Master data is the list of things a business acts on: its customers, products, suppliers, sites, accounts and the codes behind them.
Start with the smallest possible statement of the problem and the answer.
A large enterprise holds the same company in several systems. It is a vendor in SAP with one number, a supplier in Oracle with another, a party in a procurement portal with a third, and a row in the lake with a fourth. Each copy has its own spelling, its own address and its own payment terms. Every process that touches that company touches one of the four copies, so spend, risk and performance are measured four times over.
Master data management answers that with a golden record: one record for the real company, assembled field by field from the four sources, carrying the four source keys underneath so the assembly can be explained and reversed. Every process then reads the same company, and a correction made once is visible everywhere.
The same pattern applies to a product sold under four part numbers, a site that appears as a plant and a warehouse and an address, and a general ledger account that three ledgers spell differently. The golden records page shows the assembly step by step on one supplier.
These three change slowly and are shared by every process, which is what makes them worth governing centrally.
| Kind | What it is | Examples | How often it changes | Who consumes it |
|---|---|---|---|---|
| Reference data | The shared code lists every system validates against | Country codes, units of measure, currencies, payment terms, tax codes | Rarely, on a published calendar | Every system and every report |
| Master data | The things the business acts on, one record each | Customers, products, suppliers, employees, plants, accounts | Slowly, through a review step | Order entry, planning, finance, service |
| Hierarchies | How those things roll up for reporting and planning | Sales territories, product families, cost-centre trees, legal parents | At reorganisations, with dates | Planning, finance, executive reporting |
| Transactions | What happened and when, read beside the three above | Sales orders, invoices, shipments, service tickets | Constantly | Operations, and any analysis of the golden records |
The fourth row matters as much as the first three. A golden supplier is interesting to a buyer at the moment it is shown with its open orders, its invoices and its quality findings. That is why Lima keeps transactions in the same data bed as the master records, with the keys enforced between them: see the data bed and drill-downs.
The 2026 Gartner Magic Quadrant for Master Data Management Solutions lists eight mandatory features and scores four implementation styles. The list below uses that vocabulary, so a reader who knows the market can find each duty quickly.
| Duty | What it means in practice | Where it sits in Lima |
|---|---|---|
| Create or define a golden record | Assemble one record per entity from many sources, with survivorship and a reversible merge | Golden records · headless API |
| Data quality and cleansing | Standardise values, validate them against rules, detect anomalies, measure completeness | Data quality (its own section) · Rules at three levels · AI-assisted |
| Stewardship and governance support | Queues, proposals with evidence, approvals, ownership per domain | Stewardship · audit trail |
| Integration, loading and synchronisation | Read the sources, capture changes, write agreed values back through released interfaces | Integration and sync · federated display |
| Multidomain and cross-domain support | Several kinds of master data in one model, with the links between them held | The domains · the catalogue |
| Operational and analytical use | Serve a transaction at low latency and answer an analytical question on the same records | Time series and pivots · data services |
| An off-the-shelf product | Configuration reaches production, with the mechanics supplied (no custom build) | Your first domain in a day · documentation |
| Four implementation styles | Registry, consolidation, coexistence and centralised, chosen per domain | Implementation styles |
Source: Gartner Peer Insights market definition for Master Data Management Solutions, updated May 2026; Magic Quadrant for MDM Solutions, April 6, 2026. The vocabulary is the market's; the mapping is Lima's. See how analysts categorise MDM for the full analyst picture.
Master-data products have historically served one of these two and handed the other to a warehouse. Both matter, and the difference shows in the architecture.
The operational use is a lookup inside a process. An order-entry screen asks 'which customer is this, and may we sell to them today'; the answer has to arrive in milliseconds and be the current one. Lima answers that through the headless lookup, which returns the golden record, its source keys and the live related facts the caller asks for.
The analytical use is a question about the population. 'Show revenue by golden customer and month for the reorganised territories, with last year beside it.' That needs the golden records in their hierarchies, the transactions beside them, and measures over both. Lima answers it as a pivot on the same records, and a cell drills to the source rows behind it.
One model serves both because the golden records, the hierarchies, the reference data and the transactions sit in one bed with the keys enforced. The security layers and the audit trail apply at both doors, so an analytical question obeys the same masking as the operational one.
Both, and the distinction matters when budgeting. The product supplies the mechanics: matching, survivorship, hierarchies, rules, security, audit, services. The programme supplies the agreements: which source wins which field, who stewards each domain, what a duplicate is. Lima shortens the product side to days so the time goes into the agreements.
A warehouse holds copies for analysis and answers questions about the past. A master-data platform holds the agreed identity of each thing and serves it back into the processes that create new transactions. Lima does the second, and reads the warehouse or lake as one more source when history at scale is useful.
Data quality measures and fixes values inside a record. Master data decides which records are the same thing and which value survives across them. The two work together: Lima runs the quality rules and the identity decisions in one place, so a fix and a merge appear in the same audit trail.
One model, domain by domain. Most programmes start with one domain that hurts, prove it, then add the next. The lifecycle page shows the stages, and the catalogue gives an effort estimate per domain.
Bring two systems that hold the same customers, suppliers or materials. In one working day we connect them, run the match on live rows, review the queue with your steward and publish the first golden view with its API.
