3D scanning captures the surface of a physical object as millions of measured points, which are then processed into a mesh and, if you need to manufacture or modify the part, rebuilt into editable CAD geometry. The scan itself is the easy half. The part that takes real time — and where the cost sits — is converting that mesh into a clean parametric model a machinist or mold maker can work from. Understanding that split is the difference between a useful quote and a disappointed client.

When you actually need a scan

Scanning earns its cost in a few specific situations:

  • Organic or freeform shapes — a handle sculpted in clay, an anatomical form, a shoe last, a styled housing. You cannot measure these with calipers in any practical way.
  • A legacy part with no drawings — a discontinued component you need to reproduce or improve.
  • Designing something that must mate with an existing object — a bracket for a specific machine, an accessory for a vehicle interior, a device that clips onto existing equipment.
  • Physical design iteration — a hand-modified prototype where the changes were made with a file and sandpaper and now need to be captured.
  • Inspection — comparing a manufactured part against its CAD model to see where the process drifted.

Notice what is missing from this list: simple prismatic parts. A flat plate with six holes does not need scanning — it needs fifteen minutes with a caliper.

How it works: from object to point cloud

Several technologies serve different sizes and accuracy classes.

  • Structured light projects a known pattern onto the object and reads its distortion with cameras. Excellent accuracy on small and medium parts, and the most common choice for product work.
  • Laser line and laser point scanners sweep a beam across the surface. Good on a wide range of materials and often mounted on an articulated arm for larger objects.
  • Photogrammetry reconstructs geometry from many overlapping photographs. Cheap and accessible, lowest accuracy, best for large objects and visual reference.
  • CT scanning images internal structure as well as external surfaces — the only option when you need to see inside a sealed assembly, and the most expensive.

Practical scanning constraints matter more than the marketing specs. Shiny, transparent, and very dark surfaces confuse optical scanners and usually need a temporary matte coating. Deep narrow features and internal cavities are hard or impossible to reach optically. Large objects require multiple overlapping scans stitched together, and stitching introduces its own error.

From point cloud to a file you can work with

The raw output is a dense point cloud. Processing it runs roughly like this: align and merge the individual scans, delete noise and stray points, fill holes where data was missing, and generate a watertight triangle mesh. That mesh is enough for 3D printing a copy, for visual reference, or for inspection comparison.

It is not enough for manufacturing or modification. A mesh has no dimensions you can change, no flat faces that are truly flat, and no holes that are truly round — the same limitation described in STL vs STEP file formats. To get a workable model, an engineer rebuilds the geometry: fitting planes, cylinders, and true radii to the scan data, re-establishing datums and symmetry, and constructing proper features with real dimensions. The scan becomes a reference the model is built against, not the model itself. Our article on reverse engineering a part covers that rebuild strategy in depth.

How far you take that rebuild is a scope decision. Sometimes only the mating surfaces need to be exact and the rest can stay as scanned surface. Sometimes the whole part must become clean parametric geometry. Agree on this before work starts, because it drives most of the cost.

Accuracy: what you really get

Scanner specifications quote accuracy under ideal conditions on ideal surfaces. Real results depend on the object, and error compounds through the workflow:

  • Measurement error at the sensor.
  • Alignment error when merging multiple scans.
  • Interpolation error where holes in the data were filled.
  • Fitting error when the engineer approximates scan data with idealized surfaces.

For product work, high-quality structured-light scanning of a small part typically lands in the tens of microns to low tenths of a millimeter. That is excellent for form and for most fits, but a critical bearing bore or a sealing surface should still be verified with a direct measurement rather than trusted to a scan. Also remember the physical part you scanned already contains its own manufacturing deviation — you are capturing one specific object, not the designer's intent, which is why the tolerance thinking in GD&T basics matters here.

Scanning or hand measurement: how to choose

  • Prismatic geometry — flats, holes, slots, simple curves? Measure it. Faster, cheaper, and often more accurate for the dimensions that count.
  • Freeform surfaces, sculpted forms, or complex contours? Scan it.
  • Both in one part? A hybrid: scan for the organic surfaces, caliper and micrometer for the critical prismatic features. This is the most common real-world approach and usually the most economical.

Time and what makes up the cost

Scanning a small to medium object is generally a matter of hours including setup and surface preparation. Mesh cleanup is another few hours. The CAD rebuild is the variable: a simple part can be a day, a complex housing with many blended surfaces can be a week or more.

In US dollars, expect scanning alone to be a modest few-hundred-dollar service for a small part, and the scan-to-CAD conversion to run from the high hundreds into several thousand depending on complexity and how much of the model must be fully parametric. Large objects, CT scanning, and full parametric rebuilds of complex geometry sit at the top of that range. For context on where this fits in a project budget, see what a prototype costs and CAD design services cost.

What to do with the file the day after

Once you have real CAD, the normal product path opens up: modify the design, run analysis, quote it for machining or molding, or print an updated version. Make sure you receive both the native CAD and a STEP export, and store them somewhere the project will not lose them. From there, our prototype development guide and 3D printing services for prototypes cover the next steps.

Have a physical part that needs to become a manufacturable model? Tell us about it through our contact form — include the size, the material, and what you plan to do with the file, and we will scope the scanning and CAD work.