQ review is blunt. For every dimension under ±0.1 mm, name the mating feature and the failure mode if it is missed. If you cannot, the tolerance is a guess, and guesses get quoted as risk. Tighten where the function demands it, loosen where it does not, and put your measurement plan and your die-wear strategy where the tight numbers live. For the feature-level rules that sit next to this one, our stamping defect root-cause guide covers what actually goes wrong when the process window is missed.

Get a Tolerance Review Before You Quote the Whole Program

If you have a drawing with a column of ±0.05 mm and ±0.02 mm callouts and you are not certain which ones are load-bearing, send it over. We will run a tolerance rationalization against your assembly, tell you which dimensions need Cpk 1.67 and in-process gauging and which can relax to ±0.1 mm without anyone noticing, and quote the tooling to hold the ones that matter. Thirty years of stamping small precision parts teaches you that the right tolerance on the right feature is what separates a part that costs too much from a part that simply works. Request a quote and attach your drawings, and we will show you what ±0.05 mm actually buys on your program.