Case Study

Product data with provenance, not scraped guesses

Vareoprettelse sits between messy supplier inputs — Excel exports, producer feeds, bespoke sheets — and a company's PIM. It does not deliver scraped product data; it delivers a resolved, validated product proposal where every value has provenance, a confidence score, and an explanation of why it was chosen.

Many candidate values are collected per attribute as evidence. Exactly one becomes the selected truth, chosen by configurable rules — or escalated to a human. Automation handles the 90%; experts review the 10% that needs judgment.

Evidence-first intakeGrounded, governed AIHuman review by exceptionMulti-tenant with database-enforced isolation

Scroll to see how it works

The Problem

Manual product onboarding is the bottleneck on catalog growth

For distributors and retailers, onboarding a product means hunting down weights, dimensions, datasheets, images, customs codes and writing copy — per product, per supplier, in inconsistent formats.

The result is slow, expensive, and unauditable. In a flat PIM, weight = 12.5 has no story: when it is wrong, nobody knows why or where it came from.

The same supplier's files get re-keyed every time, and regulated classification like CN commodity codes is error-prone by hand.

No provenance behind published values
Repeated re-keying of the same supplier layouts
Error-prone, regulated customs classification
Every product eyeballed, regardless of confidence

The Core Idea

Collected data ≠ accepted product data

Many candidate values are collected per attribute — that is evidence. Exactly one becomes the selected truth, chosen by configurable rules, with a confidence score and a reason you can read. Nothing is silently authoritative.

Collected evidence

12.5 kg

Producer datasheet · Tier 1 · Producer

0.94

12.5 kg

Distributor PDF · Tier 2 · Distributor

0.78

12 kg

Retailer listing · Tier 3 · Retailer

0.41

Resolve

Accepted value

12.5 kg

confidence 0.94

Producer value, confirmed by a second source. Retailer value disagreed and was down-weighted.

Every value carries provenance, a confidence score, and an explanation — so your PIM stops receiving mystery data, and any published number is defensible.

The Flow

A durable, crash-safe pipeline from messy file to publish-ready product

Each product runs through a job graph whose entire state lives in database rows. Work fans out across many sources in parallel, a phase only advances when the previous one drains, and a crash is a non-event.

1
Upload
Supplier file → mapped candidates
Intake
2
Discover
Web search + match verification
Enrich (parallel)
3
Fetch
Fan-out across sources
Enrich (parallel)
4
Score
Multi-source agreement
Resolve
5
Resolve
One value per attribute
Resolve
6
Validate
Checksums + completeness
Resolve
7
Review
By exception only
Approve & publish
8
Publish
Approved proposal → PIM
Approve & publish

Inside “Fetch” — a variable, append-only fan-out

Discover
root job

→

Producer site

Datasheet PDF

Official DB

Product images

Marketplace

Children are spawned atomically as the parent completes and are only ever appended — cycles are structurally impossible, so one dead source never stalls a product.

Resolve Phase

Score, resolve, and validate before anything reaches a human

When the parallel enrichment phase drains, a short linear tail turns competing evidence into one accepted value per attribute — and decides whether the product can flow through untouched.

Step 01

Score agreement

Multi-source agreement is recomputed so values confirmed across trusted sources gain confidence and outliers lose it.

Step 02

Resolve one value

A configurable strategy — best confidence, priority source, or authoritative only — picks a single winner per attribute, with a reason.

Step 03

Convert & canonicalize

Units are converted and values normalized to internal codes so the proposal is consistent regardless of source formatting.

Step 04

Validate readiness

Checksums, type checks and required-field coverage decide: ready for auto-import, or routed to manual review.

Key Workflows

One pipeline, several operator-facing flows

The same evidence spine powers everything from remembered column mappings to grounded copy generation and PIM publishing.

Scenario context

An operator uploads a supplier file. The system fingerprints the header, looks up a remembered mapping for that sender, and only asks the operator to confirm low-confidence columns before creating one candidate per row.

Map each supplier once, never again
Raw rows preserved alongside mapped attributes
Low-confidence columns surfaced for confirmation

Governance note

Mappings are versioned per tenant and source key, so a layout change is a new version — not a silent overwrite.

Corrections on AI-sourced values are captured as a learning signal — aggregated into governed, human-approved improvements, never silent rule rewrites.

Proof Layer

Architecture built for scale, isolation, and auditability

The hard, expensive-to-retrofit parts — domain model, orchestration, tenancy and testing — are built to production standards.

Context

Operational domain: High-volume product onboarding from heterogeneous supplier catalogs
Primary users: Product data, e-commerce and PIM operations teams

Scope

Evidence spine: 52-model schema with per-language, per-channel values and JSONB evidence
Orchestration: Durable hub-and-spoke job DAG with transactional outbox and phase-drain barriers

Constraints

Isolation requirement: Database-enforced Row-Level Security via a two-role Postgres model
AI requirement: Grounded and governed — no value silently authoritative

Artifacts Delivered

Resolution engine: Three strategies, multi-component confidence, typed value system
Classifiers: Retrieve-then-rerank category and CN commodity code with kNN memory

Outcome Signals

Throughput signal: Most products flow through untouched; humans handle only exceptions
Quality signal: Multi-source agreement, checksum validation, and grounding guards catch errors early

Why It Matters

Fast and auditable and trusted — not a trade-off

Reframing product acquisition as evidence, resolution and provenance is exactly what makes AI safe to use here rather than a liability.

Faster onboarding — automate the 90%, review the 10% that needs judgment

Provable data quality with per-value provenance and confidence

Grounded AI that shows its work and never overrides a human silently

Database-enforced tenant isolation, not app-layer filtering you have to trust

Mapping-with-memory removes repeated re-keying per supplier

Shared, regulated commodity-code learning compounds across tenants

Related Infrastructure Case

LLM GatewayGoverned AI

Sovereign AI

The grounded, cost-governed AI in this platform routes through the same kind of central LLM gateway — read how one control layer governs every model call.

Explore the Sovereign AI case

Related Operational Case

In-App AIWorkflow Automation

Operational AI

See the same embedded, human-in-the-loop approach applied inside day-to-day operational workflows rather than product intake.

Explore the Operational AI case

explore further

Related capabilities

Operational AI
AI systems that integrate with existing platforms and workflows, with control, traceability, and operational reliability.
Systems Architecture
Designing architectural foundations that allow complex organizations to operate reliably and evolve safely.

Product data with provenance, not scraped guesses

Manual product onboarding is the bottleneck on catalog growth

Collected data ≠ accepted product data

A durable, crash-safe pipeline from messy file to publish-ready product

Score, resolve, and validate before anything reaches a human

Score agreement

Resolve one value

Convert & canonicalize

Validate readiness

One pipeline, several operator-facing flows

Architecture built for scale, isolation, and auditability

Context

Scope

Constraints

Artifacts Delivered

Outcome Signals

Fast and auditable and trusted — not a trade-off

Sovereign AI

Operational AI

Capabilities and reading behind this work