A prototype can look right on screen, print cleanly, and still fail at the moment of assembly. A lid binds. A shaft has too much play. A purchased bearing will not seat. These are usually not single-dimension mistakes. They are stackup problems. This tolerance stackup guide explains how dimensions vary across mating parts and how to design for an assembly that works outside of a perfect CAD model.
For startups, inventors, and small product teams, stackup analysis is not an academic exercise reserved for high-volume manufacturing. It is a practical way to prevent expensive revision cycles before you order a batch, source hardware, or commit to tooling.
What a tolerance stackup tells you
Every manufactured dimension has an allowable range. If a feature is specified as 10.00 mm plus or minus 0.10 mm, the actual feature may measure anywhere from 9.90 to 10.10 mm and still meet the drawing requirement. That range is the tolerance.
A tolerance stackup looks at what happens when several of those variable dimensions affect one functional result. For example, the clearance between a printed enclosure wall and an internal circuit board depends on the enclosure depth, the board thickness, standoff height, gasket thickness, and the location of any mounting features. Each part may be acceptable on its own, while the combined assembly still creates interference or an unacceptable gap.
The goal is not to make every dimension as tight as possible. Tight tolerances add cost, increase inspection needs, and can make a part harder to produce consistently. The goal is to assign precision where the product needs it and allow reasonable variation everywhere else.
Start with the function, not the dimensions
Before adding numbers to a drawing, define what the assembly must do at its limits. A snap-fit cover may need to close without excessive force while remaining secure after repeated use. A hinge may need free rotation without visible wobble. A bearing pocket may need enough interference to retain the bearing without cracking a resin or plastic housing.
This functional requirement sets the acceptable range for the final condition. Say an enclosure needs a visible seam gap between 0.10 mm and 0.40 mm. That gap is your functional target. The dimensions that create it are then assigned tolerances based on their contribution to the gap and the capabilities of the selected process.
This is also where assembly method matters. Parts pressed together, parts joined with screws, components bonded with adhesive, and features designed for a light sliding fit should not be treated the same way. A screw-fastened assembly can often tolerate more positional variation than a press fit. Adhesive can fill a modest gap, but it may introduce cure shrinkage, mess, and inconsistent bond-line thickness.
Where stackups come from in real products
The most obvious contributors are the dimensions shown on a drawing, but physical products have more sources of variation than a basic CAD model suggests. Material behavior, process behavior, assembly sequence, and measurement method all matter.
For 3D printed parts, orientation is especially important. FDM parts can show different accuracy and surface character in the Z direction than across the build plane. Holes may print undersize, horizontal overhangs may sag, and long thin walls may warp. SLA can produce fine detail and smooth surfaces, but post-cure behavior, support placement, and resin selection can affect fit. SLS is well suited to durable, complex nylon components, though powder-based processing and finishing can influence surface condition and feature size.
Purchased components deserve the same attention. A nominal 8 mm bearing, magnet, heat-set insert, or circuit board is not necessarily exactly 8.000 mm. Review the supplier specification, not just the catalog headline. If the component is a critical interface, measure samples from the actual source before finalizing the mating feature.
Temperature and moisture can matter as well. Nylon absorbs moisture and can change dimensions. A part designed to fit comfortably in a climate-controlled shop may tighten or loosen in its end-use environment. For consumer, outdoor, or medical-adjacent applications, define the conditions the assembly must survive before selecting a material and clearance.
Tolerance stackup guide: two ways to calculate risk
The simplest approach is worst-case stackup analysis. Add every maximum condition in the direction that creates the tightest fit, then repeat for the loosest fit. This method guarantees that every in-spec part can assemble, assuming the stated limits are real and independent.
Imagine a lid closing over a component stack. The available internal height is 15.00 mm plus or minus 0.15 mm. The combined height of the board, standoffs, and component is 14.70 mm plus or minus 0.12 mm. At the tightest condition, available height is 14.85 mm and the internal stack is 14.82 mm. That leaves only 0.03 mm of clearance. It may technically fit, but it leaves little room for surface texture, part bow, assembly force, or an overlooked adhesive label.
At the loosest condition, available height is 15.15 mm and the internal stack is 14.58 mm, producing 0.57 mm of clearance. If the lid must restrain the component, that may be too loose. The analysis reveals both risks before physical parts are built.
A root sum square, or RSS, analysis is less conservative. It combines independent variations statistically by taking the square root of the sum of each tolerance squared. RSS can be appropriate for stable, well-understood production processes where measurements follow predictable distributions and a small percentage of edge cases is manageable.
For an early-stage product or low-volume run, worst-case analysis is often the better starting point. You may not yet have enough production data to justify statistical assumptions. A practical middle ground is to calculate worst case for safety-critical, retention, sealing, and hard-interference features, then use more flexibility for cosmetic gaps and noncritical covers.
Build the stack from datums outward
A useful stackup follows the same path that a functional relationship follows. Choose a datum, or stable reference, and trace dimensions from that reference to the feature that matters. Avoid stacking dimensions through unrelated surfaces when you can dimension each critical feature directly from a common datum.
For instance, if two holes must align with a circuit board, locate both hole centers from the same enclosure datum rather than locating the second hole from the first. Chain dimensioning lets error accumulate along the chain. Datum-based dimensioning controls the relationship that actually affects assembly.
Position tolerances often matter more than size tolerances. A hole can have the correct diameter yet miss a mating screw boss because its center is shifted. Likewise, two printed halves can have correct outer dimensions but fail to close because their locating pins and sockets are not controlled from compatible references.
When reviewing CAD, identify the few features that establish location: mounting holes, alignment pins, rails, bearing seats, mating faces, and sealing lands. These deserve deliberate tolerances and a clear inspection plan. Cosmetic outside surfaces usually do not need the same level of control.
Design clearance into printed assemblies
A nominal zero-clearance fit rarely behaves as zero clearance after printing. Surface texture, stair-stepping, support marks, shrinkage, post-processing, and local geometry all consume clearance. The right allowance depends on process, material, part size, orientation, and whether the feature is printed or machined after printing.
Treat fit allowances as starting points to validate, not universal rules. A small FDM hole intended for a screw may need to be modeled larger or drilled to size after printing. An SLA enclosure with a sliding cover may need extra room for post-cure variation and surface friction. An SLS nylon snap feature may need a different design approach than a rigid photopolymer snap feature because the materials deflect differently.
Print a focused fit coupon when the interface is important. A coupon might include several hole diameters, pin sizes, snap arms, or sliding channels in small increments. It is faster and less expensive to test those features on a compact sample than to learn about a poor fit after printing a full enclosure.
This testing also separates geometry problems from process problems. If every hole is undersize by a similar amount, compensate through design rules or secondary finishing. If only one area is distorted, look at wall thickness, orientation, heat buildup, support strategy, or local stress.
Turn the analysis into a build plan
A stackup is useful only when it changes a decision. Once the critical chains are identified, document the intended fit, nominal dimensions, tolerances, process assumptions, and inspection points. Flag dimensions that require post-processing, such as reamed holes, tapped features, or machined bearing pockets.
Then test the assembly in the order it will be built. Parts can fit individually but become difficult when a technician must install them around wires, fasteners, seals, or fragile components. Design for access to screws and tools, not only geometric compatibility.
At Tech Connext, this is the kind of issue best addressed before a project moves from prototype into a low-volume batch. Hold the parts, test the interfaces, record what changed, and use the results to set production-ready dimensions. A well-chosen tolerance is not extra paperwork. It is room for your product to work reliably when real materials, real people, and real manufacturing variation enter the picture.