A batch of springs looks fine. They meet the print, they pass a visual check, dimensions are within tolerance on the sample that got measured. They ship. Three months later, a customer’s product is failing in the field, and nobody can say why — because the thing that actually would have caught it was never checked in the first place.
That’s the pattern behind almost every expensive spring failure: not a dramatic single mistake, but a performance characteristic that dimensional inspection was never going to catch. Free height and wire diameter tell you a spring was made correctly. They don’t tell you it will hold up under load, cycle after cycle, for the life of the product it’s going into.
The Testing Shortcut Nobody Talks About
Full performance testing costs time and money that dimensional inspection doesn’t. Under schedule pressure, it’s easy to treat a clean print check as good enough, especially on a part that’s “just a spring” in a much larger assembly. The assumption is that if the geometry is right, the performance will follow.
That assumption holds until it doesn’t. Spring rate, fatigue life, torque output, and solid height are all things that can drift or vary even when the part looks correct on paper — from wire lot to wire lot, from coiling machine wear, from heat treat variation, from a supplier substitution nobody flagged as significant. None of that shows up on a caliper.
Two Recalls That Started With an Untested Spring
A Material Shortcut, 56,000 Vehicles
In 2019, Volkswagen recalled 56,173 vehicles — Golf, Golf Sportwagen, Jetta, and Tiguan models from 2015 through 2019 — because rear coil springs could fracture while the vehicle was being driven. VW’s own filing with NHTSA (recall 19V-188) is specific about what happened: a supplier, Thyssenkrupp Components Technology in Mexico, used a material that hadn’t been approved for the part, and ran certain production batches through incorrect process parameters during Vanadium processing. The result was springs that met visual and dimensional expectations but had compromised long-term durability — exactly the kind of defect that a fatigue or material-verification test catches and a dimensional check doesn’t.
This is about as clean a case as exists: a named supplier, a documented process deviation, a batch-specific defect that shipped anyway, and a six-figure vehicle count as the consequence.
328,000 Vehicles, Three Recalls, One Spring
Jeep’s Grand Cherokee rear coil spring recall is worth mentioning for scale, with an honest caveat. In 2023, Stellantis recalled roughly 328,000 Grand Cherokee and Grand Cherokee L vehicles because rear coil springs could come loose while driving. What makes it notable is that this recall superseded two earlier recall attempts at the same underlying problem — the fix didn’t hold the first two times. Stellantis hasn’t published the specific failure mechanism in detail, so it’s not as clean a “testing would have caught this” story as the VW case. But it’s a useful data point on its own: even a well-resourced OEM can take three tries to close out a spring-related field failure once it’s loose in the supply chain, which says something about how hard these problems are to run down after the fact instead of catching them before shipment.
The Failures That Never Make the News
Most spring failures never become a public recall. They show up as a warranty claim, a line stoppage, or a quiet loss of a customer’s next order. These aren’t documented incidents — they’re common failure patterns worth watching for, built from how spring performance problems typically surface in production.
Torque That Isn’t There
A torsion spring in a safety guard or closure mechanism meets its target torque on the units that get checked, but output varies more than anyone realizes across the run — because torque was being judged by feel rather than measured consistently at a given angle of deflection. Somewhere in the field, a unit near the low end of that unmeasured spread doesn’t perform the way the mechanism depends on it to. The cost isn’t just the fix. It’s the inspection of everything else in the field, and a customer who starts asking whether your process can be trusted on the next order.
The Drift Nobody Was Watching
A compression spring run holds spec on the first-article sample and on the periodic checks that follow. But spring rate creeps slowly over the course of the run — coiling tooling wears, wire tension shifts a little at a time — and because testing is periodic rather than continuous, nobody sees the trend until it’s well outside tolerance. It gets caught, if it gets caught at all, at final assembly on the customer’s line: the most expensive place in the whole supply chain to find a defect, because now it’s a full lot sort instead of a process adjustment.
The Line That Stops Without Warning
An ejector spring in a customer’s injection molding tool passes solid-height verification on the first-article sample, but solid height isn’t re-checked as the run continues. A later batch is out of spec just enough that it behaves differently under repeated cycling. The customer’s press goes down mid-run, unplanned, and now it’s their downtime and your relationship on the line — not a defect that cost pennies to catch upstream.
What Actually Catches These Before They Ship
Every one of these failure modes maps to a specific test that dimensional inspection doesn’t perform:
Free height and preload testing on full production runs, not just samples, catches the drift that a first-article check misses.
Fatigue and cyclic testing validates the spring will survive the number of cycles the application actually demands — not just that it meets a static load number on day one.
Torque-at-angle testing with data logging replaces “feel” with a number, on every unit, so torsion output variation shows up before the part ships instead of after it fails.
Solid-height verification carried through the full run, not just first article, catches the kind of batch drift that took down the injection molding line in that last scenario.
SPC-linked, automated testing turns individual test results into a trend line, so a slow drift in spring rate gets flagged while it’s still a process adjustment — not a customer’s assembly-line problem.
This is the same ground covered in the testing fundamentals we’ve published on fatigue testing, force-deflection curve interpretation, and torsion spring testing — worth linking to here for readers who want the mechanics behind any one of these.
Doing the Math on Testing Versus Failure
Quality management has a long-standing rule of thumb: a defect caught at the design or process stage costs roughly one unit to fix. Caught in production, it costs roughly ten. Caught after it’s shipped and in the customer’s hands, it costs roughly a hundred — in rework, recall logistics, liability exposure, and the business you don’t get to keep. The VW recall didn’t cost Volkswagen the price of a fatigue test. It cost them 56,000 vehicles’ worth of dealer labor, parts, and reputational damage, on a defect that a properly specified test protocol is built to catch before a single spring leaves the building.
Where Testing Fits in Your Process
The fix isn’t more inspection at the end of the line. It’s testing built into the process at the points where these failure modes actually originate: verifying incoming material and early production the same way VW’s supplier should have, tracking performance data continuously instead of periodically, and treating torque, fatigue life, and solid height as production specs to be measured on every run — not just design targets to be met once during qualification.
That’s a process change, not just an equipment purchase, but it starts with having testing equipment that makes continuous, full-run verification practical instead of a bottleneck. If you’re relying on manual checks and periodic sampling because that’s what your current testing setup can handle, that’s usually the actual constraint — not the testing philosophy.
If you want to talk through where a testing gap might be sitting in your own process, that’s a conversation RK Trading is glad to have.