How to Read a CMM Report: Dimensional Data for Mold Buyers

A connector housing came back from T1 trial with a true position reading of 0.12 mm against a 0.05 mm tolerance. The buyer flagged the mold as failed. The CMM report showed PASS on every single feature. The datum plane that the whole layout hung on had never been established, and the software had defaulted to a best-fit alignment that averaged the error away. The report was technically printed, numerically wrong, and nearly cost a good tool its approval.

A CMM, or coordinate measuring machine, touches or scans a part with a probe and reports where features actually sit in space. That report is the objective record of what your trial parts measured, which makes it the evidence behind every mold acceptance decision. Mold buyers, purchasing engineers, and project engineers sign off on parts they never fully read.

This guide covers how to read a dimensional report on molded trial parts and mold steel, from the header block to the Cpk line. It explains what each column means, why the datum setup comes first, how position and form tolerances are judged, and which patterns justify pushing back. Values like Cpk thresholds and conditioning times follow common industry practice; confirm the exact requirement on your own drawing and supplier spec.

The Snapshot

• Every CMM report carries five columns per feature: nominal, actual, deviation, tolerance, and a PASS or FAIL flag. Read the columns before you read the flag.
• The datum setup decides the whole report. A best-fit alignment or an unestablished datum plane can make every feature pass for the wrong reason.
• True position is reported as a diameter, which means the tolerance zone is a cylinder and the reported value is twice the radial offset of the feature center.
• Form values like flatness and roundness are zone widths, not radial errors. A 0.02 mm flatness reading means the surface fits between planes 0.02 mm apart.
• Capability needs a process, not one part. Cpk of 1.33 is a common minimum and 1.67 is typical for critical features, and Cpk collapses fast when the mean shifts.
• Measurement conditions are part of the data. Parts condition at 23 ±1 °C for hours, and tool steel grows about 11 to 12 µm per meter per kelvin.

What a CMM Report Contains

Open any professional CMM report and the first page is a header block, not results. It lists the part number and revision, the material and grade, the cavity number, the trial stage, the operator, the date, and the machine ID. A complete header also records the CMM model, its calibration date, the probe configuration, and the measurement temperature.

Treat a report with a blank operator, a missing date, or no machine ID as incomplete. You cannot verify a dimensional claim without knowing who measured it, when, and on what instrument. DieStrike anchors every trial report to a named CMM with a current calibration certificate, because a mold shipped on unverifiable numbers is a mold shipped on faith.

The heart of the report is the feature results table. Each row covers one called-out dimension or geometric tolerance from the drawing, and each row carries five columns. Nominal is the target value from the CAD model or print. Actual is what the machine measured. Deviation is actual minus nominal. Tolerance is the allowed band or zone. The flag column says PASS or FAIL, often with a colored dot.

Do not read the flag first. Software flags are only as honest as the tolerance value typed into the program, and that value is only as honest as the drawing revision loaded. A report full of green PASS rows against the wrong revision is still a failed report.

← swipe to scroll →

Table: Standard sections of a CMM dimensional report and what a mold buyer should verify in each.

Units matter as much as columns. Mold work talks in millimeters, and precision features get tolerances in thousandths. A deviation written as 0.003 mm is three micrometers, not three hundredths. Confirm the unit declaration on every report, because a decimal place error turns a passing mold into a rejected one overnight.

Read the Datums First

Datums are the skeleton of the drawing, and the CMM report inherits them. Datum A is usually the primary plane, datum B the secondary, and datum C the tertiary. Together they form the datum reference frame, the coordinate system that every position, orientation, and profile callout is measured against.

The machine establishes that frame by probing real surfaces on your part. The primary plane datum A gets probed over its area, datum B constrains rotation, and datum C locks the last direction. If the part sits on a warped datum plane, the whole frame tilts, and every true position in the report shifts with it.

So the datum section is the first thing to read, not the results. Verify that datum features were measured as real surfaces and listed with their own results. Confirm the alignment is datum based, not best fit. A best-fit alignment floats the part until the sum of all deviations is smallest, which is exactly how the 0.12 mm housing above printed PASS on every row while its true position was 0.07 mm out.

Also check the datum features themselves for form. A datum plane that measures 0.08 mm flat against a 0.05 mm flatness callout drags every other feature with it, even when the other features are perfect. If a report shows every position feature failing by nearly the same amount and in the same direction, suspect the datum, not the mold.

Ask the supplier to print the datum alignment summary in plain language: which surfaces, how many points each, and which software routine built the frame. When a mold maker can answer that question without opening the program, the report is trustworthy. When the answer is vague, treat the whole page with suspicion.

Linear Dimensions and True Position

Linear dimensions are the easiest rows to read. A boss length of 12.00 mm nominal with a tolerance of ±0.05 mm passes when the actual sits between 11.95 and 12.05 mm. The deviation column shows whether the part ran big or small, and the margin tells you how close the edge of the band you are.

Deviation signs carry information. A positive deviation means the feature is larger than nominal or further along the axis. A negative deviation means smaller or shorter. One off-center reading is a process story. Every reading off in the same direction is a compensation story, which matters even more for plastic than for steel, because mold shrinkage is what the cavity was cut to absorb.

True position is where new readers get lost. Under ASME Y14.5, a position callout like the one on your connector drawing describes a cylindrical tolerance zone, and the callout diameter is that zone. The reported position value is the diameter of the actual error, which is two times the radial offset of the feature center from its true location.

Work the arithmetic once and it sticks. A boss center sits 0.021 mm off its true location radially. Doubled, the reported true position is 0.042 mm. Against a tolerance zone of Ø0.08 mm, the feature passes with 0.038 mm of margin. A radial error of 0.12 mm, like the housing in the opening, reports as 0.24 mm and fails a Ø0.05 mm zone by a wide margin.

Watch for the MMC modifier, written as a circled M after the tolerance. Material condition modifiers add bonus tolerance when the actual size departs from maximum material condition. A hole drilled 0.10 mm oversize at MMC gets bonus that can double the position zone, so two reports on the same feature can pass differently depending on size. Ask whether the CMM program applied the bonus, and ask for the bonus amount in the results row.

Form and Profile: Flatness, Roundness, and the Rest

Form tolerances control the shape of a single surface, with no datum required. Flatness is the classic case. The callout defines two parallel planes, and the measured value is the width of the band that contains the whole surface. A flatness result of 0.02 mm means the surface fits between two planes 0.02 mm apart, and a 0.05 mm flatness callout passes.

Flatness on a datum feature is the one to stare at longest. A seat face that is 0.04 mm flat but called at 0.05 mm passes its own row and still tilts the datum frame for everything mounted on it. For mating faces, ask what the downstream assembly actually tolerates, not just what the print allows.

Parallelism and perpendicularity are orientation tolerances, and they always reference a datum. Parallelism of 0.03 mm to datum A means the probed surface must stay within two planes 0.03 mm apart that run parallel to datum A. Perpendicularity works the same way against a datum at 90 degrees. Orientation results depend on the datum frame being right, which loops back to the second section of this guide.

Roundness measures how close a circle is to a true circle. The reported value is the radial difference between the largest inscribed and smallest circumscribed circle, or the zone width of the best-fit ring. A pin measuring 0.008 mm roundness holds its diameter variation inside 0.008 mm. Roundness depends on sampling strategy, and a three-point circle can miss a three-lobed shape entirely, which is covered in the measurement section below.

Profile of a surface is the heavyweight callout. It wraps form, orientation, and location into one zone around a 3D surface, and it is the standard way to control class-A surfaces and seal faces. The reported value is the zone width, and a profile of 0.20 mm against a 0.10 mm callout fails no matter how beautiful the part looks. Profile rows are expensive to measure and worth questioning when a buyer sees them on every feature of a commodity housing.

← swipe to scroll →

Table: Common GD&T callouts on molded part drawings, what the number means, and how the CMM decides PASS.

Every form result should carry context about how it was measured. Ask how many points the machine took on the surface and over what length the zone was evaluated. Flatness over a 10 mm pad and flatness over a 60 mm sealing face are different statements wearing the same name.

Deviation, Tolerance, and Cpk: What the Numbers Say

The deviation column is where the story lives, and the sign of the story matters. On a molded part, all deviations leaning one direction usually means the shrinkage compensation does not match the real process. If every boss measures 0.02 mm large, the cavity was likely cut for a shrinkage rate higher than the resin actually delivered, or packing pressure is holding the part above nominal size.

Center of tolerance is a design conversation, not a defect. Many mold makers aim features at the center of the band, but some programs deliberately target one side. Steel wears, cavities grow with use, and a mold cut to the low side of a critical pin diameter gives the tool a wear life before it drifts out of band. Ask what target each critical feature was cut to, then compare the report against that target, not just against the print.

That target logic is why DieStrike validates tool steel geometry at ±0.005 mm standard and ±0.002 mm precision before any molding starts. The steel is measured before plastic shrinkage enters the equation, and the molded part is measured after. Two reports, two questions: is the steel right, and is the compensation right. Our guide to holding ±0.002 mm on precision mold tolerances walks through the steel-side half of that logic.

Capability answers a different question than deviation. Deviation asks whether this part is in tolerance. Cpk asks whether the process that made it can stay in tolerance over time. Cpk compares the distance from the process mean to the nearest tolerance limit against three standard deviations of the process spread.

Work an example with real numbers. A critical gap is called at 10.00 mm ±0.05 mm, so the tolerance width is 0.10 mm. The trial run shows a standard deviation of 0.012 mm. Cp is the tolerance width divided by six sigma, which gives 0.10 divided by 0.072, or 1.39. That looks strong until you check centering.

Cpk subtracts the mean shift. If the process mean sits at 10.03 mm instead of 10.00 mm, the distance to the upper limit at 10.05 mm is only 0.02 mm. Divide by three sigma at 0.036 mm and the upper-side Cpk collapses to 0.56. A perfectly tight process, off-center by 0.03 mm, fails every capability gate it meets. This is why Cpk, not Cp, is the number on your acceptance criteria.

Common industry practice sets Cpk of 1.33 as the minimum for standard features and 1.67 for critical ones, with 1.67 roughly meaning the process runs inside about 60% of the tolerance band at six sigma. Verify the threshold in your contract, because some automotive and medical programs demand 1.67 across the board. Also check the sample size behind the Cpk line. A Cpk computed on 5 parts is a rumor; 30 parts is a common initial-study size, and production validation often uses larger runs over multiple shifts.

How the Part Was Measured: Points, Temperature, Uncertainty

The measurement method shapes the numbers, and the first question is how many points the probe took. A circle measured with 3 points fits a perfect circle through any 3 points and reports whatever radius that circle has. A three-lobed bore, which is exactly what a worn or unbalanced process can produce, is invisible to a 3-point measurement because the probe can sit on the lobes and valleys in ways that cancel.

Typical CMM practice on molded features uses 6 to 12 points on a small boss or hole, and scanning machines collect hundreds of points around a contour. More points catch lobing, out-of-roundness, and localized defects. If a report lists a critical position or roundness result and the program summary shows 3 points, ask for a re-measure with a denser strategy before you argue about the mold.

Temperature is the second variable, and it is the one most often ignored. The standard metrology atmosphere is 23 °C, and precision reports should state 23 ±1 °C, matching common laboratory practice. ISO 291 defines the general plastics conditioning atmosphere as 23 °C and 50% relative humidity, and dimensional reports assume the part and the machine reached that state.

Parts need time to get there. Molded parts leave the press warm and keep moving for hours. Industry practice for semicrystalline resins commonly calls for 24 to 48 hours of conditioning before measurement, because crystallinity continues to develop and dimensions keep settling. Amorphous grades settle faster, but even they should sit overnight before a report that feeds a mold release decision is written.

Thermal expansion explains why the temperature line matters. Tool steel grows about 11 to 12 µm per meter per kelvin. A 300 mm mold block measured at 28 °C instead of 23 °C is 5 kelvin warm, which moves a feature roughly 0.018 mm. That is four times the ±0.002 mm precision tolerance class and enough to flip a PASS into a FAIL on a tight automotive feature.

Plastic moves far more than steel. Typical unfilled engineering resins expand at 60 to 100 µm per meter per kelvin, five to eight times the rate of tool steel. A 100 mm part measured 10 kelvin off standard moves 0.06 to 0.10 mm, which is a whole tolerance band on most prints. If a report has no temperature record, treat every borderline feature as unverified.

Uncertainty is the last layer. CMM makers quote accuracy limits under ISO 10360, and a common rule of thumb says the measuring system should be about 10 times more accurate than the tolerance it checks. A feature at ±0.05 mm needs measurement uncertainty near 0.005 mm or better. Ask for the machine class and calibration date when a supplier reports thousandths-level tolerances, and treat the report as suspect if the machine cannot resolve what it claims.

Red Flags and What to Do When Features Fail

Some report patterns deserve pushback before any acceptance discussion starts. All deviations leaning one way means systematic compensation or process bias, not random chance. A datum feature out of tolerance means the frame itself is compromised. A best-fit alignment, missing datum rows, or no datum summary means the report may not answer the drawing at all.

Missing traceability is its own red flag. A report without an operator, a date, a machine ID, or a calibration date is not evidence. Three-point circles on critical features, no temperature record, no sample count behind the Cpk, and identical readings across all cavities of a multi-cavity mold all justify a question. Every cavity in a 16-cavity tool measuring within 0.001 mm of each other is either a superb process or a copy-paste program.

When a real feature fails, the sequence is simple. First, re-measure. Reseat the part, rerun the feature, and confirm the alignment is datum based. Measurement artifacts kill more trial arguments than steel errors do. Second, check the datum features themselves. Third, confirm the failed feature was measured with enough points at the right temperature after proper conditioning.

If the failure survives re-measurement, it is a mold and process problem, and that is the moment for the DFM conversation. The fix is rarely a single number. Gate size, packing pressure, cooling balance, and the shrinkage compensation value all push dimensions, and the DFM report written before steel cutting is the baseline for deciding which one to change.

Expect corrections inside the trial process itself. At T1, dimensional findings feed steel adjustments like gate enlargement or cavity compensation changes, and the mold returns for T2 sampling. The full workflow, including how many shots each stage should take and what records you should receive, is covered in our guide to running a mold trial. Validation programs then take the same dimensional data into IQ, OQ, and PQ protocols that prove the process, not just the parts.

Hold your acceptance criteria in writing before the argument starts. Agree on the Cpk threshold, the number of cavities sampled, the conditioning time, and the temperature standard in the mold specification. A buyer who defines those four items up front settles most disputes with the report itself, and a supplier who documents them delivers a report you can sign.

FAQ: Reading CMM Reports

Q1. What do nominal, actual, and deviation mean on a CMM report?

Nominal is the target from the CAD model or print. Actual is what the machine measured. Deviation is actual minus nominal, so a positive deviation means the part is larger than the drawing and a negative one means smaller. Compare the deviation against the tolerance column to judge the margin.

Q2. Why is true position reported in millimeters as a diameter?

ASME Y14.5 defines a position tolerance zone as a cylinder. The callout diameter is the zone size, and the reported value is the actual error as a diameter, which is twice the radial offset of the feature center. If the reported value is less than the callout, the feature is inside the zone and passes.

Q3. What is the difference between Cp and Cpk?

Cp compares the tolerance width to six sigma of the process spread and assumes perfect centering. Cpk also accounts for the mean shift and takes the smaller of the upper and lower side calculations. Cpk is the honest number, because a tight process that runs off-center fails capability.

Q4. Why must parts sit at 23 ±1 °C before measurement?

Dimensions depend on temperature. Plastics expand 60 to 100 µm per meter per kelvin and tool steel about 11 to 12 µm per meter per kelvin, so a warm part or a warm room shifts results by more than a precision tolerance. Conditioning for hours at the standard atmosphere makes the numbers repeatable and comparable across suppliers.

Q5. When should I reject a CMM report?

Reject when the datum setup is missing or best fit, when traceability fields are blank, when critical features were measured with 3 points, when no temperature is recorded, or when deviations lean one way across every feature. Those patterns mean the report is not yet evidence, and you should ask for a corrected one before discussing the mold.

Q6. What Cpk value should I require on a new mold?

Common industry practice is Cpk of 1.33 minimum for standard features and 1.67 for critical ones. Confirm the threshold in your contract, verify the sample size behind the number, and require the study on the features your assembly actually depends on, not just the ones that pass.

The Bottom Line

A CMM report is a claim about geometry, and every claim carries assumptions. The datum frame must be real. The columns must be read before the flags. The deviation sign must be interpreted against the compensation target, and the Cpk must be judged on enough parts at the right temperature.

Adopt one habit: read the report top to bottom, header to Cpk, before forming an opinion. Check the datum section first, then the worst feature, then the capability line, then the measurement conditions. Nine times out of ten, the failure of a report is the failure of its setup, not its mold.

When the numbers survive that reading, they become the cleanest evidence you will get that a tool is ready. When they do not, they are the cheapest discovery you will make, because a dimensional problem found on a CMM costs steel time, while the same problem found on the assembly line costs a program.

DieStrike builds precision molds under IATF 16949 across 120+ machines, with geometry at ±0.005 mm standard and ±0.002 mm precision, cavity steel at HRC 62, and mold deliveries in 2 to 5 weeks. Customers including TE Connectivity, Amphenol, and Luxshare run connector programs on the same measurement discipline. Every mold leaves with a signed dimensional report you can actually read.