Tamr is an 'AI-native' mastering product: a model trained on expert-labelled pairs clusters records, match and split rules refine the clusters, consolidation rules build golden records, and pricing is per golden record (Tamr ID).
Every Tamr cell carries its source number; the sources are listed at the foot of the page.
| Question | Tamr | Lima master data |
|---|---|---|
| Where the master data lives | Source datasets are loaded into Tamr projects (unified dataset, clusters, golden records) and the outputs are exported to the warehouse [1]. | Golden records are governed views over the sources, read in place through the federated engine, and materialised only where latency needs it, in your lake, under the same keys. The transactions sit in the same bed, so a merged customer and its orders change together. |
| Reaching SAP, Oracle and the lakes | Cloud SaaS through the GCP, AWS and Azure marketplaces; Gartner notes a connector strategy 'shifting to API-first' [2][3]. | SAP S/4HANA and ECC through released read interfaces and change capture; Oracle Fusion through REST and BI Cloud Connector extracts, E-Business Suite through JDBC and open interfaces; Snowflake, Databricks and Fabric read in place; Salesforce, ServiceNow and Workday through their APIs. Write-back through BAPIs, IDocs, REST and web services, acknowledged and logged. |
| Golden record: matching and survivorship | Model-based clustering with priority-ordered match and split rules and curation edits; consolidation rules per attribute with dataset priority, conditions and overrides; the applied rules are recorded per cluster [4][1]. | Strong keys first (VAT ID, DUNS, GTIN, manufacturer part number), then standardised attribute and address matching, with a language model for the hard pairs; two thresholds per rule (review, merge); survivorship per field by source rank, recency, verification and steward override; unmerge restores the sources; every value keeps its source, rule and time. |
| Hierarchies and reference data | Hierarchy management is absent from the mastering documentation; an Enterprise Knowledge Graph is marketed [1][5]. | Levels and several hierarchies per entity, kept as versions with validity dates and validated against the facts they aggregate; reference data and crosswalks kept as small domains with the same history, rules and security. |
| Rules, security and audit | Curation edits and applied rules are recorded in clustering metadata; a Curator Hub with queue roles; five isolated regions [4][6]. | Column, row and business rules in one language, evaluated on save and on demand, including rules over aggregates and across tables; twelve security layers for read, write and delete; an append-only audit trail with point-in-time reads and logged reads of sensitive columns. |
| APIs and data services | Tamr Cloud API and RealTime lookup referenced on the site; details were unreachable at the time of checking; MCP connectivity marketed [5]. | REST, SOAP, SQL (JDBC and ODBC), events and an agent endpoint over one policy engine; the golden-record lookup answers with the record, the cross-reference and live related facts; page sizes and rate limits published per endpoint. |
| Analytics: pivots, measures, transactions | Model accuracy metrics and cluster curation inside; analytics after export; pivots and measures are absent [1]. | Pivots on any hierarchy level with time on the columns and measures in the cells; calculated measures; data entry with disaggregation by weights, equal shares or a time profile; drill-down from a cell to the transactions and to the source row; charts beside grids; natural-language and SQL querying. |
| Pricing unit and trial | Per golden record (Tamr ID) with unlimited users, onboarding and support included, prices unpublished; a guided proof of concept and a value calculator in place of a trial; Forrester TEI: 643% ROI (2021) [7][8]. | Platform capacity tiers (32C to 128C); domains, records, attributes and users included; the price is written before the work starts; a guided first domain on your data in one day; the showcase estate for demonstrations; a self-service sandbox tier is a decision pending (see the review notes). |
| Analyst placements 2025–26 | Gartner Magic Quadrant for MDM Solutions, April 6, 2026: Niche Player, with 'AI-native mastering at massive volumes' as the noted strength [3]. | Analyst evaluation is the purpose of this page set; placements to date: none. The platform pages are organised on Gartner's eight mandatory features and four implementation styles so a briefing lands in the right market. |
Choose Tamr when customer or supplier mastering at very large volume with model-based matching and external B2B reference data is the whole job, and when paying per golden record with unlimited users suits the budget. Its trained models and its outcome-based unit are real advantages for a large B2B data-science-led programme [3].
Lima suits the enterprise whose master data should be governed on the rows in place across SAP, Oracle, the CRM and the lakes, with the transactions in the same bed: the golden record and its orders in one query, rules over totals, pivots on golden dimensions with data entry and disaggregation, drill-downs to the source row, and a headless lookup that answers with live related facts. Twelve security layers and an append-only audit trail apply to every door, and the price is written before the work starts.
The domain catalogue shows thirty domains with, for each, the objects read in place in SAP S/4HANA, Oracle Fusion, Salesforce, Workday, ServiceNow, Snowflake and Databricks and the interface used for the write-back.
Facts change; re-check the sources at each review and re-date the page.
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.
