Master data · For CEOs and CIOs
Master data your business can trust
Nexus helps you clean up item, supplier and customer records: find duplicates, build a single trusted record for each entity and prepare data for your target systems. Source references, legacy codes and decision history are preserved
- ARBA Nexus · Master Data
- One platform for the full workflow · In your own cloud
The business cost of poor master data
One item, five codes — excess purchases and unreliable reports
- Five codesThe same item was created five times, each with a slightly different description.
- Fragmented inventoryStock is recorded under three codes. The other two show zero.
- Unnecessary purchasingProcurement orders an item already in stock under another code.
- Reporting errorsReports count the item twice or miss it when records cannot be matched.
- Errors carried forwardMigration carries all five codes, and the duplicates, into the new ERP.
The business pays through excess inventory, purchasing delays and conflicting reports. As duplicate records accumulate, the problem grows.
Illustrative example: one item is registered under several codes, with its attributes recorded only in text descriptions.
When data quality becomes a priority
Three business priorities that depend on reliable master data
An ERP migration
Records need to be reviewed, mapped and loaded as part of migration. This is an opportunity to fix data quality issues before duplicates reach the new system.
A merger or a consolidation
Each company has its own item and supplier records. Shared purchasing, inventory management and reporting need consistent records linked to both companies' original codes.
AI on company data
Agents and copilots rely on the records available to them. Duplicates and conflicting values can lead to incorrect conclusions, even when the answer sounds convincing.
Each case requires verified master data by an agreed date, with a traceable basis for every decision.
What Nexus delivers
Verified master data, ready to load — with legacy codes preserved
A golden record is the single trusted record for an item, supplier or customer. It retains the contributing sources, applied rules and approval history. Each original code remains linked to its golden record.
Nexus prepares data between extraction from your current systems and loading into the target platform. Records are matched, checked and approved before handover.
One platform covers the workflow from the first data extract to final reconciliation
The full workflow in one platform
From source data to golden records for your systems
Every stage runs in the same platform, with its results and supporting evidence preserved. Your data owners approve the basis for treating different records as the same item
- Assess qualityVersion the extract and measure data quality
- Extract attributesIdentify part numbers, sizes and colours in descriptions
- Match recordsApply rules, statistical methods, then AI
- ApproveData owners approve rules for recurring cases
- Build golden recordsKeep one record linked to its sources and original codes
- Publish and maintainDeliver records with code mappings; check new entries
A domain pack holds the configuration for a business data domain. Supported rules can be updated without changing the platform code.
What sets it apart
What Nexus offers your business and IT teams
One platform for the full workflow
Data intake, quality assessment, matching, approval, golden records and duplicate prevention form a connected workflow.
Start with your first extract
An existing platform and configurable rules let you begin assessing quality and preparing golden records from your first data extract.
Your cloud, your models
Nexus runs in your cloud account and accesses AI models through your chosen provider. You control access.
Your data owners approve the rules
Approved rules handle recurring cases, reducing repetitive review. Ambiguous cases are examined individually by your data specialists.
AI under control
Rules and statistical methods come first. AI is used where they are insufficient, with source evidence attached to its proposals for review.
Traceable decisions
Each golden record retains its sources, the applied rule and who approved the decision. Its change history remains available for review.
You retain control
Your infrastructure. Your decisions
Your systems today
- ERP item, supplier and customer masters
- Databases in your internal systems
- Inventory, purchasing and usage history to support matching
Nexus receives extracts without changing the source systems.
Nexus in your cloud
- Quality assessment and rule configuration
- Matching with supporting evidence
- Golden records linked to their sources
- Code mappings and data packages
Your account, your regions, your model provider.
Where the result goes
- Your new ERP, loaded by your migration team
- Your reporting and your AI
- Record counts reconciled between source and target
We deliver the prepared data; your team loads it.
Your approvalInfrastructure and AI model costs appear in your own cloud billing. You pay the provider directly, without an ARBA markup.
Approaches to MDM implementation
A traditional MDM programme and Nexus: where the work begins
| A traditional MDM programme | Nexus | |
|---|---|---|
| Starts with | Configuring the data model, responsibilities, change requests and approval workflows | Assessing your extract and preparing the first golden records |
| Rule configuration | Specialists configure and deploy changes in the MDM system | Supported rules are configured in a domain pack without a platform release |
| Matching inputs | Values held in structured fields | Structured fields and attributes extracted from descriptions: part number, size, colour, grade |
| Who decides | Data stewards (data quality specialists) review records and change requests | Data owners approve rules for recurring cases; ambiguous cases receive individual review |
| Where it runs | The vendor's cloud or a licensed deployment in your infrastructure | Your own cloud account, your own model provider |
| What it is for | Long-term, centralised master data governance | Data cleanup and handover, with ongoing quality control available by subscription |
An overview of implementation approaches based on vendor documentation; capabilities vary by product and configuration. The appendix covers the strengths of established MDM systems.
Project deliverables
Four deliverables for business, IT and audit
Golden records
One trusted record for each item, supplier or customer, linked to its sources and original codes.
Code mappings (crosswalk)
A mapping from original codes to golden records and target-system identifiers, supporting migration and reconciliation.
Decision history and open issues
The basis for each decision to merge records or keep them separate, plus a register of issues awaiting a data owner's decision.
Quality before and after
The same quality measures applied to the source data and the prepared output, so the change can be assessed.
How we work together
Clear responsibilities, agreed deliverables and a completion date
- From you
Your team's contribution
- A designated data owner and weekly decision reviews
- Extracts with their control totals
- Cloud infrastructure for the Nexus deployment
- From us
Platform and methodology
- Nexus deployment and rules configured for your master data
- Data packages with code mappings and reconciliation results
- Supporting evidence for records and a written estimate of AI model costs
- At project completion
Your choice
- A fixed completion date; delays do not automatically extend the project
- Continue by subscription to check new records for duplicates
- Or receive a full export and end your use of the platform
The first discussion defines the data scope, responsible owner, schedule and expected results.
Where to begin
Let's discuss your master data
Tell us which records you need to prepare — items, suppliers or customers — and by when. We will discuss your data sources, the scope of work, your team's involvement and the deliverables.
You do not need to install software or design a data model for the first meeting. The discussion will help you decide whether the approach fits your needs.
Discuss your requirementsFor IT and data teams
Appendix
Processing stages, the role of AI, the strengths of established MDM systems and the scope of Nexus
- The data processing workflow
- AI support and human responsibility
- Strengths of established MDM systems
- Deployment, access and data languages
- What Nexus does not do
Data processing · stages 1–4
From source data to candidate matches with supporting evidence
- 1Receive and version the dataThe original extract is preserved, versioned and reconciled against control totals. Pattern-based checks flag records that may contain personal data and hold them for review.
- 2Assess data qualityMeasure completeness, repeated descriptions and codes, unrecognised units and classification gaps. Retain these measures as a baseline for comparison with the final output.
- 3Extract and normalise attributesExtract part numbers, sizes, colours, grades and pack quantities from descriptions into separate fields. A unit reference list maps equivalent labels such as “no”, “nos” and “pcs”.
- 4MatchRules identify candidate matches, followed by statistical scoring. Semantic similarity adds a lower-weight signal. Differences in part number, size or colour block a merge regardless of the overall similarity score.
Data processing · stages 5–8
From approval to ongoing data quality
- 1Review and approveCandidate matches are grouped into recurring cases. Your data owner approves a rule that applies to records meeting its conditions. Ambiguous and high-impact cases receive individual review.
- 2Build the golden recordApproved results form a golden record with its contributing sources, selected values, applied rule, rationale and approver. Published identifiers remain unchanged.
- 3Publish and reconcileDeliver versioned data packages with manifests and code mappings between source records, golden records and target identifiers. Reconcile record counts and explain discrepancies.
- 4Check new recordsCheck proposed records against the current master data before creation. If a likely duplicate is found, pause creation and show the requester the matches and supporting evidence.
AI support and human responsibility
AI proposes. Your data owners approve
What the platform does
- Assesses data quality and finds potential duplicates across sources
- Extracts part numbers, sizes and colours from descriptions
- Shows source evidence, the applied rule and a confidence score
- Refers ambiguous cases to an authorised specialist
What the platform does not do
- Merge records without an approved basis
- Treat similar descriptions alone as sufficient grounds for a merge
- Replace a value without preserving the original
- Write results directly into your ERP or other systems
Identifiers, part numbers, units and usage history take priority over wording. AI may propose a match or a group split. Merging requires a human decision or a human-approved rule, and every proposal can be reviewed.
Strengths of established MDM systems
Where established MDM systems have a broader role
Long-term data governance
Change requests, multi-stage approvals and distributed data stewardship teams across domains and countries, designed for long-term operation.
Deep ERP integration
Master data creation and updates within ERP interfaces, with built-in validation and distribution across the vendor's systems.
Broad domain coverage
Packaged capabilities for financial, employee, asset and reference data, supported by implementation partner ecosystems.
Nexus focuses on cleaning master data, preserving the basis for decisions and maintaining quality afterwards. Where an MDM system is already in place, Nexus can prepare data for handover to it.
Deployment, access and data languages
In your infrastructure, under your control
Cloud infrastructure
Nexus runs in your cloud account, in approved regions, and accesses AI models through your provider. Your data is not used to train foundation models.
Access you grant and revoke
You manage identities, network access and keys. Access to contact details and bank information is restricted by role, with actions recorded in an audit log.
Languages and scripts
Latin, Cyrillic and Arabic text is normalised for search and matching, including character variants, digits, punctuation and text direction. The original text is preserved.
After the initial data assessment, we provide a written estimate of AI model costs to help you set budgets and spending alerts in your cloud account.
Scope and responsibilities
What Nexus does, and what it does not
Nexus does
- Validate and prepare item, supplier and customer records for loading
- Preserve supporting evidence, code mappings and decision history
- Check proposed records for duplicates against the current master data
- Deliver results in an agreed format at project completion
Nexus does not
- Replace your ERP or maintain transaction records
- Extract from or load into your systems; your team or partner handles this
- Confirm bank accounts with banks or company registrations with official registries
- Operate without your team's input; your data owner approves the rules
Your team or migration partner handles integration with your systems. Requirements for ongoing data governance are discussed separately.
Sources for the product description and comparison
Sources
- [1]ARBA Nexus. Product architecture and domain packsDescribes implemented platform capabilities. Readiness for a customer's environment is confirmed during deployment and acceptance.
- [2]SAP. Master Data Governance for Material — configuration guide, SAP LibraryThe guide describes data model activation, change request configuration, workflow tasks and rule-based workflows. SAP and SAP Master Data Governance are trademarks of SAP SE.
- [3]IBM. IBM Match 360 — product documentationThe documentation describes source onboarding, data model generation, matching configuration and review by data stewards — the specialists responsible for data quality. IBM and IBM Match 360 are trademarks of IBM Corp.
- [4]Informatica. Master Data Management — product pageThe product page describes AI-assisted matching and steward review, and includes the vendor's statements on implementation time. Informatica is a trademark of Informatica LLC. Neither ARBA nor Nexus is affiliated with, endorsed by or a partner of SAP, IBM or Informatica.
Information about other products is drawn from their vendors' documentation; the Nexus description covers implemented platform capabilities