How the plate check works

A green checked mark on a code means one specific thing: that blank has been compared against the drawing published in the FEFCO Code, 12th Edition, by machine, and it matched. 311 of our 479 templates carry it. This page explains exactly what was compared, and — just as importantly — what the check cannot tell you.

Why bother

A parametric generator can draw a box that folds, exports cleanly and looks entirely convincing, and still be the wrong box. Nothing about a tidy drawing tells you it matches the standard the code refers to. We know that because it happened here: this library once carried hand holes, vents and stacking tabs on codes whose plates draw none of them. They were invented, they looked plausible, and they survived until something measured them.

So a claim like "generated by our software" is circular — it says the picture agrees with itself. The only claim worth making is that the drawing agrees with the standard, and the only way to make that claim is to check.

What actually happens

1. The plate is read as geometry, not as a picture. The standard's own PDF is parsed into line segments. FEFCO draws creases in red and cuts in black, so the standard itself tells us which is which.
2. Our blank is generated from the template. The same code that produces your download.
3. The proportions are solved for, not assumed. A plate is drawn at whatever length, width and depth the standard chose; you will use your own. So the check searches for the dimensions that best reproduce the plate's crease layout. The question it answers is the one a badge should mean: do box dimensions exist for which this template draws this plate?
4. The cut line is then tested against those proportions. Fitting on the creases and testing on the cuts matters — using the same evidence for both would let the drawing be bent until it agreed with itself.
5. The verdict is the worst of the measurements, not the average. Cut coverage, crease coverage and the blank's outer profile are each scored in both directions. A code cannot pass by getting most of itself right and omitting one whole feature.

How we know the check is not simply agreeing with everything

A checker that says yes to everything is worse than no checker, because it produces confidence instead of information. So it is tested by being given the wrong plate: each template is scored against a sibling code's drawing, which it must fail.

The first version of this check accepted the wrong plate 53% of the time. That number is why the current one compares the cut line and the blank's outline rather than the folds alone — within a group the codes largely share their creases and differ in the cutting. It now records no genuine false matches across those pairs.

What the check cannot do

These are real limits, and a badge should be read with them in mind.

• A shared fold pattern is not enough on its own. Some plates carry identical crease layouts and differ only in a ratio, and because the proportions are solved for rather than assumed, one blank can be stretched onto the other's plate — our 0201 scores the same against 0202's drawing as against its own. A match then confirms the construction without saying which code was drawn, so nothing is claimed on the score alone. Those codes are settled separately, by scoring the blank against every sibling's plate and by comparing its proportions with the plate's: 0201's blank sits within 0.1% of its own plate's aspect and 5.4% away from 0202's, which is what separates them. Where that still cannot decide — where two plates really are one drawing under two numbers, or where a sibling remains as good a fit — no badge is granted.
• It is close to blind to a knife slit. The comparison is rasterised, and a slit has no width on the finished blank, so a short internal cut can be missing — or present when the plate has none — at very little cost to the score. Measured on the worst case we know of: B30.02 and B48.02 differ by four cuts, 11% of that plate's total cut length, and one drawing scores 1.000 against both. Where two codes differ only by small internal cuts, treat the badge as confirming the panels and folds rather than every lock.
• It cannot verify a flat piece. Many interior fitments — pads, liners, partition strips — have no folds at all, so there is nothing for a crease comparison to work with. Their outline is compared, which is weaker evidence, and they are not badged on it.
• It cannot judge profile detail. Whether a corner is radiused or chamfered, or a taper straight or swept, is below what it measures. A converter adjusts those anyway.
• It is about the flat drawing, not the 3D preview. The badge compares the blank against the plate. The folded picture beside it is a convenience, built by rotating panels about the lines the drawing scores — so where a box bends without a score, or a tab flexes through a slot, or two plies meet at a corner that board thickness would make room for, the model cannot show it. 3 of our 479 drawings say “preview approximate” on screen for that reason. A correct drawing can have an imperfect picture; the badge is about the drawing.
• It says nothing about your job. It compares two drawings. It knows nothing about your board, your machine, your product or your tolerances.

What a code without a badge means

Not that it is wrong. It means one of three things: it has not been checked yet, it has been checked and did not match, or it cannot be checked by this method — because its plate is flat, because it shares a fold pattern with a sibling, or because no published plate exists for it at all. Envelopes, bags, rigid set-up boxes and most trade styles are not in either standard, so there is nothing to check them against and there never will be.

We would rather tell you which of our 479 drawings have been checked than imply all of them have. 244 of the 325 FEFCO codes are built so far, and the count of checked ones goes up as the work continues.

Browse every FEFCO code, or open the generator.