{ "format": "hyperize/v1", "@id": "https://www.hyperize.ai/en/dax40-index/brands/merck.json", "type": "BrandScore", "pageType": "BrandScore", "releaseState": "pilot_probe", "releaseLabel": "Pilot Probe", "title": "Merck — Pilot Probe (Single-Task Observation, Wave Q2 2026)", "pilotProbe": { "status": "pilot_probe", "rationale": "Single-task observation under a narrower provider track than a full BrandScore. The Agent Success Score is computed but should be read as an early signal, not a fully-qualified score.", "promotionCriteria": "Confidence promotes to B (and releaseState to brand_score) when the next wave adds a second task and the fourth provider on the Discoverability track.", "nextWave": "WAVE-Q3-2026" }, "hook": { "en": "Sigma-Aldrich runs one of the cleanest B2B product catalogues in the DAX 40 and the simplest agent classes never see it. A plain reader times out before any response arrives. A coding agent gets HTTP 000. A browser reaches the page, finds TRI Reagent BD T3809 for whole-blood RNA at 200µl, adds three sizes to cart without login. The full-automation run completes the same path. The catalogue is reference-grade. The HTTP wall is the entire story.", "de": "Sigma-Aldrich betreibt einen der saubersten B2B-Produktkataloge der DAX 40 und die einfachsten Agent-Klassen sehen ihn nie. Ein einfacher Reader läuft in einen Timeout, bevor eine Antwort kommt. Ein Coding-Agent bekommt HTTP 000. Ein Browser erreicht die Seite, findet TRI Reagent BD T3809 für Vollblut-RNA bei 200µl, legt drei Größen ohne Login in den Warenkorb. Der Full-Automation-Run macht denselben Pfad. Der Katalog ist Referenzqualität. Die HTTP-Wand ist die ganze Geschichte." }, "summary": { "en": "A reference-grade B2B catalogue that text and code agents can't reach at all.", "de": "Ein Referenz-B2B-Katalog, den Text- und Code-Agenten gar nicht erreichen." }, "brandName": "Merck", "legalEntity": "Merck KGaA", "daxTicker": "MRK", "sector": "pharma_healthcare", "sectorLabel": "Pharma & Healthcare", "inLanguage": "en", "datePublished": "2026-05-18", "dateModified": "2026-05-24", "nextReview": "2026-09-30", "confidence": "C", "evidenceTier": "proprietary", "wave": "WAVE-Q2-2026-W5-LIFE-SCIENCES", "url": "https://www.hyperize.ai/en/dax40-index/brands/merck", "alternateLanguage": { "de": "https://www.hyperize.ai/de/dax40-index/brands/merck" }, "isPartOf": { "@id": "https://www.hyperize.ai/en/dax40-index", "name": "DAX 40 Agent Success Index" }, "lanes": [ { "id": "commerce", "name": "Commerce", "measured": true }, { "id": "talent", "name": "Talent", "measured": false }, { "id": "aftersales", "name": "After-sales", "measured": false }, { "id": "procurement", "name": "Procurement", "measured": false }, { "id": "investor_relations", "name": "Investor Relations", "measured": false }, { "id": "press", "name": "Press", "measured": false } ], "commercialLane": { "visibility": { "question": "Will an agent find and recommend the brand?", "score": 38.33, "dimension": "Discoverability (D)", "methodVersion": "gate1/v1.1", "sampleSize": 18, "providers": [ "openai", "perplexity", "anthropic" ], "provenance": "measured" }, "usability": { "question": "Once an agent arrives — how far does it get toward the deal?", "score": 54, "scoreNote": "Derived from the access profile below + close state (usability-derivation/v1). NOT a hand-averaged number — a reproducible summary of observed agent behaviour.", "accessProfile": [ { "agentClass": "text", "outcome": "blocked" }, { "agentClass": "search", "outcome": "pending" }, { "agentClass": "code", "outcome": "blocked" }, { "agentClass": "browser", "outcome": "success" }, { "agentClass": "full_automation", "outcome": "success" } ], "blockedClasses": [ "text", "code" ], "passClasses": [ "browser", "full_automation" ], "closeState": "cart_ready", "closeStateNote": "Reaches a working cart with no login gate — the transaction path is real, not just order-ready. Final payment is a separate step.", "methodVersion": "fleet/measured", "provenance": "measured" }, "evidence": { "score": 68, "basis": "estimate — cross-method ground-truth consistency, not a separate measured run", "note": "A confidence layer on the two axes above, not a third sales axis." } }, "score": { "compositeForIndexRanking": { "value": 52.27, "scale100": 52.27, "scale10": 5.2, "status": "computed", "formula": "Agent Success Score = (AI Visibility × 0.20) + (AI Usability × 0.70) + (Evidence × 0.10)", "methodVersion": "ars-methodology/v1.1", "usabilityDerivation": "usability-derivation/v1", "role": "index-ranking", "note": "Weighted composite, used for ranking the index. Read the two commercialLane axes (visibility + usability) for the standalone verdict — this single number blends them and is not the headline." }, "dimensions": { "aiVisibility": { "name": "AI Visibility (Discoverability)", "weight": 0.2, "value": 38.33, "methodVersion": "gate1/v1.1", "sampleSize": 18, "providers": [ "openai", "perplexity", "anthropic" ], "queryVariants": 3, "runsPerVariant": 2, "taskCount": 1 }, "aiUsability": { "name": "AI Usability (reach + completion)", "weight": 0.7, "value": 54, "derivation": "usability-derivation/v1", "basis": "derived from the access profile (per-class outcomes) + close state" }, "evidence": { "name": "Evidence", "weight": 0.1, "value": 68, "methodVersion": "fleet/measured" } } }, "thirdPartyInterception": { "classification": "none", "classificationLabel": "None", "directShare": 0, "intermediaryShare": 0, "intermediaryCaptureExamples": [], "narrative": { "en": "Channel position derivation on the Wave 5 Cody Gate-1 response texts is queued; the displayed share values are the unmeasured default. The measured task closes directly on Sigma-Aldrich (Merck's own catalogue), so the close-state is brand-owned rather than intermediated; the discovery competition with peer catalogues (Qiagen, Thermo Fisher, Bio-Rad) will be characterised in the next pass.", "de": "Channel-Position-Ableitung aus den Wave-5-Cody-Gate-1-Antworttexten ist eingereiht; die angezeigten Share-Werte sind der ungemessene Default. Der gemessene Task schließt direkt auf Sigma-Aldrich (Mercks eigenem Katalog), der Close-State ist also markenbezogen, nicht intermediär; die Discovery-Konkurrenz mit Peer-Katalogen (Qiagen, Thermo Fisher, Bio-Rad) wird im nächsten Pass charakterisiert." } }, "bottleneck": { "type": "visibility", "classification": { "en": "Discovery", "de": "Discovery" }, "classificationNote": { "en": "Rarely found for the category.", "de": "Wird für die Kategorie selten gefunden." }, "sentence": { "en": "A reference-grade B2B catalogue that text and code agents can't reach at all.", "de": "Ein Referenz-B2B-Katalog, den Text- und Code-Agenten gar nicht erreichen." }, "executiveSummary": { "en": "The B2B path is built. The protection layer above it deletes half the agent classes before they see anything. Browser-class research succeeds end to end; everything below it hits an aborted connection.", "de": "Der B2B-Pfad ist gebaut. Die Schutzschicht darüber löscht die Hälfte der Agent-Klassen, bevor sie überhaupt etwas sehen. Browser-basierte Recherche gelingt end-to-end; alles darunter trifft auf eine abgebrochene Verbindung." } }, "testNarrative": { "verdict": { "en": "A €154 reagent you can order with a browser, invisible to anything simpler.", "de": "Ein €154-Reagenz, das man im Browser bestellt, unsichtbar für alles Einfachere." }, "narrative": { "en": "We asked five kinds of AI agent to find an RNA-extraction kit for human whole blood at 200µl on Sigma-Aldrich. The plain reader's connection aborted before any response landed. The coding agent's curl returned HTTP 000, no file written. The browser navigated to sigmaaldrich.com/DE/de/product/sigma/t3809, identified TRI Reagent BD (catalog T3809), confirmed whole-blood compatibility, and added a 25mL, 100mL, and 200mL pack to cart, with the cart counter ticking from 0 to 1 and an alert confirming \"1 Artikel wurde in den Warenkorb gelegt.\" No login required. The full-automation run completed the same flow.", "de": "Wir haben fünf Typen von KI-Agenten auf Sigma-Aldrich geschickt, ein RNA-Extraktions-Kit für humanes Vollblut bei 200µl zu finden. Die Verbindung des einfachen Readers brach ab, bevor eine Antwort kam. Das curl des Coding-Agenten lieferte HTTP 000, keine Datei geschrieben. Der Browser navigierte zu sigmaaldrich.com/DE/de/product/sigma/t3809, identifizierte TRI Reagent BD (Katalog T3809), bestätigte Vollblut-Kompatibilität und legte eine 25mL-, 100mL- und 200mL-Packung in den Warenkorb, der Cart-Counter sprang von 0 auf 1 und ein Alert bestätigte \"1 Artikel wurde in den Warenkorb gelegt\". Kein Login nötig. Der Full-Automation-Run machte denselben Flow." }, "scopeShift": { "en": "This is one reagent on one surface (Sigma-Aldrich). Merck Life Science sells hundreds of thousands of SKUs across reagents, chemicals, analytical, biopharma and lab equipment, each on parametric-search-driven product pages.", "de": "Das ist ein Reagenz auf einer Surface (Sigma-Aldrich). Merck Life Science verkauft hunderttausende SKUs über Reagenzien, Chemie, Analytik, Biopharma und Laborausrüstung, jeweils auf parametrisch-such-getriebenen Produktseiten." }, "agentMatrix": [ { "type": "text", "status": "blocked", "note": "Connection aborted at sigmaaldrich.com (likely Cloudflare or Akamai); zero response received" }, { "type": "search", "status": "pending", "note": "Wave 5 Cody Gate-1 in progress" }, { "type": "code", "status": "blocked", "note": "curl returned HTTP 000 (connection failed); no file written" }, { "type": "browser", "status": "success", "note": "3 of 3 runs reached T3809 product page, Score avg 96.3" }, { "type": "full_automation", "status": "success", "note": "ACT added to cart without login (Warenkorb 1, alert confirmed), Score 100" } ] }, "fairnessDeclaration": { "reviewPassed": false, "note": "Wave 5 Q2 2026 audit complete on a single task (TRI Reagent BD T3809, cart-ready close state). AI Visibility from Cody Gate-1 on an unbranded informational probe (RNA-Extraktion-Vollblut-Lab-Anbieter). AI Usability from a fleet wave (2026-04-07) full phase 1-4. Fairness Review pending the sector fairness grid.", "methodologyUrl": "https://www.hyperize.ai/en/methodology/task-selection", "sectorGridRef": null }, "activeTasks": [ { "title": { "en": "TRI Reagent BD T3809 for whole-blood RNA extraction (200µl)", "de": "TRI Reagent BD T3809 für Vollblut-RNA-Extraktion (200µl)" }, "closeState": "cart_ready", "intermediaryMode": "none", "businessRelevance": null, "difficultyBand": null, "engpass": { "en": "WAF blocks the plain reader and coding agent at the connection layer; browser and full automation reach the cart without login.", "de": "WAF blockiert den einfachen Reader und den Coding-Agenten auf der Connection-Layer; Browser und Full Automation erreichen den Warenkorb ohne Login." } } ], "scoreHistory": [ { "date": "2026-04-07", "quarter": "q1-2026", "waveId": "a fleet wave", "protocolVersion": "fleet/measured", "summary": "First-pass FLEET measurement: TRI Reagent BD T3809 on Sigma-Aldrich across ceiling + http + coding + browser-agent + act breeds. Found connection-layer blocking for text/code agents (HTTP 000); browser success 3/3 with Score 96.3 average; ACT cart-ready with no login gate." }, { "date": "2026-05-24", "quarter": "Q2 2026", "waveId": "WAVE-Q2-2026-W5-LIFE-SCIENCES", "protocolVersion": "ars-methodology/v1.1", "summary": "Wave 5 kickoff. Cody Gate-1 audit running on unbranded informational probe (RNA-Extraktion-Vollblut). AI Usability carried from a fleet wave full phase 1-4 access profile." } ], "methodology": { "formula": "Agent Success Score = (AI Visibility × 0.20) + (AI Usability × 0.70) + (Evidence × 0.10)", "usabilityDerivation": "usability-derivation/v1", "protocolVersion": "ars-methodology/v1.1", "taskSelection": "https://www.hyperize.ai/en/methodology/task-selection", "agentSurface": "https://www.hyperize.ai/en/methodology", "dimensions": { "aiVisibility": "AI Visibility (Discoverability) — audit-pipeline-derived. 3–4 providers × 3 query variants × 2 runs.", "aiUsability": "AI Usability — derived from the per-class access profile (which agent classes reach the close state) modulated by close-state depth. The profile is the truth; the score is a reproducible summary, not a hand-rating.", "evidence": "Evidence — cross-method ground-truth consistency. A confidence layer, not a sales axis." } }, "consequenceClose": { "archetype": "discovery", "archetypeLabel": { "en": "discovery", "de": "Discovery" }, "source": "brand", "diagnosis": { "en": "On B2B life-sciences sourcing, Merck is not lacking a catalogue, T3809 carries the kind of structured product data the Hyperize cookbook calls reference-grade. It is losing the agent classes that don't run a browser. WAF protection at sigmaaldrich.com aborts the HTTP connection before any product content is served, the simpler agents that LLM tools call out to first see nothing. A specifier who lets ChatGPT do the lookup against a text-only fetcher gets routed to Qiagen, Thermo Fisher, or Bio-Rad by default.", "de": "Im B2B-Life-Sciences-Sourcing fehlt Merck kein Katalog, T3809 trägt die Art strukturierter Produktdaten, die das Hyperize-Cookbook Referenzqualität nennt. Es verliert die Agent-Klassen, die keinen Browser fahren. WAF-Schutz auf sigmaaldrich.com bricht die HTTP-Verbindung ab, bevor irgendein Produkt-Content ausgeliefert wird, die einfacheren Agenten, die LLM-Tools zuerst aufrufen, sehen nichts. Ein Spezifizierer, der ChatGPT die Recherche gegen einen Text-only-Fetcher machen lässt, wird per Default auf Qiagen, Thermo Fisher oder Bio-Rad geroutet." }, "evidence": { "en": "A reference-grade B2B catalogue that text and code agents can't reach at all.", "de": "Ein Referenz-B2B-Katalog, den Text- und Code-Agenten gar nicht erreichen." }, "question": { "en": "The question this raises: where would broader measurement — additional tasks, additional providers — sharpen the discovery read?", "de": "Die Frage daraus: Wo würde breitere Messung — zusätzliche Tasks, zusätzliche Provider — die Discovery-Lesart schärfen?" }, "confidenceHedge": { "en": "Based on the current pilot slate, this read is directional.", "de": "Basierend auf dem aktuellen Pilot-Slate ist diese Lesart direktional." }, "closes": { "en": "The fix is not removing WAF, the security trade-off is real. It is publishing the catalogue's agent-readable shadow alongside it: Reference Pages per high-traffic SKU carrying spec + price + cart URL in plain HTML, Answer Pages for the parametric questions a buyer asks (\"Welches Kit für RNA aus Vollblut?\"), Concept Pages for the technique vocabulary so the brand owns the terms upstream of product selection.", "de": "Der Hebel ist nicht das Entfernen der WAF, der Sicherheits-Trade-off ist real. Es ist das Veröffentlichen des agentenlesbaren Schattens des Katalogs daneben: Reference Pages pro High-Traffic-SKU mit Spec + Preis + Cart-URL in einfachem HTML, Answer Pages für die parametrischen Fragen (\"Welches Kit für RNA aus Vollblut?\"), Concept Pages für das Technik-Vokabular, damit die Marke die Begriffe vor der Produktauswahl besetzt." }, "proof": { "en": "The proof is not a new commerce flow. It is plain readers and coding agents extracting \"T3809, TRI Reagent BD, whole blood compatible, €154/25mL, add-to-cart URL\" directly, alongside the existing browser path, re-measured each wave.", "de": "Der Beweis ist nicht ein neuer Commerce-Flow. Es ist, dass einfache Reader und Coding-Agenten \"T3809, TRI Reagent BD, Vollblut-kompatibel, €154/25mL, Add-to-Cart-URL\" direkt extrahieren, neben dem bestehenden Browser-Pfad, neu gemessen mit jeder Wave." }, "ctas": [ { "weight": "primary", "eyebrow": "Audit · €1,900", "title": { "en": "Commission an audit.", "de": "Audit beauftragen." }, "body": { "en": "Where the BrandScore opens the question, an audit closes it. An interpretive engagement on your full surface, scored under the same methodology.", "de": "Wo die BrandScore die Frage öffnet, schließt sie ein Audit. Ein interpretatives Engagement auf der vollen Surface, gemessen unter derselben Methodologie." }, "cta": { "en": "Get the audit", "de": "Audit anfragen" }, "href": "mailto:hello@hyperize.ai?subject=Audit%20%C2%B7%20Merck" }, { "weight": "secondary", "eyebrow": "Founding · €4,500", "title": { "en": "Found with us.", "de": "Founding Program." }, "body": { "en": "Strategic partnership for brands building agent success as a long-term capability, not a one-off engagement.", "de": "Strategische Partnerschaft für Marken, die Agent-Success als langfristige Capability aufbauen — nicht als Einmal-Engagement." }, "cta": { "en": "Apply", "de": "Bewerben" }, "href": "mailto:hello@hyperize.ai?subject=Founding%20Program%20%C2%B7%20Merck" }, { "weight": "secondary", "eyebrow": "Snapshot", "title": { "en": "Audit an adjacent property.", "de": "Eine angrenzende Property messen." }, "body": { "en": "The BrandScore covers the primary domain. Get the same methodology applied to an adjacent property: a country site, a sub-brand, a category beyond the DAX-40 slate.", "de": "Die BrandScore deckt die primäre Domain ab. Die gleiche Methodologie für eine angrenzende Property: eine Länderseite, eine Sub-Brand, eine Kategorie außerhalb des DAX-40-Slates." }, "cta": { "en": "Get a Snapshot", "de": "Snapshot anfragen" }, "href": "mailto:hello@hyperize.ai?subject=Snapshot%20%C2%B7%20adjacent%20property%20for%20Merck" } ] }, "sources": [ { "id": "S1", "label": "Gate-1 audit run · Merck Wave Q2 2026 (Cody, unbranded informational)", "accessedAt": "2026-05-24", "supports": [ "AI Visibility score (audit-derived, 18 datapoints, 3 providers)", "AI platforms queried (openai/perplexity/anthropic)", "close state reached (cart_ready)" ], "url": null, "kind": "internal", "internalLabel": "Internal · Hyperize audit pipeline" }, { "id": "S2", "label": "a fleet wave Life Sciences wave · Merck phases 1-4 (Giorgio repo)", "accessedAt": "2026-04-07", "supports": [ "per-breed access profile (text/code blocked at connection layer; browser success 3/3; ACT add-to-cart no login)", "WAF finding (HTTP 000 timeout for both text and coding agents)", "cart-ready ACT (Warenkorb 1 + alert 'Artikel wurde in den Warenkorb gelegt')" ], "url": null, "kind": "internal", "internalLabel": "Internal · Hyperize fleet (Giorgio)" }, { "id": "S3", "label": "Ground-truth ceiling run · TRI Reagent BD T3809 specification", "accessedAt": "2026-04-07", "supports": [ "Ground-truth catalog number (T3809)", "Whole-blood compatibility, 25mL/100mL/200mL pack sizes, €154/€472/€863 pricing" ], "url": null, "kind": "internal", "internalLabel": "Internal · Hyperize fleet (Giorgio)" }, { "id": "S4", "label": "Task Selection Doctrine", "accessedAt": "2026-05-24", "supports": [ "fairness declaration", "Third-Party Interception framing" ], "url": "https://www.hyperize.ai/en/methodology/task-selection", "kind": "public", "internalLabel": null } ], "crossLinks": { "index": "https://www.hyperize.ai/en/dax40-index", "taskSelection": "https://www.hyperize.ai/en/methodology/task-selection", "agentSurface": "https://www.hyperize.ai/en/methodology", "foundingProgram": "https://www.hyperize.ai/en/founding-program" }, "scope": { "publishes": [ "Brand identity (name, legal entity, DAX ticker, sector)", "Composite (index-ranking) + AI Visibility + AI Usability (as access profile + derived score) + Evidence, with weights", "Per-agent-class access profile (which agent classes reach the close state)", "Third-Party Interception classification, direct/intermediary share, narrative", "Hyperize-selected task titles, close state, intermediary mode, task score, bottleneck sentence", "Measurement timeline (date, wave, protocol version, summary)", "Sources with public URLs or internal-evidence pointers", "Confidence grade + Evidence tier + Next review date", "Bottleneck classification (Discovery / Interception / Usability / Multi-axis / No dominant gap)", "Bottleneck archetype + consequence-claim diagnosis + operator question", "Test narrative (verdict, narrative, agent matrix, scope shift) — scene-level prose, no prompts or formulas" ], "doesNotPublish": [ "Task identifiers (internal)", "Raw and derived measurement artifact paths (internal repos)", "Per-phase outcome tables (ceiling / http / coding / browser_standard / act)", "Individual prompts and provider-specific transcripts", "Scoring derivation steps below the dimension level", "Sector task grid file references when still draft" ] } }