Deep Dive

The Enrichment Engine

50-129 enriched fields per product. 10-20 web sources. Confidence-scored. Anti-hallucination checked. Follow a real product through every phase.

50-129

Enriched fields per product

10-20

Web sources scraped

67.6%

Multi-source validated

~10 min

Per batch, fully automatic

The 7 Phases

Follow the Schuberth C5 through every phase

Watch a real product transform from a sparse CSV row into the most complete product profile on the internet.

1

Phase 1: Import

CSV → Structured Records

What the seller uploads

title: "Schuberth C5"

ean: "4017765145231"

price: "549.00"

brand: "Schuberth"

description: "Premium modular helmet"

What Central creates

Category auto-detected: Motorcycle Helmets → Modular

5 fields imported at confidence 1.0

Product record created with UUID

2

Phase 2: Field Suggestion

Category Intelligence → Schema

AI analyzes the “Modular Motorcycle Helmets” category and suggests 87 fields that should exist:

Weight Shell sizes Shell material Noise level (dB) Certification Visor type Pinlock ready Communication system Ventilation channels Chin bar mechanism Liner material Reflective elements Glasses channel Neck roll Breath guard Wind deflector UV protection Field of vision Retention system Helmet bag included ...67 more
3

Phase 3: Web Scraping

Product Identity → 10-20 Web Sources

SERP search discovers 14 relevant sources for the Schuberth C5:

schuberth.com Manufacturer
revzilla.com Retailer
fc-moto.de Retailer
louis.de Retailer
webbikeworld.com Review site
motorcyclenews.com Review site
rideapart.com Editorial
fortnine.ca Review + Retail

+ 6 additional sources scraped and stored

4

Phase 4: Field Discovery

Web Pages → New Fields Found

Scraped pages reveal 12 additional fields the schema didn’t anticipate:

SC2 ready Emergency release cheek pads Anti-fog insert included Micro-ratchet buckle Aerodynamic whisper count Removable sun visor Anti-scratch coating Drop-down sun visor range Chin curtain removable ECE 22.06 test date Helmet speaker pockets Wind tunnel tested speed

Total schema now: 99 fields

5

Phase 5: Extraction

Web Pages + Schema → Structured Values

AI extracts values for every field from every source. Example for “Weight”:

Source Extracted Value Context
schuberth.com 1,640g (±50g) Official spec sheet
revzilla.com 3.6 lbs (1,633g) Product listing
webbikeworld.com 1,648g (weighed) Hands-on review, own scale
fc-moto.de 1,640g Technical data table
6

Phase 6: Consolidation (Truth Engine)

Multiple Values → One Canonical Truth

Four sources reported weight. The Truth Engine consolidates:

Canonical value: 1,640g Confidence: 0.97

4 independent sources agree within ±0.5% tolerance. Weighted by source authority. Manufacturer data prioritized.

7

Phase 7: Optimization

Intelligence → Channel-Ready Content

The enriched profile generates channel-specific outputs:

Google Shopping title

Schuberth C5 Modular Helmet | ECE 22.06 | 1,640g | Matte Black

Amazon keywords

schuberth c5 modular helmet ece motorcycle touring flip...

Smart Negative

Not for track racing — no SNELL/FIM certification

Contextual Spec

1,640g — lighter than 72% of modular helmets

Living FAQ

87 product-specific Q&As generated

Schema.org JSON-LD

Product, AggregateRating, FAQPage markup

Multi-Source Intelligence

Spies in 20 fortresses

Imagine you need to know the true weight of a helmet. You could ask the manufacturer. Or you could send spies to 20 independent fortresses — each one reporting back what they found. When 4 spies from 4 different fortresses independently report “1,640g,” you know it’s true.

That’s exactly what Central does. Each web source is an independent observer. They don’t coordinate. They can’t collude. When they agree, it’s because the truth is the truth.

Manufacturer Primary

Reports: 1,640g ±50g

Retailer A Secondary

Reports: 3.6 lbs (1,633g)

Review Site Independent

Reports: 1,648g (weighed)

Retailer B Secondary

Reports: 1,640g

Forum Post Low

Reports: about 1.6kg

Consensus: 1,640g · Confidence: 0.97

5 independent sources, weighted by authority

Conflict Resolution

When sources disagree, the truth engine decides

Real data is messy. Sources report different values. The Truth Engine uses weighted consensus, authority scoring, and recency analysis to resolve conflicts.

Example: Certification Conflict

Retailer A ECE 22.05 Outdated
Manufacturer ECE 22.06 Current
Retailer B ECE 22.06 Confirms
Review Site ECE 22.06 Confirms

Resolution: Manufacturer data (authority: highest) + 2 confirming sources override 1 outdated retailer listing. Canonical value: ECE 22.06 at confidence 0.97.

Enrichment by the numbers

50-129

Fields per product

10-20

Sources scraped

67.6%

Multi-source validated

0.82+

Display threshold

6

Violation types blocked

~10 min

Per batch

We understand your products better than you do

Let us enrich your catalog. In 30 minutes, you’ll see your products the way the internet sees them — and what’s missing.