ShadeVeto

Catalog coverage, brand view

Study data only: this screen counts the decision engine's verdicts over the six synthetic study tones, ranked against the licensed catalog: 617 real foundation shades from The Pudding foundation shades dataset, swatches recorded in 2018, used under the MIT License. No API call is made and no units are spent. It is a coverage report on a small test set, not a claim about shoppers in general.

A shopper-facing check answers one person. This screen answers the brand: if faces like the six study tones came to this catalog, how many would leave with a shade the engine can stand behind, and for the rest, exactly why not? Every figure below traces to a receipt the engine produced in this same computation, one per tone, listed in full.

Coverage gap

3 of 6 study tones had no safe BUY in this catalog.

Counted separately, never merged: SHADE-RISK 3 of 6, NO 0 of 6, CAN'T TELL 0 of 6. A shade risk, an empty range and an unreadable gap are three different problems, with three different fixes.

Verdicts across the six study tones

VerdictShare of study tonesWhich tones
BUY3 of 6 (50%)tone1 (deep espresso), tone2 (warm medium), tone3 (golden tan)
SHADE-RISK3 of 6 (50%)tone4 (olive tan), tone5 (light beige), tone6 (natural porcelain)
NO0 of 6 (0%)none
CAN'T TELL0 of 6 (0%)none

Every tone, every receipt

ToneReading (skin_color)Face pixels (study truth)VerdictReceipt headline
tone1 (deep espresso) #6e4c37 #663d2cBUY
cluster-match		Buy: top pick is Licensed catalog 724A32 at 2.03; 14 shades sit within measurement noise of each other here and all of them match your measured tone. This is a cluster pick, not a unique winner.
tone2 (warm medium) #a77d60 #bc7d58BUY
cluster-match		Buy: top pick is Licensed catalog AD7D59 at 2.37; 26 shades sit within measurement noise of each other here and all of them match your measured tone. This is a cluster pick, not a unique winner.
tone3 (golden tan) #af8569 #d39c77BUY
cluster-match		Buy: top pick is Licensed catalog B2856F at 2.85; 28 shades sit within measurement noise of each other here and all of them match your measured tone. This is a cluster pick, not a unique winner.
tone4 (olive tan) #af8e73 #b58669SHADE-RISK
candidates-within-noise		Shade risk: Licensed catalog B19277 matches your measured tone at 1.39, but 8 shades within measurement noise of it fall outside the match range of 6 (closest: Licensed catalog C1A07E at 6.02). The reading cannot separate a match from a near miss here.
tone5 (light beige) #b29078 #d4a58cSHADE-RISK
candidates-within-noise		Shade risk: Licensed catalog B19277 matches your measured tone at 2.02, but 16 shades within measurement noise of it fall outside the match range of 6 (closest: Licensed catalog C1A07E at 6.1). The reading cannot separate a match from a near miss here.
tone6 (natural porcelain) #b7937d #e4b7afSHADE-RISK
candidates-within-noise		Shade risk: Licensed catalog C09578 matches your measured tone at 2.85, but 50 shades within measurement noise of it fall outside the match range of 6 (closest: Licensed catalog CB9070 at 6.08). The reading cannot separate a match from a near miss here.

What this screen is

  • How this was counted: the same decision engine that prints the check page receipt ran once per study tone, with the study framing that passed and no owned bottles listed, and the verdicts on those six receipts were counted. Nothing on this screen is computed any other way.
  • Study set: the six tones are the synthetic faces from the C088 measurement study, each generated to a known skin tone. The set is small by design: six tones, one per band the study covered.
  • Catalog: the shades ranked here are a real, licensed catalog of 617 foundation shades: The Pudding foundation shades dataset (Beauty Brawl, June 2018), shades.csv, https://github.com/the-pudding/data/tree/master/makeup-shades. The swatch colours were recorded in 2018 and are used under the MIT License (The Pudding, the-pudding/data repository), so they are dated reference data, not a reading of today's shelf. Display labels carry shade hex codes only; brand and product names from the dataset are never shown on this screen.
  • These are study results on six synthetic faces, not population statistics. They describe how this catalog covers this test set, and nothing wider.
  • Every share is a count of 6 first. The percentage beside it is that same count rounded to the nearest percent, never a separate measurement.