Snap fits are the cheapest fastening method in plastics — no screws, no inserts, no secondary operations — and they are also the feature most likely to fail on the first drop test. The difference between a snap that clicks 10,000 times and one that cracks on assembly is almost never the concept. It is strain, beam geometry, and whether the undercut was designed for the tool that will actually cut it.
The four snap types worth knowing
Cantilever snaps are the workhorse: a beam molded into a wall that deflects and returns. They suit small to medium plastic housings and are easy to model. U-latches are a double-cantilever variant that doubles the effective beam length inside the same package height, which matters when you have no room for a long beam. Torsional snaps use a twisting bar and are compact but harder to mold consistently. Annular snaps — the ring-on-shaft style used on caps and pen lids — give a 360° hold but demand tight concentricity between core and cavity.
For plastic injection prototypes, cantilever and U-latch designs cover most of what engineers actually need. Start there.
Strain is the number that decides whether it survives
A snap works by bending plastic below its elastic limit. Every material has an allowable strain, and the geometry has a maximum strain it will experience. If the second exceeds the first, you get stress whitening, then a crack — usually at the root fillet, usually on the first assembly.
For a straight cantilever of uniform thickness, the maximum strain at the root is:
ε = 1.5 × t × y / L²
where t is beam thickness at the root, y is the deflection needed to clear the undercut, and L is the effective beam length. The 1.5 collapses to about 0.7 for a beam tapered in thickness from root to tip — which is why tapered beams are standard practice: same deflection, roughly half the strain, and less material.
Practical allowable strain for a single assembly at room temperature, with a well-radiused root:
- PC: 4% (unfilled), drops sharply with notch sensitivity — avoid sharp internal corners
- ABS: 2.5–3%
- PP: 6% or more, and it tolerates repeated flexing better than anything else in this list
- PA66: 6–8% dry, but it absorbs moisture and properties shift; design to the conditioned value
- POM: 8% at low strain rates, excellent fatigue, poor for anything you need to glue
- Glass-filled grades (PA66-GF30, PBT-GF): treat as 1.5–2%. The fibers do not stretch, and every snap becomes a notch test
If the part must be assembled and disassembled repeatedly, cut the allowable strain roughly in half and add a lead-in radius on the return face so the beam is not gouged each cycle.
Geometry rules that keep strain in range
Length-to-thickness ratio. Because strain scales with 1/L², beam length is your strongest lever. Aim for L/t between 8 and 15. Below 8 the strain climbs fast; above 15 the beam gets floppy in the other direction and can be damaged by handling.
Root fillet. A fillet of 0.5 × t at the root is the single highest-value change you can make. Sharp roots concentrate stress and turn a 3% design into a failure at 1.5%.
Retention depth. 0.3–0.5 mm of engagement is enough for most small housings. Deeper engagement increases y, which increases strain quadratically — do not add 1 mm of engagement "for safety."
Lead-in and return angles. A 25–30° lead-in lets the mating part push the beam over gradually. The return face — the side that holds — should be 60–90° so it resists pull-out. Two symmetric 45° faces will click nicely and then release under load.
Support the base. A beam hanging off a 1.5 mm wall will flex the wall instead of the beam. Add a rib or a local boss at the root, and keep the beam axis aligned with the direction of mold opening so it does not need a side action.
The undercut question, and what it costs
Every snap fit is an undercut. A cantilever that deflects in the direction of mold opening needs nothing extra — the mold halves simply pull apart. The moment the hook faces sideways, you need a slider, a lifter, or a strip-off, and that changes both tooling cost and lead time.
For prototypes, the cheapest path is usually to orient the snap so it releases in the pull direction, or to accept a short-lived aluminum tool with a simple insert. Where the geometry truly will not cooperate, a secondary operation — machining the hook after molding — is sometimes cheaper than a slider for 50 parts, but it will not survive the transition to production. Decide early whether the design is a prototype shortcut or the production intent.
If you are still choosing a process, soft production tooling lets you test a real snap in real resin at low volume, and 3D printing is fine for fit checks but misleading for snap force — layer adhesion and anisotropy make printed snaps either too stiff or too brittle, rarely representative.
Test it the way it will fail
Two tests catch nearly every bad snap. First, assemble and disassemble 20 times by hand and inspect the root under magnification for whitening; whitening means you are above the material's yield and the next step is a crack. Second, a drop test from the height the product will actually see, with the housing closed and loaded. Snaps that pass assembly often fail on impact because the engaged beam carries the whole housing load.
If a snap is borderline, change the geometry before changing the material. Switching from ABS to PC to buy strain allowance usually costs more per part than adding 2 mm of beam length — and on a CNC machined prototype of the housing, a longer beam is a five-minute program edit.
Send the snap detail with the rest of the drawing and say what the part has to survive. A supplier who knows the assembly cycle and the drop height will design to the right strain instead of guessing — and that conversation is cheaper than a second tool.
