Data lineage (in Italian lignaggio o tracciabilità dei dati) is the body of information describing a piece of data’s life cycle: the system where it originates, the steps it passes through, the transformations it undergoes and the reports or applications where it ends up. Read backwards, it lets you trace a number back to its source; read forwards, it shows what depends on that data and what breaks if you change it.
Online, most of what you find on the subject is a description of software. Yet our starting point here is more concrete: a number in a Board of Directors report and someone asking where it comes from. Data lineage makes it possible to answer by looking it up rather than reconstructing it by hand, but only if every step of the path has a person accountable for it. That is why it is part of corporate governance before it is an IT project.
What data lineage is for
The benefits of data lineage become clear the day a number does not add up. The margin per customer in the monthly report changes for no apparent reason, two spreadsheets give different values for the same fact, an auditor asks where a figure comes from. Without a written history of the data, the answer depends on the memory of whoever knows the systems. With lineage, you simply open the path.
It supports three decisions:
- Assessing the impact of a change. Deleting a column used in a table join can break the reports that depend on it (Informatica): lineage tells you in advance which ones.
- Finding the source of an error, tracing back from the incorrect value to the step where it was altered (IBM).
- Proving to an auditor or an authority where a number comes from.
It is not equally valuable for all data. It pays off where errors are costly or where the question comes up often: financial statements, regulatory reports, KPIs presented to the Board, customers’ personal data. For data used by a single person in a spreadsheet, keeping it up to date costs more than it is worth.
Data lineage in Italian: lignaggio, tracciabilità and provenance
Lignaggio dei dati is the literal Italian translation, but Italian texts almost always use the English term, sometimes alongside “tracciabilità dei dati” (data traceability). The closest concept is data provenance, which refers to the first instance of the data, that is, its origin; lineage also covers the journey and the intermediate transformations (IBM). Provenance is the starting point; lineage is the whole road.
Related concepts, and what sets them apart
Five terms that come up alongside lineage and are often confused with it.
| Concept | What it is | Difference from lineage |
|---|---|---|
| Data provenance | Where the data comes from | Lineage tells the whole journey, not just the origin |
| Data mapping | Mapping between source and target fields | It is an ingredient: it indicates a relationship, lineage shows the chain |
| Data catalog | Inventory of data and its meanings | The catalogue says what exists; lineage says where it comes from |
| Audit trail | Chronological record of who did what in a system | It records events, not the relationships between data across different systems |
| ETL process | Extraction, transformation and loading of data | It produces part of the path; lineage documents it |
Types of data lineage: horizontal, vertical, business and physical
Horizontal and vertical data lineage have a precise meaning in Italian sources, with a third type bridging the two layers. According to Irion EDM, there are three types:
The three types of lineage
Business horizontal lineageBusiness layer
How business processes exchange information.
Which processes produce this indicator?
Physical horizontal lineageTechnical layer
How data moves between applications through IT flows.
Which systems and tables does it come from?
Vertical lineageConnection
How a field in a table represents a data element defined by the business.
Which field implements this definition?
The three do not carry the same weight. According to the same source, software vendors long treated only the physical type as relevant, even though its automated detection has technical and semantic limits, while inspectors look more at how data passes between processes than at files moving between systems. Answering a regulator’s inquiry requires business and vertical lineage together (Irion EDM).
How data lineage works
Metadata is the raw material. Technical metadata describes the structure and format of data, such as column name, data type and access permissions; business metadata describes its meaning, such as business name, process, data steward, applicable regulation and confidentiality level (Erika Pinto, Cyber Security 360, 2019). The former is extracted from systems. Much of the latter lives in people’s heads and has to be written down.
There is also a practical limit: automated detection cannot see what moves outside the systems. The spreadsheet sent by e-mail and the manual correction made before the report are the points where a number changes without leaving a trace, and they must be recorded separately, with a procedure before a tool. To make different tools work together there is the open standard OpenLineage, which describes lineage in terms of datasets, jobs and runs (OpenLineage).
The Bliss view
Who is accountable for the number
For us, data lineage is first and foremost a matter of accountability. Behind every number that reaches a decision-maker lies a chain of steps, and each step has an author: whoever wrote the calculation rule, whoever corrected the file by hand, whoever approved the definition. If none of them can be identified, the mapped path is accurate and unusable, because decision-makers do not ask which table feeds the number: they ask whom they can trust.
Two choices follow from this that matter more to a company than the tool.
- Scope. You trace what the Board uses to make decisions or must be able to demonstrate, not the entire data estate.
- Assignment. Every critical data element has a data owner, accountable for its meaning, and a data steward, who manages it day to day. These are the roles of data governance: without them, lineage lasts only as long as the project that produced it.
A test that takes half an hour: take three numbers from the latest Board report and, for each one, ask which systems it comes from, who touched the data along the way and who is accountable for it.
Where the answer is missing, lineage is already needed.
The Bliss Agency perspective
Data lineage, GDPR and BCBS 239
No regulation mandates data lineage by that name, but many require you to be able to demonstrate what you do with data. The GDPR requires the controller to be able to demonstrate compliance with its principles (Article 5(2)) and to keep a record of processing activities (Article 30); knowing which systems hold personal data is the precondition for being able to erase it (Cyber Security 360, 2019).
In banking, the Basel Committee’s report of 28 November 2023 on the BCBS 239 principles notes that the 31 global systemically important banks are at different stages of alignment, almost ten years after the principles were published (Bank for International Settlements). The page does not mention data lineage: reading it as the most direct response to the demand for reliable risk data is our interpretation, not part of the principles. We draw a practical criterion from this: lineage is built to answer a specific question from an auditor or a regulator, not for the sake of compliance.
How to build data lineage
Building data lineage starts with a short list, not a map of the systems. You select the critical data, meaning the data on which the financial statements, legal obligations and senior management decisions depend, and assign a data owner and a data steward to each. Only then are the metadata collected.
Next, the path from source to report is reconstructed, marking which steps are automated and which are manual, and represented in two views: a business view for decision-makers and a physical view for those who work on the systems, with a vertical link between the two. Take the “margin per customer” indicator as an illustrative example. The business view shows which processes produce it and with which definition of margin; the physical view shows which systems and tables it comes from; the vertical link ties the definition of margin to the field that calculates it. Then there is maintenance, the part that gets forgotten: someone must be responsible for updating the lineage when a system changes and for checking that it still matches the actual systems.
Data lineage tools: the tool comes later
A data lineage tool collects metadata and presents it as a graph or table; catalogues, governance platforms and data integration tools offer it to varying degrees. Before choosing one, three answers are needed: which systems to cover (if the answer is “all of them”, the problem is the scope), who the representation is for and how it stays up to date. To judge whether it works, counting mapped systems is not enough: what matters is coverage of critical data, alignment with the actual systems and the time needed to answer a question about where a piece of data comes from.
Data lineage and related concepts: the differences
Five terms are often used interchangeably with lineage, and each covers only part of it:
Related concepts, and what sets them apart
Five terms that come up alongside lineage and are often confused with it.
| Concept | What it is | Difference from lineage |
|---|---|---|
| Data provenance | Where the data comes from | Lineage tells the whole journey, not just the origin |
| Data mapping | Mapping between source and target fields | It is an ingredient: it indicates a relationship, lineage shows the chain |
| Data catalog | Inventory of data and its meanings | The catalogue says what exists; lineage says where it comes from |
| Audit trail | Chronological record of who did what in a system | It records events, not the relationships between data across different systems |
| ETL process | Extraction, transformation and loading of data | It produces part of the path; lineage documents it |
Domande frequenti
What is data lineage?
It is the reconstruction of a piece of data’s journey: where it originates, what transformations it undergoes and where it is used. It helps to assess the impact of a change, trace the source of an error and prove where a number comes from. It is information to be kept up to date over time, not a diagram to be produced once and filed away.
What does data lineage mean in Italian?
The literal translation is lignaggio or discendenza dei dati, but in practice the English term is used, sometimes alongside tracciabilità dei dati. It is not the same as provenance, which refers only to where data originates: lineage also covers the journey and the intermediate transformations, through to final use.
What are the types of data lineage?
According to Irion EDM there are three: horizontal business lineage, which shows how processes exchange information; horizontal physical lineage, which follows data across systems; and vertical lineage, which connects the business layer to the technical one. Some English-language sources use the same terms with a different meaning.
What is the difference between data lineage and data provenance?
Provenance indicates where data originates, that is, its first instance. Lineage tells the full journey: origin, intermediate transformations and destination. Provenance is therefore one part of lineage, useful above all when you need to prove where a piece of data came from.
What is the difference between data lineage, data mapping and a data catalog?
Mapping establishes the correspondence between the fields of a source system and those of a target system; the catalogue is the inventory of data with their meanings. Lineage uses both: it chains together the correspondences from the mapping and is easier to read if every data element has a clear meaning in the catalogue.
Is data lineage part of data governance?
Yes, it is one of the tools that make data governance verifiable: governance sets rules and responsibilities, and lineage shows how the data complies with them along the path. Without someone accountable for each step, however, it remains a representation and does not become a governance tool.
Is data lineage mandatory for compliance?
No regulation requires it under this name. The GDPR requires organisations to be able to demonstrate compliance with its principles and to keep a record of processing activities; the Basel principles require banks to have reliable risk data. Lineage is one way of meeting these requirements, not the only one.
Do you need software to do data lineage?
Not at the outset. The decisive work is selecting the critical data, assigning owners and mapping the steps, including manual ones, that no tool detects on its own. Software helps when the scope is clear and the systems are many; without these choices, it automates a path that nobody knows how to read.
Who is responsible for data lineage?
The content is the responsibility of those who manage the data: the data owner for meaning and quality, the data steward for day-to-day management. Technical functions extract the metadata and keep it up to date. Deciding which data to trace ultimately rests with those who govern the business.
Fonti e riferimenti
- IBM, Cos’è il data lineage?
- Informatica, What Is Data Lineage and Why Is It Important?
- Irion EDM, Le dimensioni del Data Lineage
- Secoda, What distinguishes vertical lineage from horizontal lineage in data management?
- Erika Pinto, Il data lineage nella gestione del patrimonio informativo aziendale: la strategia corretta, Cyber Security 360, 15 febbraio 2019
- OpenLineage (LF AI & Data Foundation), About OpenLineage
- Basel Committee on Banking Supervision, Progress in adopting the Principles for effective risk data aggregation and risk reporting, 28 novembre 2023
- Parlamento europeo e Consiglio, Regolamento (UE) 2016/679 (GDPR)

