Scope & coverage

Answers to the three questions Jason raised on 23 July: which colors actually need Lab values, where the Pantone values can come from, and whether unused swatches can be stripped automatically. Every number here is measured from the live A2000 color master and two months of real Color Direct art — nothing is estimated.

1 · Which colors actually need a Lab value

A2000's color master carries every decoration method in one list, which is why the browser looks like 3,000+ colors. Its own ART_COLOR_TYPE already separates them — thread, flock, twill, rhinestone and foil follow their own RIP protocols and never need a Lab value.

colors in A2000

what the CMS browser lists today

digital-relevant

type INK or PMS

other methods

thread, flock, twill, foil… — filter these out

house inks actually used

distinct INK codes in 2 months of art

The working list is inks, not

Across every Color Direct file in the window, only distinct house ink codes appear — and of them (%) now have a canonical Lab value. Not one thread, flock, twill, rhinestone or foil color showed up in a single digital file, so the concern about specialty swatches is real for the browser but never reached the standard: the CMS holds values only for INK and PMS colors, and no other decoration method has ever been given one.

Filtering the CMS art-colors browser to ART_COLOR_TYPE IN ('INK','PMS') removes rows (%) from view. That is a display filter — no color data is deleted.

The entire remaining backlog inks

House inks that appear in real Color Direct art but still have no canonical Lab value. This is the complete list — there is no long tail behind it.

Everything else is now valued. INK006 and INK013 were resolved in the CMS on 23 July (LAB Black (Spot) and LAB White), and the Pantone seed below filled in the PMS side, so the standard went from 215 colors to .

2 · Pantone Lab values — solved

Done — no sourcing effort needed, and Kaitlyn's books closed the gap

Two independent sources already carry Pantone's own Lab values, so nothing has to be looked up or translated. The Adobe book Tim supplied on 15 July (MV Sport PMS C.acb, colors) got us to 96%; the two .ase books Kaitlyn supplied on 23 July — Solid Coated and Solid Uncoated, both stored as Lab spot colors — supersede it.

They now cover of the distinct Pantone swatch names (%) in real Color Direct art and % of all Pantone swatch instances. Where the two books overlap they agree to a median ΔE 1.07, which is byte-rounding noise — so this is corroboration, not a coin flip.

These values are now seeded into the CMS itself — PMS colors written on 24 July with source = 'pantone', so the standard covers Pantone as well as house inks. Together with the black/white values added on 23 July, ungradeable files fell from to — a % reduction — with nobody hand-entering a color. Jay does not need to start this work.

The seed is deliberately conservative: puff, halftone and translucent variants are excluded (a puff ink prints differently from its base, so inheriting a flat book value would assert something false at press), existing hand-entered values are never overwritten, and the Pantone number in each A2000 description must agree with its color code before anything is written. That last check caught 7 rows where the two disagree — e.g. PMS218 is described as "PMS248 Azalea" and PMS371 as "PMS357 Olive". Those are worth a human look.

The uncoated book matters more than its size suggests: only four uncoated swatch instances appear in the corpus, but grading a U color against its C twin is wrong by ΔE 17.7–22.9. The audit now keys on the book suffix so that substitution can't happen silently.

Pantone colors in use

distinct names across the corpus

covered by the book

% of names

still missing

swatch instances total

The residual gap names

The complete list of Pantone swatches we still cannot value. None are Solid Coated or Solid Uncoated misses — they are other books entirely (TCX is Pantone's textile system, P 7-8 C and 179-12 C are from the CMYK book), legacy names Pantone has retired (Process Black C, Medium Yellow C), or a swatch written without its book suffix. Between them they account for swatch instances across the whole corpus.

3 · Stripping unused swatches

Measured on a random 140-file sample by comparing each file's swatch book against the spot colors actually painted into the artwork.

files already clean

of — nothing to purge

of swatches unused

of declared

swatches actually painted

across the sample

Yes — and it can be detected reliably

An .ai file is a PDF. The swatch panel lives in the file's XMP metadata, but a spot color that is actually painted must also appear as a /Separation or /DeviceN colorspace. Comparing the two sets identifies unused swatches exactly, with no guessing.

The check validates itself: on files whose designer already ran Illustrator's purge, the two counts agree precisely (19/19, 23/23, 24/24 observed). % of files are already clean — the waste is concentrated in a minority of files that carry Illustrator's untouched default palette (C=0 M=0 Y=0 K=100, CMYK Red, …), which alone accounts for of the swatch entries in the corpus.