Survivorship in master data: the rules that decide which value survives in the golden record

The other day a client asked me which supplier data was the correct one. The CNPJ (Brazil's company registration number) was right in every system. The problem was the address, the phone number and the legal name, each one different in the ERP, in the purchasing records and in the spreadsheet that finance kept on its own.

I turned the question around: correct according to which system?

That was when I explained what survivorship is in master data. It is the discipline that decides, attribute by attribute, which value survives when the same customer, supplier or material is recorded in different ways in different systems. The result of that decision is the golden record, the single record the company starts to treat as true.

Survivorship is not a synonym for deduplication. Deduplication, or matching, is the step that finds out that three records are actually the same entity. Survivorship comes afterward, to resolve what to do with the values that do not match across those three records.

How it works in practice

The golden rule of survivorship is simple to state and hard to apply: the decision is made per attribute, not per whole record.

That means a supplier's golden record can carry the legal name from the ERP, the address from an external validation service and the purchasing category defined by a master data analyst. There is no single winning system. There is a winner for each field.

There are four common ways to decide that winner.

The first is source priority. You build a ranking of systems for each attribute. The ERP is usually the source of truth for tax data and legal name, because it is the one that talks to the invoice and to Sintegra (Brazil's state-level tax registration lookup system). The CRM, on the other hand, tends to be more reliable for phone and business email, because it is updated by the salesperson who talks to the customer every week. Defining this ranking forces the company to decide, once and for all, who rules each type of information, and that alone already settles half of the master data arguments I see in governance meetings.

The second is the most-recent criterion. The value updated last prevails. It works well when the most recent source is reliable, but it is dangerous when it is not. I have seen a company apply this rule to a supplier's phone number and discover that an intern had typed a wrong number the day before, and that wrong number became the official data across the whole company just because it was the newest.

The third is the completeness and quality criterion. Between two values, the winner is the one that passes the validations, the one with more characters filled in, the one that does not trigger an alert in the quality check. An address with CEP (Brazilian postal code), number and complement beats an address with just the street name.

The fourth is human decision. When no automatic rule is defensible, the conflict goes to the data steward to resolve manually. This is the typical case of two legal names that are similar but not identical, which require someone to look at the contract or call the supplier.

The point that usually goes unnoticed is that survivorship depends on well-done matching beforehand. If matching merges two different suppliers as if they were one, the survivorship rule will pick values from a group that should never have existed, and the golden record is born wrong with all the technical elegance in the world. That is why I always recommend conservative thresholds in the match step, especially in the material master, where two bolts that look alike in text can be completely different items on the shelf.

Who is responsible for this decision

In most of the companies I work with, the survivorship rule is born from a conversation among three roles.

The data steward knows the business process and knows, for example, that the customer's delivery address cannot come from the sales record, because the salesperson fills in anything just to close the deal. The data architect or MDM specialist translates that rule into configuration inside the tool. And the process owner, usually someone from purchasing, sales or finance, signs off on which system rules which attribute.

When that conversation does not happen, the IT team configures the rule on its own, by the easiest path, which is normally the newest system or the one implemented last. Result: a perfect technical rule deciding something that should be a business decision.

Which tools do this

The market's MDM platforms already come with a built-in matching and survivorship engine, because they are two sides of the same coin.

SAP Master Data Governance works this logic inside the SAP ecosystem itself, with consolidation rules configurable per master data domain. Informatica MDM has one of the most mature match and merge engines on the market. Stibo Systems is strong in product data, where attribute completeness weighs more than source priority. Reltio works with dynamic survivorship, recalculated as new data arrives. Semarchy and Profisee gained ground through the flexibility of configuring rules without depending so much on code. And Ataccama comes in from the data quality side, which feeds directly into the completeness criterion.

None of these tools solves the problem alone. I have seen a company buy the most expensive license on the market and keep poor master data, because nobody sat down to define which system owns which attribute. The tool applies the rule. Governance decides the rule.

How to learn about the subject

Survivorship is an advanced MDM topic, so the study path goes first through the basics of governance and data quality before getting here.

I recommend starting with the fundamentals of Master Data Management and the dimensions of data quality, then moving on to matching and deduplication, and only then studying survivorship and golden record construction. DAMA's CDMP certification, with the DAMA-DMBOK body of knowledge, covers this path in a structured way and is today the most recognized certification in the field. For those who already work with SAP, specific training in SAP Master Data Governance is also worthwhile, because the syntax changes, but the decision logic is the same.

In practice, nothing replaces taking part in a real master data consolidation project, sitting with the business team, discussing line by line why one system should win over the other.

Career opportunities

Those who master matching and survivorship for master data are no longer an ordinary master data analyst. They are the professional the company calls when two companies merge and the ERPs have to become one, when the SAP S/4HANA project requires deciding what goes into the new master, or when the company finally decides to tackle the supplier duplication that has been blocking purchasing for years.

That knowledge separates the operational master data analyst from the MDM specialist, the data architect and the governance consultant, roles with higher pay precisely because the wrong decision is costly. A poorly configured rule can overwrite a customer's correct tax record with the wrong one, and that becomes an invoicing, tax and audit problem, not just an IT one.

Main questions

Is survivorship the same thing as deduplication?

No. Deduplication, or matching, finds the records that represent the same entity. Survivorship decides which values from those records go into the final record.

What is a golden record?

It is the single, consolidated record of a customer, supplier or material, the result of applying survivorship rules to the records that matching identified as duplicates.

What is the most widely used survivorship rule?

Source priority, because it forces the company to decide, in writing, which system owns each attribute, instead of leaving that decision to whoever configured the system last.

Can survivorship be done without MDM software?

It can, manually or with a spreadsheet, but it does not scale. In companies with thousands of suppliers and materials, the rule has to be automated in a tool, with an exception for the cases that require a human decision.

Who defines the survivorship rules in the company?

Ideally it is a joint decision among the data steward, the business process owner and the technical MDM team, never an isolated IT choice.

MDM Academy

This kind of content, which mixes technical concept with business decision, is exactly what we teach at MDM Academy, 4MDG's school of governance and master data.

MDM Academy's proposal is not to train people who memorize tool names, but professionals who understand why a survivorship rule exists, who know how to sit down with purchasing, sales and IT to decide which system rules which data, and who come out ready to lead a real master data consolidation project, not just operate software.

If you want to go deeper into MDM, governance and data quality with this practical view, it is worth getting to know MDM Academy or joining Data Club, our community of people who live master data in the day-to-day of Brazilian companies.

Conclusion

Every large company already has its data scattered across three, four, five different systems. That is not the problem. The problem is not having decided, before any consolidation project, which of those systems rules each piece of information.

Survivorship is that decision written down, tested and automated. Without it, every golden record project becomes a technical rule applied over an organizational mess that nobody had the courage to resolve before.

At your company, when master data diverges between systems, who decides which value counts: the rule, the most recent system, or is it still a shouting match between departments?