Data cleansing
Data cleansing: turn a dirty database into a reliable asset.
Data cleansing is the process of cleaning, standardizing, deduplicating, validating and enriching a company's master data to make it unique, accurate and reliable. 4MDG cleanses material, supplier and customer data with AI and more than 600 public sources — and installs the governance that keeps bad data from coming back.
In just a few days you see the duplicates, inconsistencies and risks hidden in your database. No commitment required.
Trusted by 160 clients
Integrations
What we cleanse
Three cleansing fronts, with individuals and legal entities in the first two.
Customer data cleansing
Individuals and legal entities
Identity validated against public registry databases, registration status normalized, address and postal code standardized, verified contacts and duplicates merged into a golden record. A database ready for billing without rejections.
Customer onboarding →Supplier data cleansing
Individuals and legal entities
Business registration status, ownership structure, tax regime, certificates, sanctions lists and verified bank details. Duplicate suppliers unified and irregular records flagged before payment.
Supplier qualification →Cleansing of materials, products and services
The entire catalog, item by item
Rule-based standardized descriptions, attributes by family, unit of measure, packaging, product classification codes and UNSPSC, translation and deduplication. Covers materials, MRO, resale products and service catalogs.
Material standardization →
You can start with a single domain: the diagnostic shows which one delivers the most return.
The invisible cost
Your dirty database is already costing you — it just isn't on the invoice.
The same customer registered three times. A supplier with no validated registration. "Screw", "scr." and "SCREW M6" as different items.
The result shows up as duplicate purchases, wrong payments, stockouts, tax fines and reports nobody trusts.
US$ 12,9M
is the average annual cost of poor-quality data per company. Gartner.
up to 25%
of annual revenue is what incorrect data can consume. 4MDG reference.
Bad data stalls ERP, CRM, BI — and now AI
A model trained on dirty data doesn't make fewer mistakes: it fails at scale, with the appearance of certainty.
A diagnostic on a sample of your database, with nothing to install.
Concept
What is data cleansing?
Data cleansing, also called data hygiene or data scrubbing, is the set of techniques that corrects and organizes a master database: it removes duplicates, standardizes descriptions and formats, fills in missing fields, validates information against official sources and eliminates invalid records.
The goal is the golden record: a single reliable version of each material, supplier and customer, published to every system that consumes that data.
Saneamento, limpeza, higienização e qualidade de dados: nomes diferentes para o mesmo trabalho. Qualidade de dados (data quality) é o resultado medido; saneamento de cadastro é o projeto que entrega esse resultado; e data cleansing é o nome do mesmo processo em inglês.
Terminology
Scrubbing, deduplication and enrichment: what's the difference?
Scrubbing
Corrects and standardizes what is wrong or off-standard: formats, descriptions, units, required fields left blank.
Deduplication
Identifies and merges repeated records into a single golden record, with similarity scores and survivorship rules.
Enrichment
Completes and updates records with data from reliable sources: tax status, CNAE, address, certificates, technical attributes.
At 4MDG, all three happen together and continuously — not as a one-off cleanup that unravels the following month.
Scrubbing, deduplication and enrichment in the same flow.
PDM and PDS
There is no clean database without standard descriptions.
PDM and PDS are the descriptive standardization disciplines that make cleansing last. Deduplicating without standardizing treats the symptom: next month the same item returns under another name.
PDM
Descriptive standardization of materials
Describing each item in a unique, structured way: taxonomy by family, attribute templates, short and long descriptions generated by rule, unit of measure, packaging and tax classification. It's what makes items findable in search, prices comparable and duplicates visible before they are created.
See the PDM page →PDS
Descriptive standardization of services
The same discipline applied to the service catalog: clear scope, unit of measure, measurement criteria and tax linkage. Without it, "preventive maintenance" becomes thirty records, each with its own price and terms.
See the PDS page →Standardize before deduplicating
With structured attributes, matching compares apples to apples: similarity scores become far more precise.
A standard that protects the database
Rule-generated descriptions keep the next requester from creating the same item with different free text.
Services dirty the database too
Contracting services without scope or unit of measure creates duplicates and disputed measurements, just like materials.
The diagnostic measures how many items are missing attributes, off-standard or duplicated.
No SAP
Saneamento de cadastro no SAP.
Mestre de materiais (SAP MM)
No SAP, o cadastro de materiais vive no mestre de materiais, e cada item chega a ele com descrição padronizada, atributos e classificação.
Parceiro de negócios (BP)
No SAP S/4HANA, clientes e fornecedores ficam no parceiro de negócios, e o saneamento valida CPF e CNPJ e unifica duplicados antes da carga.
Carga por IDoc, BAPI e arquivo
A base tratada entra no SAP ECC ou S/4HANA por API, IDoc, BAPI ou arquivo, no dia a dia e nos projetos.
Rollout e migração para S/4HANA
Em rollout ou migração, a padronização acontece antes da carga: o sistema novo nasce limpo.
Governança depois do go-live
Regra na entrada e workflow de aprovação mantêm o cadastro limpo depois da carga.
How to do it
How to cleanse your data in five steps.
STEP 01
Diagnostic
We measure the state of the database: duplicates, inconsistencies, empty critical fields, invalid records and hidden tax risks.
STEP 02
Standardization
We apply taxonomy and rules: short and long descriptions, formats, units, classifications and naming by family.
STEP 03
Deduplication
We identify and merge repeated records into a golden record, with similarity scores and attribute-by-attribute survivorship.
STEP 04
Validation and enrichment
We verify and complete records against more than 600 public sources, with AI support: registration status, CNAE, address, certificates and technical attributes.
STEP 05
Continuous governance
We install entry rules, approval workflow and indicators so bad data doesn't come back the following month.
The fifth step is what sets 4MDG apart from a common cleanup.
Your database doesn't get clean just once — it stays clean.
The fifth step is what keeps the database clean after the project.
Why 4MDG
Different from a common cleanup.
Cleansing and governance together
We solve the cause, not just the symptom. Entry rules and workflow keep the problem from coming back.
AI and more than 600 public sources
Validation and enrichment at scale, checked at the source with evidence stored per record.
All three domains
Materials, suppliers and customers — not just a marketing mailing list.
Native SAP integration
The treated database enters the ERP clean, via API, IDoc, BAPI or file.
Software and people (DaaS)
We can run your cleansing continuously for you, under an agreed SLA.
ISO/IEC 27001 and LGPD
Your data handled under a certified management system, in compliance with LGPD and GDPR.
Proof
Clean data, measurable results.
−87%
Wella Company · supplier SLA, with −78% for customers and −84% for materials, and 100% of records managed by 4MDG.
+174k
Industry (SAP) · records standardized and cleansed, with optimized e-procurement and fewer purchases from wrong descriptions.
60M
records under management, for 160 clients in 6 countries.
Cristália · indirect materials cleansing: a clean, standardized MRO and consumables database, with operational improvement and cost reduction.
Uma indústria com SAP saneou e padronizou mais de 174 mil registros, com e-procurement otimizado e menos compra por descrição errada.
Cristália · saneamento de materiais indiretos: base de MRO e uso e consumo limpa e padronizada, com melhoria operacional e redução de custos.
Quero esses resultados na minha baseComo escolher
Como escolher uma empresa de saneamento de cadastro.
Metodologia de PDM documentada
Peça o padrão descritivo por escrito: taxonomia, atributos por família e a regra da descrição curta e longa.
Deduplicação com score e survivorship
A unificação precisa mostrar o score de similaridade e a regra que decide qual valor sobrevive no golden record.
Validação em fontes oficiais
CPF, CNPJ, inscrição estadual e situação cadastral conferidos na origem, com evidência guardada por registro.
Integração ao ERP sem planilha
A base tratada precisa entrar no ERP por API, IDoc, BAPI ou arquivo, sem retrabalho manual em planilha.
Governança depois do projeto
Sem regra na entrada e workflow de aprovação, a base volta a sujar em poucos meses.
Segurança certificada: ISO 27001 e LGPD
Seu cadastro deve ser tratado sob sistema de gestão certificado ISO/IEC 27001 e em conformidade com a LGPD.
Casos no seu setor
Peça resultados medidos em empresas parecidas com a sua, com número e nome do cliente.
Equipe que opera o cadastro
Se você quiser terceirizar, a mesma empresa deve poder operar o saneamento de forma contínua, com SLA.
What you gain
What changes when your data becomes reliable.
Fewer duplicate purchases
Unique, findable items cut repeat purchases and idle inventory.
A month-end close that adds up
Spend by category, ABC curve and supplier views stop diverging between areas.
Faster onboarding and qualification
Records done right the first time shorten the path to the first purchase or sale.
Less tax and compliance risk
Registration status, certificates and tax classification checked at the source and revalidated.
Data ready for BI
Reports built on a single database stop being a debate over which number is right.
Data ready for AI
AI models and agents only deliver value on reliable, traceable master data.
BI and AI only deliver value on reliable master data.
Investimento
Quanto custa um saneamento de cadastro.
O preço depende do tamanho e da profundidade do trabalho, não de uma tabela fixa. Cinco fatores definem o investimento.
Por isso o primeiro passo é medir: o diagnóstico gratuito conta os registros, as duplicidades e os campos críticos da sua base antes de qualquer orçamento.
Volume e domínio
Quantos registros e de quais cadastros: materiais, clientes ou fornecedores.
Profundidade
Só deduplicação ou padronização descritiva completa, com NCM e UNSPSC.
Fontes e formato
Quais fontes externas são consultadas, e se é projeto único ou Data as a Service mensal.
O diagnóstico mede o volume antes de orçar, sem instalar nada.
Contact
Let's look at your database before proposing anything.
A specialist analyzes your scenario and shows the current state of records in your industry. No commitment required.
Av. Queiroz Filho, 1700, Torre E, Conjuntos 715 a 718
Vila Leopoldina, São Paulo, SP, 05319-000
FAQ
Questions about data cleansing
It is the process of cleaning, standardizing, deduplicating, validating and enriching a master database to make it unique, accurate and reliable — the so-called golden record for each material, supplier and customer.
They are nearly synonyms. Cleansing usually covers the full process, including deduplication and enrichment; scrubbing emphasizes correcting and standardizing whatever is off-standard.
In five steps: a diagnostic of the current state, standardization with taxonomy and rules, deduplication into golden records, validation and enrichment against official sources and, finally, continuous governance so bad data doesn't come back.
It depends on the volume and state of the database. 4MDG's free diagnostic shows the real size of the job within days, with the number of duplicates and incomplete or irregular records.
Yes, all three master data domains, plus services and reference data. The model, rules and data sources change by domain.
With entry rules, approval workflow and monitored quality indicators — or with 4MDG's Data as a Service, which runs cleansing and data maintenance continuously, under an agreed SLA.
Yes, natively. The treated database enters SAP ECC or S/4HANA and other systems clean, via API, IDoc, BAPI or file, both day to day and in migrations and rollouts.
Yes. ADAM, 4MDG's AI, interprets free-text descriptions, builds the standardized description, suggests classifications such as product classification codes and UNSPSC and flags similar records — always with human validation before publishing.