Beyond CRM.

Stars tell you how good.
Receipts tell you what sells.
Taste tells you who.

Taste, Decoded

A traditional CRM tells you what your members did.
Kornerway tells you what they want — and where that want is heading.

Overview deck (PDF) ↗

Three-Vector Taste Engine · Patent Pending · HG001
Scroll

An introduction to Kornerway.

Aggregate ratings
are noise.

Marketplaces lose conversion to noisy 5-stars. Loyalty programs blast generic offers and burn member trust. Both feel the same pain in different shapes — and both pretend it's a personalization problem when it's really a measurement problem.

Average preferences don't exist. Taste does.

01 / THE AGGREGATION TRAP

Averaged opinions erase taste signals

When you average 3,000 reviews, individual taste profiles cancel out. The result reflects mass consensus, not personal fit.

02 / THE IDENTITY PROBLEM

No platform knows who you eat like

Recommendation engines use behaviour (clicks, bookings) not taste attributes. They optimise for the next transaction, not the right match.

03 / THE COLD START TRAP

New users receive the worst advice

Without historical behaviour, collaborative filters default to popularity. The new user gets the most generic possible list.

Trust Vector · For Marketplaces

Stop ranking restaurants. Start matching tastes.

Convert browsers into bookings by showing them recommendations from people who actually share their taste. Cross-domain prediction means we know what hotel they'll love after just three restaurant reviews.

Built for Booking Platforms · Review Apps · Travel Marketplaces

Resonance Vector · For Loyalty Programs

Your members aren't a recency score. They're a taste.

Stop sending the same offer to every member who engaged last month. Kornerway clusters your members by what they actually enjoy and routes the right offer to the right cohort — including the sparse-data members RFM ignores.

Built for Airlines · Hotels · Social Clubs

Pathway Vector · For Organisations

Your people aren't a seniority band. They're a pattern.

Route talent to the right next role, specialty, or team based on what they actually engage with — not just tenure. Kornerway reads voluntary participation signals and preference patterns to augment the decisions that determine who goes where.

Built for Professional Services · Disciplined Services · Enterprise

One engine. Three vectors.
Same architecture.

Every Kornerway deployment runs the same core: attribute-decomposed trust propagation over a 10-dimension Universal Taste Vector. Whether the input is public reviews or private loyalty receipts, the math is identical. Trust Vector, Resonance Vector, and Pathway Vector share the same patent, the same engine, and the same calibration gate. The only thing that changes is what the buyer routes through it.

01

Behavioural Ingestion

Partner platforms pipe in the signal — reviews, receipts, reservations, or promo engagement. Each event is parsed into structured attributes via our two-pass LLM extraction pipeline. Runs once at ingest.

02

Taste Vectorisation

Extracted attributes are embedded as 10-dimension Universal Taste Vectors and stored in PostgreSQL with pgvector. Reviewer and member profiles accumulate over time.

03

Trust & Cohort Layer

Trust Vector forms reviewer-to-reviewer edges by attribute-level agreement. Resonance Vector clusters members into taste-coherent cohorts. Both update incrementally, not on every query.

04

Real-Time Routing

At query time: pure pgvector cosine similarity + context reweighting. No LLM calls. Sub-200ms response. Matches arrive with full provenance — trust path or cohort lineage.

Validated in POC

0.92–0.95
Cross-Domain Taste Prediction
Restaurant → hotel preference recovery

800 reviewers · 4,408 reviews validated

+234%
Precision Over RFM Baseline
Cohort-static loyalty intelligence

401 households · 8 cohorts · 366 offers scored

A traditional CRM tells you what your members did.
Kornerway tells you what they want.

Demos are partner-confidential. Request access if you don't already have credentials.