Case study · Ammo
We automated cartridge inspection for a premium ammunition maker.
- 180 PPM
- 30+ DEFECT TYPES
- 12+ SKUS
They wanted every round leaving the plant “looking like jewelry.”
Up to 60 hand inspectors across two shifts checked cartridges on curved, polished brass, where glare hides the flaws that matter and many calls rely on subjective judgment.
Inspection set the pace for the whole plant. They could make ammunition faster than they could inspect it, so rounds that couldn’t be checked right away went into bins and storage to wait their turn.
When we first met, they were skeptical, with good reason. For 20 years, vision companies had come through the door promising to automate their inspection. None of them had delivered.
Before
After
What we built
A complete inline inspection system, mounted at the end of their assembly line and proven there before they bought it.
Imaging
Engineered for curved, polished brass, where glare hides defects.
Defect models
Trained on their own cartridges. 30+ defect types, identified by name.
Severity scoring
Every defect is scored, so cosmetic and critical are handled differently.
Tiered thresholds
One line, different standards for different markets.
Automatic sorting
Good rounds go to packaging. Rejects go to scrap.
Reporting + alerts
By caliber, machine, defect type and time, with alerts when a threshold is crossed.
A sample of the defects the system catches.
The result
- 100%
Every round inspected inline at 180 parts per minute. No sampling, no bottleneck.
- 3 → 1
Loading, storage, and inspection and packaging now share one area of the floor.
- 60
Hand inspectors moved into other roles in the plant.
The data improved production, not just quality.
Once every round was counted and logged, the system started answering production questions too. For the first time, they could see exactly how much each shift produced, and one shift was producing nearly twice as much as the other. No one had known, because rounds went into bins to be dealt with later. With the numbers in front of them, they found what the stronger shift was doing differently and brought the other shift up to match.
We trained their team to run it themselves.
They didn’t want to rely on us for updates, retraining or new SKUs, so we made them self-sufficient.
- 01We defined the role and wrote the job posting.
- 02We interviewed candidates with them and helped hire two people.
- 03We trained the team for two months, from daily operation to model updates.
- 04Their team now runs and updates the system on its own.
Have an inspection problem nobody’s solved yet?
We’ll tell you whether we can solve it and what it would take.