Sensium’s Band G stack already published fifteen confusion cuts (Study 1), atlas size (Study 3), sensory vocabulary (Study 4), and evidence/bilingual trust (Study 5). This page is the next non-separator asset: structure fingerprint cohorts — reproducible groups of dossiers defined by the six structure scales every grape already carries. It answers a different question than “which pairs confuse?” It answers: which grapes share the same structural trap before aroma poetry starts?
Headline finding (export 2026-07-09): Under documented predicates, the catalog yields 218 dual acid+tannin reds, only 5 pale+high-tannin reds (Nebbiolo-class), 172 deep-color+high-tannin reds, 198 soft low-tannin reds, 291 high-acid whites, 28 light-body+high-acid whites, 10 low-acid aromatic whites, 9 full+low-acid whites, 57 alcohol=`high` dossiers, 489 high aromatic-intensity dossiers, and 97 rosé-frame dossiers.
This is still catalog research, not population miss-rates. Study 2 (topic `137`) remains volume-gated. Do not cite cohort sizes as “how often candidates miss Nebbiolo.” Cite them as the size of each structural teaching set the product encodes.
Companions: dual-structure pairs (1i) · how to taste tannin/acid/body · Study 3 atlas · Italian reds fingerprints.
Methodology (read this before citing)
| Field | Value |
|---|---|
| Source | `grapes.json` `structure` scales + `color` |
| High bands | `high`, `medium_plus` |
| Low bands | `low`, `medium_minus` |
| Order in tables | Rising `globalPopularityRank` (more planted first) |
| Export date | 2026-7 / 8 / 9 |
| Catalog size | 1,534 dossiers |
| Re-run | `node scripts/data/export_structure_fingerprint_cohorts.mjs --pretty --limit 12` |
| Not claimed | Live exam miss-rates, or that cohort membership equals “always tastes like this in every bottle” |
Study 1i filters confusionPair separator strings that name tannin and acid together. Study 6 groups dossiers by their structure fields — a grape can sit in a dual-acid-tannin cohort even when its first separator to a neighbor is fruit-shaped. Different objects; complementary drills.
Why structure cohorts after five other studies?
Google’s 2026 non-commodity bar still applies: dated predicates, re-run command, clear “not telemetry” label. Candidates fail when they smell Cabernet and write “Left Bank” before logging acid and tannin — or when they see pale Nebbiolo and invent Pinot. Structure cohorts turn the six scales into named teaching sets you can drill without inventing another aroma-pair cut.
Practically: open Compare inside one cohort, not across random famous names. Force the shared scales first (structure tasting).
Cohort map (all 11)
| Cohort | Count | Teaching question |
|---|---|---|
| Dual acid + tannin reds | 218 | Both pillars high before fruit? |
| Pale + high-tannin reds | 5 | Light look, hard grip? |
| Deep color + high-tannin reds | 172 | Inky + firm? |
| Soft low-tannin reds | 198 | Soft grip without full body? |
| High-acid whites | 291 | Acid drive first? |
| Light-body + high-acid whites | 28 | Lean and sharp? |
| Low-acid aromatic whites | 10 | Loud nose, soft acid? |
| Full-body + low-acid whites | 9 | Broad and soft-acid? |
| Alcohol = high | 57 | Heat/weight as fingerprint? |
| High aromatic intensity | 489 | Nose volume high/medium_plus? |
| Rosé-frame dossiers | 97 | Color frame = rosé? |
Teaching tables (top by planting rank)
Separators and aromas omitted on purpose — these tables are structure-only. Open the grape path for the full dossier.
Dual acid + tannin reds (218) — top 12
Predicate: `color=red` AND acidity high/medium_plus AND tannin high/medium_plus.
| Rank | Grape | Acid | Tannin | Body | Color | Arom. |
|---|---|---|---|---|---|---|
| 1 | Cabernet Sauvignon | medium_plus | high | high | high | medium_plus |
| 3 | Tempranillo | medium_plus | medium_plus | medium_plus | medium_plus | medium |
| 11 | Sangiovese | high | medium_plus | medium | medium | medium_plus |
| 12 | Bobal | medium_plus | high | medium_plus | high | medium |
| 13 | Carignan | high | medium_plus | medium_plus | high | medium |
| 29 | Nebbiolo | high | high | medium_plus | medium_minus | high |
| 30 | Cabernet Franc | medium_plus | medium_plus | medium | medium | medium_plus |
| 31 | Petit Verdot | medium_plus | high | high | high | medium |
| 37 | Touriga Nacional | medium_plus | high | high | high | high |
| 38 | Touriga Franca | medium_plus | medium_plus | medium_plus | high | medium_plus |
| 40 | Aglianico | high | high | high | high | medium_plus |
| 49 | Blaufränkisch | high | medium_plus | medium | medium | medium_plus |
Drill note: Nebbiolo appears here and in the pale cohort — that double membership is the exam trap.
Pale + high-tannin reds (5) — full set
Predicate: `color=red` AND colorDepth low/medium_minus AND tannin high/medium_plus.
| Rank | Grape | Acid | Tannin | Color depth |
|---|---|---|---|---|
| 29 | Nebbiolo | high | high | medium_minus |
| 150 | Xinomavro | high | high | medium_minus |
| 374 | Caino Bravo | medium | medium_plus | medium_minus |
| 574 | Amaral | medium | medium_plus | medium_minus |
| 721 | Ferrón | medium | medium_plus | medium_minus |
Five dossiers. Memorize the set. Do not invent a sixth from Pinot nostalgia.
Deep color + high-tannin reds (172) — top 8
| Rank | Grape | Tannin | Color depth |
|---|---|---|---|
| 1 | Cabernet Sauvignon | high | high |
| 3 | Tempranillo | medium_plus | medium_plus |
| 6 | Syrah | medium_plus | high |
| 12 | Bobal | high | high |
| 13 | Carignan | medium_plus | high |
| 17 | Mourvèdre | high | high |
| 18 | Malbec | medium_plus | high |
| 21 | Montepulciano | medium_plus | medium_plus |
Contrast with the pale set: same tannin pillar, opposite color frame.
Soft low-tannin reds (198) — top 8
Predicate: red + tannin low/medium_minus + body not high/medium_plus.
| Rank | Grape | Tannin | Body |
|---|---|---|---|
| 9 | Pinot Noir | low | medium |
| 19 | Cinsault | low | medium_minus |
| 33 | Gamay | low | medium_minus |
| 78 | Schiava | low | low |
| 79 | Marzemino | low | medium |
| 88 | Vaccarèse | low | medium |
| 99 | Poulsard | low | low |
| 112 | Frappato | low | low |
High-acid whites (291) — top 8
| Rank | Grape | Acid | Body | Arom. |
|---|---|---|---|---|
| 8 | Sauvignon Blanc | high | medium_minus | high |
| 10 | Trebbiano Toscano | high | medium_minus | medium_minus |
| 14 | Macabeo | medium_plus | medium | medium |
| 15 | Pinot Gris | medium_plus | medium | medium |
| 16 | Riesling | high | medium_minus | high |
| 20 | Chenin Blanc | high | medium | medium_plus |
| 34 | Albariño | high | medium | medium_plus |
| 35 | Verdejo | high | medium | medium_plus |
Low-acid aromatic whites (10) — full teaching spine
| Rank | Grape | Acid | Body | Arom. |
|---|---|---|---|---|
| 25 | Viognier | medium_minus | medium_plus | high |
| 26 | Gewürztraminer | medium_minus | medium_plus | high |
| 86 | Clairette | medium_minus | medium | medium_plus |
| 89 | Muscadelle | medium_minus | medium_plus | high |
| 219 | Bacchus | medium_minus | medium | high |
| 231 | Irsai Olivér | medium_minus | medium_plus | medium_plus |
| 241 | Muscat Ottonel | low | medium | high |
| 361 | Siegerrebe | low | medium_plus | high |
| 638 | Albillo Real | medium_minus | medium_plus | medium_plus |
| 937 | Vugava | medium_minus | medium_plus | medium_plus |
Pair with aromatic whites: perfume without acid drive is a named cohort, not a vibe.
High-alcohol dossiers (57) — top 8
| Rank | Grape | Alcohol | Color |
|---|---|---|---|
| 6 | Syrah | high | red |
| 7 | Grenache | high | red |
| 13 | Carignan | high | red |
| 17 | Mourvèdre | high | red |
| 22 | Zinfandel | high | red |
| 24 | Nero d'Avola | high | red |
| 25 | Viognier | high | white |
| 31 | Petit Verdot | high | red |
How Study 6 relates to Study 1i
| Study 1i | Study 6 | |
|---|---|---|
| Object | First separator strings naming tannin and acid | Dossier structure fields |
| Output | ~12 teaching pairs | Cohort membership lists |
| Use | Pair stop rules | Flight building inside a trap |
A grape can be dual-structure in Study 6 and still have a fruit-led first separator to a neighbor in Study 1o. Drill both.
How to cite this census
- Name it: “Sensium Study 6 (structure fingerprint cohorts), export 2026-07-09.”
- Link this URL.
- Name the cohort id (e.g. `pale_high_tannin_red`).
- Keep the label: catalog structure inventory — not miss-rates.
- Re-run `node scripts/data/export_structure_fingerprint_cohorts.mjs --pretty --limit 12`.
How this sits beside Studies 1–5 and Study 2
| Study 1 | Study 3 / Study 4 / Study 5 | Study 6 (this page) | Study 2 | |
|---|---|---|---|---|
| Object | Confusion cuts | Atlas / vocab / evidence | Structure cohorts | Wrong answers |
| Status | Complete | Drafted | This export | Volume-gated (`137`) |
A two-week drill
| Week | Focus | Habit |
|---|---|---|
| 1 | Red traps | Pale set (5) + dual top 12 + soft low-tannin top 8 — write scales before grape |
| 2 | White traps | High-acid top 8 + low-acid aromatic full 10 — perfume ≠ acid |
Service: home flights · Blind.
Frequently asked questions
Why only five pale high-tannin reds?
Because the catalog’s colorDepth + tannin encoding is strict. Nebbiolo and Xinomavro lead; three Iberian/Atlantic names complete the set. That scarcity is the teaching point.
Is Pinot Noir “dual acid + tannin”?
No — it sits in soft low-tannin under these predicates. Do not smuggle it into the Nebbiolo pale set.
Does cohort membership mean every bottle matches?
No. It means the dossier fingerprint encodes those bands. Site and winemaking still move real glasses.
When does Study 2 ship?
When anonymized Train/Blind volume clears a documented threshold. Until then, prefer Studies 1 and 3–6 — and do not invent miss-rate tables.
Bookmark this page beside the Study 1 field guide, drill the five pale high-tannin reds this week, and force acid+tannin lines in Blind before any place fantasy. Structure first — then stop rules — then fruit poetry.