--- title: "Enrich — Master Data Cleanup & Standardization for Supply Chain AI | bluefabric" description: "bluefabric uses purpose-built AI to clean, standardize, deduplicate, and backfill supply chain master data — items, suppliers, customers, locations, units — across WMS, TMS, and ERP, so AI agents reason on operational truth instead of broken records." url: https://bluefabric.ai/enrich/ source: content/enrich/index.html --- [bluefabric](/)/ [Product](/ingest/)/ Enrich Pillars [Ingest](/ingest/) [Enrich](/enrich/) [Unify](/unify/) [Context](/context/) [Calculate](/calculate/) [Use](/use/) Layer 2 of 5 · Enrich # Make broken supply chain data _usable for AI._ Getting the data in is only the first step. Once supply chain data lands in bluefabric, our purpose-built AI analyzes it for the issues that make agents unreliable — **duplicates, inconsistent naming, missing attributes, bad descriptions, conflicting values, incomplete product records, and master data gaps.** bluefabric cleans, standardizes, enriches, and backfills the data before agents use it. Because AI agents should not reason over a messy SKU master. [See bluefabric Live →](/demo/) 15-min walkthrough // raw records shaped into clean form Step 1 · Clean ## First, bluefabric cleans the _obvious mess._ Most data problems are not dramatic. **They are small, boring, and everywhere.** bluefabric scans your SKU masters, product files, order data, supplier records, and operational datasets to detect the inconsistencies that make downstream AI unreliable. If the record is messy, the answer will be messy. SKU master"6-Pack Lager" · "6 pack lager" · "6 PK LAGER"duplicate ProductDescription: _(empty)_missing Price feed$5.00 (ERP) ≠ $10.00 (WMS)conflict CategoriesBeer · Beverages · Alcohol · BEERinconsistent OrderItem #4821 · qty: _null_incomplete ![](https://pub-6d0b5b97762c4335b5b515672d21523f.r2.dev/img/technology/network-graph-nodes-cyan-orange-dark.jpg) Step 2 · Resolve ## Then it finds the inconsistencies _hiding in plain sight._ The same product should not tell two different stories. A six-pack should not be $5 in one file and $10 in another. A case should not have missing dimensions in one system and a weight value in another. bluefabric compares related records across systems, identifies conflicts, and flags what needs resolving. Duplicates Same product, two records. Exact and near-duplicate items, name variants, capitalization drift, abbreviation differences, and partial matches across feeds. "6-Pack" vs "6 PK" "Pepsi 330ml" vs "PEPSI 330 ML" SKU-4821 vs 4821-A Conflicts Two systems, two answers. Conflicting prices, dimensions, weights, units, descriptions, and categories — flagged with source so a human can decide which is truth. price: $5 vs $10 units: case vs each weight: kg vs lb Gaps Missing fields agents shouldn't guess. Suspicious blanks and obviously-incomplete records, surfaced before they break a downstream calculation or a customer-facing answer. missing HS code missing pack hierarchy missing barcode Example · de-duped SKU Raw source SKU:SKU-4821 SKU:sku 4821 SKU:4821-A ← three records, one item → After Enrich SKU:SKU-4821 (canonical) aliases:\[sku 4821, 4821-A\] vendor:ACME-23 → one resolved entity Clean context in. Better agents out. ![Cereal boxes lined up on a warehouse shelf in dramatic low light — SKU records with attributes to be inspected, backfilled, and augmented](https://pub-6d0b5b97762c4335b5b515672d21523f.r2.dev/img/cpg/cereal-boxes-warehouse-golden-light.jpg) // half a product record is not enough Step 3 · Backfill & Augment ## Then it fills the blanks — _the way each product needs it._ Most supply chain data is incomplete. **Missing dimensions, weights, pack hierarchy, HS codes, barcodes, handling rules.** bluefabric backfills using your connected systems, similar products, historical records, external references, and supply chain-specific logic. But a beer case, a lithium battery, a frozen meal, and a fragile spare part do not need the same attributes. bluefabric enriches based on **what the product actually is** and what decisions agents need to make about it. Seasonality Demand patterns over time. Peak periods, weekly cycles, category drift, historical movement signals. Trade & compliance What it takes to move it. HS codes, tariff categories, customs attributes, packaging regs, hazmat flags. Physical handling What it takes to store and ship it. Dimensions & weights, pack hierarchy, palletization, temperature, GTIN. 100s of enrichments out of the box Agents need complete context, not half a product record. 📦 Products C SKU master catalogue 3/5 Records synced 34k Completeness 47% ●  Field completeness 10 fields ✓SKU100% ✓Name100% ✓Weight100% ✓Dimensions L · W · H100% !Min stock level21% ○HS codemissing // every value tagged with source · freshness · confidence Step 4 · Validate ## Every enrichment keeps its _source and confidence._ **bluefabric does not silently overwrite your systems.** Every cleaned value, inferred attribute, resolved conflict, and backfilled field keeps context: **where it came from, how confident the match is, and whether it should be reviewed** before being pushed back to a source system. Teams use enriched data immediately while keeping control over what becomes system truth — with a live data-quality grade per entity, so you always know what is ready and what still needs a human eye. Better data without losing control. From messy records to usable context ## The full enrichment _journey._ bluefabric does not just clean data for reporting. It prepares operational data for AI agents that need to reason, calculate, and act. 01 Ingest bring sources in → 02 · here Enrich clean · resolve · backfill → 03 Unify common data model → 04 Calculate trusted methods → 05 Use MCP · any agent Clean records. Richer context. Better agents. [Explore the Data Model →](/unify/) [Architecture deep-dive](/architecture/) ![](https://pub-6d0b5b97762c4335b5b515672d21523f.r2.dev/img/cpg/warehouse-supermarket-aisle-banners.jpg) AI agents need clean context ## Better data, _without losing control._ bluefabric prepares the messy operational data your business already has into the clean, attributed, traceable context AI agents need to reason, calculate, and act. Because AI agents should not reason over a messy SKU master. [See bluefabric Live →](/demo/) [Next: Data Model →](/unify/) [ ← Back: Layer 1 Ingest How bluefabric connects every operational source — WMS, TMS, ERP, files, APIs — without rip-and-replace. ](/ingest/)[ Next: Layer 3 → Unify · Data Model How clean, enriched records become a common supply chain data model any AI agent can query. ](/unify/)