How to clean an industrial ERP database before a SAP migration

The deadline is approaching. SAP has confirmed that mainstream support for SAP ECC 6.0 will end in late 2027, with an extended maintenance option available until 2030 at an additional cost. Yet, according to SAPinsider’s 2025 research, only 28–33% of SAP ECC customers have completed their migration to SAP S/4HANA. For most industrial companies, the transformation is still ahead, even as demands for performance, traceability, and effective data management continue to grow.

These figures reflect a reality that CIOs and Manufacturing Directors know all too well: migrating without cleaning your data means carrying the same problems into your new system. Duplicate Part Numbers, Bills of Materials (BOMs) frozen at obsolete revisions, duplicate suppliers with conflicting addresses, all of these issues will follow you into the new system unless they are addressed beforehand.

Cleaning a general-purpose ERP database and cleaning an industrial ERP database are two fundamentally different exercises. Manufacturing companies face multiple layers of complexity that conventional IT teams consistently underestimate.

Industrial environments also involve a multi-system dimension: an industrial ERP interacts with PLM platforms (Teamcenter, Windchill, ENOVIA), CAD tools, MES solutions, and sometimes legacy AS/400 systems containing business logic buried in decades of code. Cleaning ERP data without mapping these dependencies is like repainting a building’s exterior without inspecting its foundations.

Effective industrial ERP data cleansing cannot be improvised. It requires a structured approach, from initial assessment through to final validation, with each phase laying the groundwork for the next. Here is the methodology we apply at KLERIA as part of our Zero-Backlog Engine™ approach.

Before any work begins, we map the actual state of your database: data volumes by entity (items, suppliers, BOMs, inventory, manufacturing routings, technical drawings, and quality documents), field-level completeness rates, duplicate detection, and identification of missing mandatory fields.

This phase produces a Data Maturity Assessment Report, an objective baseline of your existing data quality.

Not all data needs to be migrated. This step determines what should be migrated, archived, or deleted, together with the associated business validation rules.

Without formalized rules validated by business teams, each consultant or operator may interpret requirements differently, allowing inconsistencies to accumulate.

Part Number deduplication is often the most time-consuming phase in an industrial environment. It involves comparing item descriptions, reconciling coding systems across sites, and consolidating variants. For a database containing 50,000 items, an estimated 15–30% may consist of duplicates or unusable records in legacy databases, often resulting from mergers, acquisitions, or repeated manual data entry.

📋 Example: one component, four different Part Numbers In a multi-site ERP database following a merger, the same 125W electric motor exists under four different references:

From the ERP system’s perspective, these were four separate components with independent inventory records. Result: 3 duplicate references eliminated, inventory consolidated, and one enriched master item record created with all technical attributes merged.

Bills of Materials are at the operational heart of an industrial ERP system. This phase addresses obsolete revisions, phantom components (listed in the BOM but with no available inventory or active supplier), broken links to CAD drawings, and inconsistencies between manufacturing BOMs and engineering BOMs. Every poorly cleansed BOM represents a risk of non-conformity or production line downtime after migration.

Standardize supplier, customer, and site data according to the target SAP S/4HANA data model. This phase also includes enriching missing data and ensuring regulatory compliance — particularly important in light of the 2026–2027 deadlines: the rollout of mandatory e-invoicing and the end of SAP ECC support.

📋 Example: standardizing units of measure In a multi-site database resulting from the merger of three ERP systems, the same unit of measure “millimeter” appears in six different forms:

The same approach applies to product families, currencies, Incoterms, supplier address formats, and intra-community VAT numbers. Result: No unit conversion errors during the SAP migration.

Validation goes far beyond a superficial check. It involves several migration testing cycles using real data (not fictitious test data), quantitative and qualitative reconciliation between source and target systems, and quality certification for each data domain. Only data that has passed all validation rules is loaded into the production environment.

Industrial AI projects : predictive maintenance, production planning optimization, non-conformity detection, and technical assistants — all rely on one fundamental prerequisite: reliable, consistent, and well-governed data. Deloitte reports that 70% of companies are already experimenting with generative AI, but fewer than a quarter consider themselves ready to deploy it at scale. Data quality and governance remain among the main obstacles to wider adoption.

This is not a technology issue. Even the most sophisticated AI models fail when fed poor-quality ERP data. A predictive maintenance AI trained on incomplete manufacturing routings will generate inaccurate recommendations. A replenishment model relying on inconsistent inventory data will generate costly ordering errors.

The SAP S/4HANA migration is therefore a twofold strategic opportunity: it not only modernizes your management system, but also creates, when properly prepared, a clean data foundation that your AI, BI, and Industry 4.0 initiatives can leverage immediately. Missing this opportunity means delaying your organization’s ability to unlock the value of its industrial data by several years.

Developed by VAGANET Group teams through projects involving SAP, Teamcenter, Windchill, and Opcenter environments, the Zero-Backlog Engine™ method structures ERP data cleansing into traceable, measurable phases. It is based on a simple observation: in industrial environments, poor-quality data backlogs accumulate silently over the years — until a migration forces companies to address everything at once, under pressure.

Our approach anticipates this challenge through rigorous sequencing, domain-specific quality indicators, and operating procedures tailored to the field constraints of industrial engineering and IT teams.

Tools don’t define the methodology, but they help accelerate it. Here are the main categories of tools used in industrial ERP data cleansing projects, depending on the context and data volumes.