Home / Notes / Factory Floor
Factory Floor

Why We Still Hand-Check Every Needle

A machine can tell you a needle exists. It cannot tell you whether it will hurt.

Melody
Melody Founder & CEO
Nov 20, 2026 · 7min read

There is a station on our line that has no name on the process sheet. It is a stainless table, a 3000K lamp, and a woman who has done this for nine years. Everything that comes off the moulding machine passes through her hands before it goes anywhere near a pouch.

Visitors usually ask two questions. The first is why we still do it. The second, after I answer, is whether they can skip it and save the cost.

What the machine is good at

Our in-line vision system is genuinely good, and I will not pretend otherwise. It measures needle length to ±0.01 mm, counts the array, checks that every tip is present, and rejects a head with a bent needle in about 90 milliseconds. For a line running 400 heads a day, that is not a nice-to-have. It is the reason we can hold a tolerance at all.

What it cannot do is judge the surface of the tip. A needle can be perfectly straight, perfectly long, and still have a microscopic burr from the grinding step. The burr is 4 to 8 microns. It does not affect depth. It does not show up in a dimension report. It does affect what the patient feels, and on a 0.5 mm stamp going into the forehead, they will feel it.

Three things she is looking for

  • The reflection test. Head tilted against the lamp at about 30 degrees. A clean tip throws a single point of light. A burred tip throws a smear. You cannot photograph this reliably; the camera's own lens introduces flare that looks identical to a burr.
  • The drag test on the pad. Each head is pressed once into a silicone pad at production pressure. A head with one short needle leaves a shallower dimple. That dimple is visible to the eye and invisible to a 2D camera looking straight down.
  • The seam. Where the needle plate meets the housing, there is a glue line. If it has crept more than about 0.3 mm inward, the head can trap serum and will be hard to clean. This is a cosmetic defect with a functional consequence, which is exactly the category that automated QC handles worst.

The number that matters

She rejects between 1.5 and 2.5 percent of a normal batch. On a bad day with a worn grinding wheel, closer to 4. That is not a small number, and it is not free — every rejected head is material we paid for.

If a supplier tells you their defect rate is 0.1 percent on microneedling heads and they have no manual station, ask to see the rework bin. There is always a bin.

We keep ours in the open. It is the least attractive part of the factory and the most useful thing a buyer can look at, because it tells you what the process actually catches rather than what the certificate claims.

What this costs you

Hand inspection adds roughly 2 to 3 US cents per head at our volume. On a 5,000-piece order that is about 150 dollars. I have lost orders over that number, and I understand why — on paper it looks like paying for something a camera already does.

The honest version: if your product is a 0.25 mm roller for home use, and your customers are price-driven, you may genuinely not need this step. The burr is smaller, the depth is shallower, and the tolerance for a rough experience is higher. If you are selling a 0.5 mm or 1.0 mm stamp into clinics, where one bad session costs your customer a client, three cents is the cheapest insurance on the whole bill of materials.

I would rather tell you which side of that line you are on than sell the same process to everyone.

Melody
Written by

Melody

Founder & CEO

In this industry since 2011 — the only one I have ever worked in. I answer every enquiry myself, and I sign off the batches that leave.

Factory Floor

Same series

Next step

Send the spec, get a real number

Not a catalogue price. The number after your quantity, your packaging and your port are in the calculation.