Print-to-Build vs. Sample-to-Build: The Reverse-Engineering Trap and the Drawing That Lied

The Two Paths, Defined at the Shop Door

Print-to-build means the customer sends a dimensioned drawing — a PDF, a SolidWorks file, a fully toleranced print — and says, “Make this.” The authority is the drawing. If the part matches the print, it’s accepted. The drawing is supposed to contain everything: material, surface finish, heat treatment, critical dimensions, revision history. It’s a contract written in views and datums.

Sample-to-build means the customer sends a physical part — often an old, used, or broken component — and says, “Copy this.” There is no drawing, or perhaps a rough sketch. The authority is the physical artifact. The machinist or engineer must measure the sample, interpret its features, guess at the material and heat treatment, and produce a replica that fits and functions like the original. It’s essentially reverse engineering, and it’s full of traps that a drawing would have avoided.

Old Li, when he first saw me struggling with the greasy shaft, wiped it with a rag and held it up to the light. “A drawing is a promise. A sample is a suggestion. A worn sample is a suggestion from a tired old man who has forgotten his own name. Copy it literally, and you’ll reproduce every mistake its maker made, plus the mistakes the last ten years of service added.” He was right, and we’ve since built a formal process for each path.

Print-to-Build: The Clarity of a Good Drawing, and the Chaos of a Bad One

When a customer sends a clean, well-dimensioned drawing with reasonable tolerances, material callouts, and surface finish specs, print-to-build is a joy. We can program directly from the model, set up inspection to the print, and know exactly what “good” looks like. There’s no ambiguity about whether a feature is functional or cosmetic, because a good drawing tells you.

The problems start when the drawing is poor. We’ve seen prints where every dimension is ±0.001″ because the designer didn’t understand which tolerances mattered, or where critical holes are dimensioned from a corner that will be removed in the first operation, or where the material is listed simply as “steel” without a grade. A bad drawing is an expensive guessing game. We’ve learned to treat every new customer drawing with a skeptical eye: redline it, question the tolerances, and ask for missing material specs before we quote. If a dimension seems unnecessarily tight, we call. If a surface finish is specified on a non-functional surface, we ask why. The conversation often results in a revised print that saves the customer money and saves us from making a scrap part that was “to print” but functionally useless.

Old Li keeps a red pen in his shirt pocket specifically for drawing review. He’ll circle an impossible sharp internal corner and write, “Tool radius?” or point to a blind threaded hole and ask, “Chip evacuation?” He treats every drawing like a puzzle that might have a missing piece. His favorite saying during print reviews: “The CAD model is perfect. The real world has drills, endmills, and gravity. Bridge the gap, or the gap will eat your margin.”

Sample-to-Build: The Detective Work Behind Copying a Part

Reverse engineering a sample is part measurement, part materials science, and part intuition. The shaft disaster taught us that simply measuring a used part with calipers and a micrometer is not enough. A worn shaft has lost material; the keyway edges are rounded; the thread pitch diameter may have been distorted by years of tightening. If you measure it and cut to those numbers, you’re cutting a pre-worn part.

Now, our sample-to-build process starts with a formal incoming inspection of the sample itself. We document every feature under a microscope, photograph the wear patterns, and estimate the original dimensions based on the least-worn areas. For the shaft, we cut off a small section and examined the thread form on an optical comparator, which revealed the original Class 3 fit and the root radius we’d missed. We also sent a small chip to a materials lab for spectrometry to confirm the alloy — it was 17-4 PH, not 303 as we’d assumed. The lab report cost $150 and saved us from making 50 shafts from the wrong material.

We also learned to ask the customer contextual questions: “What does this part mate with? Can you send us the mating component? What was the failure mode that caused you to replace it?” Those answers often reveal design intent that the worn sample has lost. In the case of the shaft, the mating gear had a hardened bore with a sharp corner that had fretted against the shaft groove, wearing away the radius. The gear gave us the original groove geometry.

Old Li, during the shaft investigation, placed the worn shaft and the gear on a granite plate, slid them together, and pointed to the fretting marks. “The sample lies,” he said. “The assembly tells the truth. Always measure the hole that the shaft goes into. It remembers what the shaft forgot.”

The Hidden Traps of Each Approach

Print-to-build traps:

• Over-dimensioning and false precision. Tighter than necessary tolerances that increase cost without adding function.

• Missing process information. The print may not specify that a hole must be drilled before heat treat, or that a surface must be masked during plating.

• Material substitutions. A print might call out an obsolete alloy, or a grade that’s unavailable in the required stock size.

• Assumptions about setup. The print shows the finished part, not how it’s held. Fixturing may be impossible without added tooling tabs.

Sample-to-build traps:

• Wear and distortion. The sample has changed dimensionally during its service life.

• Material ambiguity. Without a lab test, you don’t know the alloy, heat treat, or coating.

• Incomplete reverse engineering. Internal passages, blind hole depths, or spring tensions are hard to measure non-destructively.

• Design intent lost. You can copy the shape but not necessarily the engineer’s reasoning behind a subtle radius or a specific surface finish.

• Tolerance guessing. Without a print, you don’t know whether a dimension was ±0.05mm or ±0.5mm. You have to infer from the fit.

We’ve learned that the worst scenario is a hybrid that has the weaknesses of both: a customer who sends a sample and a “sketch” with hand-scrawled dimensions that don’t match the sample. That’s a recipe for disaster. We now refuse to accept a sketch as a drawing; if there’s a sample, we reverse-engineer it and create our own CAD model, then send it to the customer for sign-off before we cut metal. That sign-off step has caught countless mismatches.

When to Trust the Print, When to Trust the Sample, and When to Ask for Both

In an ideal world, the customer provides both: a detailed, toleranced drawing and a reference sample. The sample confirms what the drawing intends, and the drawing explains what the sample can’t (material, heat treat, tolerances). When we have both, we measure the sample against the print, note any discrepancies, and resolve them before machining.

When we only have a print, we follow it exactly, but we proactively flag any features that seem unreasonable and ask for clarification. The redline loop is built into our quoting process. Old Li calls it “measure twice, cut once, but ask three times before you quote.”

When we only have a sample, we follow a strict reverse-engineering protocol: measure and document the as-received condition, estimate original dimensions, identify material and heat treatment (via lab testing if necessary), create a CAD model and a provisional drawing, and send both to the customer for written approval. Only then do we machine. It adds a day or two to the lead time, but it eliminates the “we copied your worn part” conversation.

The Horror Shelf Gains a Liar and a Worn-Out Truth

The original worn shaft that started this lesson, along with one of the incorrectly copied new shafts, now sits on our horror shelf. Old Li mounted them side by side on a small aluminum plate, with labels. The worn shaft: “This one lied.” The new shaft: “This one believed the lie.” Between them, a small sign: “Sample-to-Build: Reverse-engineer the design intent, not the wear.” Next to it is a drawing from another job that was entirely covered in redline marks — Old Li’s work — with a note: “Print-to-Build: Good drawing = good part. Bad drawing = expensive art.”

When a new customer asks us to copy a part, we walk them past the shaft display and explain the process. When a customer sends a print full of questionable tolerances, we show them the redline-marked drawing and explain that we’re not being difficult — we’re preventing a horror shelf exhibit with their name on it.

A Short Guide for Shops and Customers

If you’re a machine shop navigating these two worlds, here’s the guide we now follow:

For print-to-build:

1. Redline every new drawing. Check for impossible geometry, missing tolerances, unclear material specs, and over-dimensioning.

2. Ask about functional intent. Which dimensions are critical? What does the part mate with? This guides tolerance decisions.

3. Confirm material and condition. Is the material available? Does it need heat treatment, and at what stage?

4. Generate an inspection plan from the print. If you can’t measure it, you can’t prove you made it right.

For sample-to-build:

1. Document the as-received condition. Photograph, measure, note wear and damage. Don’t trust a single measurement.

2. Identify material and hardness. Use a lab if necessary. Guessing “stainless” isn’t enough.

3. Infer original dimensions from the least-worn areas and from mating parts.

4. Create a CAD model and provisional drawing, and get customer sign-off before machining. This is non-negotiable.

5. If the sample is a one-of-a-kind artifact that cannot be sacrificed, use non-destructive testing (CT scanning, laser scanning) for internal features.

6. Protect the sample. Treat it like evidence. Return it with the finished parts.

The shaft re-order? We remade the batch with the correct material, the Class 3 thread, the radiused groove, and the proper keyway tolerance. The customer paid the difference, acknowledging that our first attempt was based on a worn sample with incomplete information. They’ve since sent us a drawing for every new part, and we’ve kept that worn shaft as a permanent reminder that copying is not the same as understanding.

Old Li summed it up last week while packing a box of precision dowel pins: “A drawing is a language. A sample is a memory. Both can be wrong, but a drawing wrong is a mistake in translation. A sample wrong is a legend passed down by a storyteller with a bad memory. Trust neither completely, but verify both against the machine that will hold them.”

Have you ever been burned by a reverse-engineering job where the sample was worn, or by a print that had an impossible tolerance stack? How do you bridge the gap between what the customer sends and what the machine needs to make? Share your print-to-build and sample-to-build stories in the comments — we’re all students of the worn shaft and the redlined drawing.