Partner Demo
Restricted Access
Invalid credentials
← Back to Intro
Confidential Partner Demo · Kornerway.ai

Reviews that match
your taste.

Seeking the Way

Existing review platforms show you what everyone thinks. Kornerway shows you what people like you think — by building a trust network of reviewers whose taste aligns with yours, across every attribute that matters.

↓ Scroll to explore
The Problem We Solve

A five-star review from someone with different taste tells you nothing.

⭐
Today's platforms
Average all reviews equally. A noise-loving reviewer and a quiet-seeking reviewer both score the same restaurant five stars — for completely opposite reasons. You have no way to know which review applies to you.
🎯
Kornerway's approach
Match you to reviewers who share your specific preferences across food quality, ambiance, service, value, and noise — then propagate that trust through their network, so indirect reviewers become useful too.
🏨
The context problem
A business traveller and a family tourist rate the same hotel entirely differently. Reviews without context are noise. Kornerway's AI pipeline detects occasion context and adjusts scores accordingly.
🔗
The trust chain
You and Reviewer A agree on three restaurants. A and Reviewer B agree on two more. B's recommendations become 72% relevant to you — even though you've never reviewed the same place. This is the Kornerway trust graph.
Two Streams · One Engine

Two vectors. Two engines. One mathematical core.

Trust Vector — For Marketplaces

Cross-domain taste prediction for hospitality marketplaces.

Scroll to Resonance Vector — For Loyalty Programs ↓
01
Core Algorithm
Trust Propagation Network
Build your taste profile by rating restaurants across five attributes. Watch the trust network compute direct and indirect reviewer relevance scores in real time. The D3 graph visualises your personal trust chain — hover any node to see exactly why each reviewer is trusted.
Cosine similarity Multi-hop propagation Attribute-specific trust Live graph
→
02
Statistical Validation · Static Only
Statistical Proof of Concept
500 simulated users, 40 reviewers, 80 restaurants. The trust propagation engine is tested head-to-head against simple average ratings and random selection. MAE, Pearson correlation, bootstrap confidence intervals, and p-values — all statistically significant. Upload your own CSV to re-run against real data.
Mean absolute error Pearson r Bootstrap CI CSV upload
→
03
AI Pipeline · Static or Live
Review Intelligence Pipeline
Paste any restaurant review. The two-pass AI pipeline detects dining context (romantic, family, business, solo), extracts raw attribute scores, then produces context-adjusted scores — showing exactly how a business dinner rating should differ from a family outing rating of the same restaurant.
Context detection Attribute extraction Context adjustment 4 example reviews
→
04
Live API · Static or Live
Live Review Decoder
Submit any review text and watch the two-pass AI pipeline extract structured attribute scores in real time. See raw scores, context-adjusted scores, confidence levels, and signal strength per attribute — live against the Kornerway API.
Live API calls Attribute scoring Context adjustment Confidence levels
→
05
Analytics · Static Only
Graph Density Analyser
Explore the trust network topology. Visualise reviewer graph density, clustering coefficients, and trust edge distributions. Understand how network structure affects recommendation quality and cold-start scenarios.
Network topology Graph density Clustering Cold-start analysis
→
06
Simulation · Static Only
Recommendation Simulator
Simulate full recommendation cycles for synthetic user cohorts. Adjust trust thresholds, network depth, and attribute weights — see how each parameter affects recommendation accuracy and coverage in real time.
Parameter tuning Cohort simulation Trust thresholds Coverage vs accuracy
→
07
Partner Portal · Static or Live
Partner Portal
Partner-facing pilot dashboard with three tabs: real-time ingestion metrics, trust network health, and recommendation performance. White-label ready for partner deployments.
Pilot dashboard 3-tab view Partner-facing White-label
→
How the Engine Works
Five steps from raw review to trusted recommendation.
01
User taste profile
User rates places they've visited across configurable attributes. Ratings become an n-dimensional preference vector stored in the system.
02
Context-adjusted ingestion
AI pipeline extracts structured scores from reviewer text, adjusting for occasion context before storing — eliminating systematic input bias.
03
Per-attribute trust scoring
Cosine similarity computed independently per attribute. Food trust and noise trust are separate — a reviewer trusted for food may not be trusted for ambiance.
04
Trust propagation
Trust scores propagate through reviewer-to-reviewer chains. 90% × 80% = 72% indirect trust. Chains extend until scores fall below the 20% confidence floor.
05
Ranked recommendations
Restaurants rated by trusted reviewers — direct and indirect — are surfaced and ranked by trust-weighted predicted satisfaction for the specific user.
Start the Trust Vector Demo →

Resonance Vector — For Loyalty Programs

Cohort-static loyalty intelligence. Built for Airlines · Hotels · Social Clubs.

Demo runs on a real social club dataset; your data would surface its own cohorts and offerings.

01
Resonance Vector · 8 Taste-Based Cohorts
Cohort Intelligence Report
Cluster loyalty members by what they actually enjoy. k-anonymity ≥20. Top offers per cohort.
→
02
Resonance Vector · +234% Precision
RFM Lift
The taste-based engine vs the loyalty industry's recency-frequency-monetary baseline. Honest layer included.
→
03
Resonance Vector · Real Household, Sparse Data
Case Study
Real household, 6 receipts, 18 months. Sparse data, coherent recommendations. Includes Family 2 dense-data comparison.
→
04
Resonance Vector · Individual Member Drill-Down
Signal
Radar chart with long-run baseline vs recent activity. Trajectory dim slopes. Event history. Cohort peer count.
→
05
Resonance Vector · RFM vs Kornerway Routing
Routing View
Side-by-side promo allocation comparison for one member, with trajectory-nudge slots highlighted.
→
06
Resonance Vector · Catalog Readiness Check
Calibration Gate
The cosine test that determines whether your catalog is ready for category-level routing.
→
07
Resonance Vector · Partner-Configured Rules
Compliance & Suppression
Per-cohort and per-household rule enforcement. Built for regulated loyalty programs.
→
Start the Resonance Vector Demo →

Pathway Vector — For Workforce Routing

Vocation-matched career routing. Built for disciplined services · professional organisations.

Demo runs on 200 synthetic HKPF officer profiles; real data surfaces its own posting recommendations.

01
Pathway Vector · 10-Dim Vocation Profile
Officer Signal
Career vocation radar, CVA trajectory sparkline, posting history with alignment rationale. Attrition risk score per officer.
→
02
Pathway Vector · +60pp Routing Precision
Routing View
Pathway routing vs seniority baseline side-by-side. Precision@3: 86.7% vs 26.7%. Admin-track systemic attrition signal.
→
Start the Pathway Vector Demo →