Why do parts from the same file come out different the second time?

Because the file is the only thing your order actually fixed. A quote fixes price, quantity and a ship date — it does not reserve a machine, a material lot, a build orientation, an operator or a finishing recipe. Any of those can change between batch one and batch two without anybody making a decision about your part. The most common culprits are a different machine in the same farm, a fresh material lot, and a different nesting of the build plate. All three are routine shop behaviour, and all three move dimensions by amounts that matter in a press fit.

Can two batches both be in spec and still not fit together?

Yes, and this is the single most misunderstood part of buying printed parts. A tolerance is a band, not a target. If the band is ±0.3 mm, a batch averaging 0.15 mm under nominal and a batch averaging 0.15 mm over nominal are both perfectly in spec and are 0.3 mm apart from each other. Fixtures, mating parts and assembly jigs built around the first batch do not care that the second one passed inspection. If batch-to-batch interchangeability is what you need, you have to buy it explicitly — spec does not supply it by default.

What should I put on a purchase order so a reorder repeats?

Five lines, and they are all things the shop already knows internally. Name the process and the machine family that ran the original job. Name the material by full grade and colour, not by family — PA12 natural, not nylon. State the build orientation, or ask the shop to record and repeat the one they used. List the two or three dimensions that are actually critical, with their tolerance and datum. And name the finish by operation, not adjective — bead blast, 120-grit, dyed black. A reorder quoted against those five lines is a repeat, not a fresh job.

Is first-article inspection worth paying for on a reorder?

On anything that mates with something else, usually yes — it is a day of delay against a scrapped assembly run. Ask for a dimensional report on the first part of the batch covering your critical dimensions, before the rest of the run is finished. The cost is typically modest next to the batch, and the real value is that it converts a dispute into a measurement: if the report shows batch two running at the opposite end of the band from batch one, you have a factual basis for a remake rather than an argument about whether the parts feel different.

How do I check the difference if I don't have a CMM?

Calipers and discipline get you most of the way. Measure the same dimension on five parts from each batch rather than one, write down every reading, and compare the averages as well as the spread. One part proves nothing because you may have grabbed an outlier from either run. Measure at room temperature, and let nylon parts sit a day if they arrived sealed, since they take up moisture and grow slightly. If the geometry is organic or the critical feature is a surface rather than a dimension, that is the point to pay for a scan-based inspection report instead.

Should I reorder from the same provider or switch?

Switching resets the problem rather than solving it, because a new shop is a new machine, a new lot and a new operator all at once — a guaranteed second drift. The better move is to give the current shop the information that lets it repeat the job, then judge the answer. A provider who can tell you which machine ran the first batch, what lot the material came from and how the plate was nested is a provider who can reproduce it. One that cannot answer any of those questions is not keeping the records that repeatability requires, and that is the real reason to move.

Why Your 3D Printing Reorder Doesn't Match

The file has not changed. The shop has not changed. The purchase order says the same thing it said four months ago. And the parts that arrived on Tuesday will not go into the fixture that the first batch dropped into. A 3D printing reorder that comes back different is not usually a mistake anybody made — it is the gap between what you thought you bought and what the quote actually fixed. Closing that gap takes about five lines on a PO, and knowing which five is the whole job. If the shop cannot tell you what changed, browse the 3D printing provider directory and ask two or three others the questions at the end of this post before you place the repeat.

The Quote Fixed Less Than You Think

When a bureau quotes a job, it commits to a price, a quantity, a material family and a date. It does not commit to the machine, the material lot, the position on the build plate, the operator, the slicer profile version, or whether the job stays in the building. Those are scheduling decisions, remade every morning, and nobody involved considers them a change to your part — because against the drawing, they are not.

That is the asymmetry. You experience a reorder as "run it again." The shop experiences it as a new job with a known geometry. Both are reasonable descriptions, and only one of them produces interchangeable parts.

Six Things That Moved Between Batch One and Batch Two

1. A different machine ran it

Bureaus run farms, not single machines. Two printers of the same model, calibrated the same week, still differ — belt tension, nozzle wear, a resin vat with a few hundred more hours on it, a laser that has drifted a fraction since its last calibration. Within a farm those differences are small and real, and they land squarely on the dimensions you care about. A shop that can tell you which machine ran the original is a shop that can run it there again.

2. The material lot changed

Polymer lots vary. Shrinkage, colourant loading, moisture content on arrival and filler dispersion all move slightly between batches, and powder processes add a second variable: the ratio of virgin to recycled powder in the blend. A nylon part printed on a fresh powder refresh and the same part printed late in a powder cycle are not quite the same part. Naming the grade and colour on the PO closes most of this; the trade-offs between grades are in how to choose a 3D printing material.

3. The part was oriented or nested differently

This is the big one, and the most invisible. Build orientation decides which direction the layers run, which faces carry support witness marks, how much a long dimension shrinks and which axis is strongest. A shop nesting a full plate economically will rotate your part to fit the other work on it — a completely ordinary decision that can move a 100 mm dimension by more than the tolerance you assumed. The same effect, at hobby scale, is covered in why the same file prints differently on two printers.

4. The job was subcontracted

Overflow work goes to partner shops, especially on a tight date or an unusual process. The parts are still made to your file — by a different machine on a different floor with a different material supplier. Nothing about this is dishonest and almost nobody mentions it unprompted. Ask whether the work is run in house; the answer is a fact about repeatability, not an accusation.

5. Finishing was done differently

As-printed parts vary less than finished ones. Bead blasting, sanding, tumbling, vapour smoothing and dyeing are hand-and-judgment operations, and "the same finish" to a shop means the same process, not the same number of minutes. A sanded part can lose a few hundredths of a millimetre more than the last one did, and a dyed part can land a shade off. Spell the finish out as an operation rather than an adjective — this is the stage most often agreed by implication, as 3D printing post-processing and finishing services sets out.

6. Software moved under the job

Slicers update. Firmware updates. Shrinkage-compensation profiles get retuned after a calibration study. Six months is plenty of time for the shop's standard profile for your material to have been revised for good reasons that have nothing to do with you. This one is invisible even to the operator, and it is the reason a long-running part is worth re-measuring on every reorder rather than only when something feels wrong.

Both Batches Were In Spec. That Is the Problem.

A tolerance is a band, not a target, and a batch tends to sit somewhere inside the band rather than in the middle of it. Two runs can therefore pass the same inspection and still be nearly a full band apart from each other:

	Nominal	Tolerance	Batch mean	In spec?
First order	40.00 mm	±0.3 mm	39.86 mm	Yes
Reorder	40.00 mm	±0.3 mm	40.19 mm	Yes
Difference between them	—	—	0.33 mm	—

Nothing failed. Every part shipped inside the quoted capability. And the fixture that was drilled around the first batch, or the mating part machined to suit it, is now 0.33 mm out — which is the difference between a press fit and a part that falls out. Tolerances and accuracy in 3D printing explains how providers publish these bands as a percentage of nominal with a floor, and why the absolute number grows with part size.

The practical consequence is simple: spec does not buy interchangeability — a tighter band on the two dimensions that matter does. If batch two has to mate with hardware built around batch one, say so, and expect to pay for the inspection that enforces it.

Prove It Before You Argue It

Measurement is what turns a complaint into a claim. Before emailing the shop:

Measure five parts per batch, not one. A single part might be an outlier from either run. Write down every reading.
Compare the averages and the spread separately. A shifted average is a process change at the shop. A wider spread on the same average is a stability problem, and it is a different conversation.
Keep a golden sample. One part from every batch, labelled with the date and PO number, in a drawer. It is the cheapest quality system that exists, and it is the only thing that lets you measure batch three against batch one rather than against memory.
Measure in the same conditions. Room temperature, and let sealed nylon parts sit a day — powder-process nylon takes up moisture from the air and grows slightly as it does.
Use a scan when the feature is a surface. Calipers cannot describe a warped face or a drifting organic contour; 3D scanning services for quality inspection covers when a scan-based report earns its cost.

With numbers in hand the question stops being "do these feel different" and becomes "batch two is running 0.33 mm above batch one on the bore spacing — what changed?" That question gets answered. The other one gets a shrug. If the parts are not merely different but actually wrong, why your 3D printing order came back wrong covers recourse.

How to Order So the Next Run Repeats

Put these on the purchase order rather than in the email thread:

Reference the original job number. It is how the shop retrieves the machine, profile and lot records in the first place.
Name the material in full. Grade, colour and supplier if you know it. "Nylon" is a family; "PA12, natural" is a material.
Fix the build orientation. Either state it, or ask the shop to record and repeat the orientation used on the original — most will, and almost none will volunteer it.
Mark the critical dimensions on a drawing. Two or three, with tolerance and datum, not every dimension on the part. A drawing that flags what matters gets respected; one that demands ±0.1 mm everywhere gets ignored or priced as impossible. The conventions are in design for 3D printing guidelines.
Ask for first-article inspection on those dimensions. On the first part of the run, before the rest is finished.
Order the full quantity at once where you can. The cheapest way to guarantee that 200 parts match each other is for all 200 to come off the same machine, from the same lot, on the same day. Where cash flow prevents that, a blanket order with scheduled releases gets most of the benefit.

When the Change Was On Your Side

Rule this out before raising it, because it happens often enough to be embarrassing. A re-exported STL can carry different unit assumptions or a coarser chord tolerance, which quietly rounds off small features. A CAD revision made for an unrelated reason may be sitting in the file you sent. A scaling factor applied once to compensate for shrinkage can end up applied twice.

Check the file you sent against the file that produced batch one — byte for byte if you still have it — before you ask the shop what they changed. Finding it yourself is much cheaper than finding it in the reply.

Buy From a Shop That Can Answer the Question

Repeatability is a records problem as much as a machine problem. Ask any provider three questions before a reorder: which machine ran the original, what material lot it used, and how the plate was nested. A shop with a quality system answers all three from the job file in minutes. One that cannot answer any of them is not keeping the records a repeat requires, and goodwill does not substitute for the data.

Browse 3D printing providers, filter by process, and ask those three questions alongside your price request. The one that answers them is the one whose second batch will match its first. If you are choosing a supplier from scratch, how to choose a 3D printing service covers the rest of the qualifying criteria, and 3D printing service certifications explains which quality standards actually imply the record-keeping described here.

Related Resources

Tolerances and Accuracy in 3D Printing — what each process can hold, and why the band grows with size
Why Your 3D Printing Order Came Back Wrong — when the parts are not merely different but defective
3D Printing Service Certifications — the standards behind repeatable records
3D Scanning Services for Quality Inspection — measuring surfaces calipers cannot describe
How to Choose a 3D Printing Service — qualifying a provider before the first order
Browse 3D Printing Providers — ask which machine ran the last batch

Hero photograph by Greg Rosenke via Unsplash.