Why the vocabulary matters
Data discussions often stall because people use the same word for different things. One team’s “customer record” is another team’s “account”, and a request for “clean data” can mean anything from fixing typos to rebuilding a whole process. Agreeing on a few terms makes conversations about ownership and quality shorter and less heated.
The definitions below are general and describe common usage. Individual organizations and published standards may define the words more narrowly.
Terms for what the data is
Data. Recorded facts or values, such as a name, a date, an amount or a status. Data becomes useful information when it is placed in context.
Record. A single complete set of related facts about one thing, such as one customer or one purchase order. In records management, the word also has a narrower meaning: a document or entry that must be kept as evidence of an activity.
Field. One named piece of a record, such as “postal code” or “order date”.
Master data. The core reference information that many processes use over and over, for example customers, products, suppliers, employees and locations. It changes relatively slowly compared with day-to-day activity. The discipline of keeping it consistent is the subject of master data management.
Transactional data. Data about individual events, such as a sale, a payment or a delivery. Each transaction usually points back to master data, for example to the customer who made the purchase.
Reference data. Agreed lists of allowed values, such as country codes or order statuses. Using the same list everywhere helps different systems agree.
Structured and unstructured data. Structured data fits into defined fields, as in a database table. Unstructured data, such as emails, documents and images, does not, and is often handled through document management systems.
Terms for describing and tracing data
Metadata. Data about data. Examples include a field’s meaning, who created a file, when it was last changed and how often a table is refreshed. Good metadata helps people find data and understand it before they rely on it.
Data dictionary. A document or tool listing each field with its meaning, format and allowed values. It is a practical way to settle what “active customer” means.
Data catalog. A searchable inventory of an organization’s data sets, usually built from metadata, so people can find what exists and who looks after it.
Data lineage. The path data takes from its origin through each system or transformation to where it is finally used. If a figure in a report looks wrong, lineage helps trace where it went astray. It is also useful when someone needs to know which reports will be affected if a source system changes.
System of record. The system treated as the official home for a given type of data. Other systems may hold copies, but when copies disagree, the system of record is the reference. This idea underlies the search for a single source of truth, which is easier to name than to reach fully in organizations with many systems.
Terms for responsibility
Data owner. A person, usually a senior one, who is accountable for a data set: who may use it, how it is protected and how important decisions about it are made.
Data steward. A person who looks after the day-to-day quality and correct use of a data set, applying the rules the owner and governance forums set. The responsibilities are discussed under data stewardship, and are often added to an existing job instead of being a full-time role.
Data custodian. The person or team that runs the technology where data sits, such as storing it, backing it up and controlling technical access. Custodians usually keep data safe but do not decide what it means.
Data governance. The overall set of roles, policies and decision processes for managing data as an organizational asset. Governance answers questions such as who decides on a new definition and how disagreements are resolved. It is mostly about accountability, not tools.
Terms for quality and movement
Data quality. How well data fits its intended use. Common dimensions include accuracy, completeness, consistency, timeliness and uniqueness. Quality is judged against purpose: data that suits one use may be inadequate for another.
Data cleansing. Correcting or removing data that is wrong, incomplete or duplicated. Cleansing without fixing the cause of the errors often leads to the same problems returning, which is why data quality issues are usually traced back to how data is captured.
Duplicate record. More than one record for the same real-world thing, such as a supplier entered twice with slightly different spellings.
Data integration. Combining data from different sources so it can be used together.
ETL. Short for extract, transform, load: pulling data from a source, reshaping it, and placing it in a target such as a reporting database. Variations of the order exist.
Data migration. Moving data from one system to another, normally during a replacement or consolidation. It involves mapping fields, cleansing and checking, and it is a distinct task from day-to-day integration.
Terms for keeping and removing data
Retention. How long data or records are kept. Periods are usually set out in data retention schedules and may be influenced by legal, contractual or business requirements, which a qualified adviser can interpret for a specific organization.
Archiving. Moving data that is rarely used into cheaper or separate storage while keeping it retrievable.
Disposal. Deleting or destroying data when it is no longer needed or required to be kept. Doing this consistently, and recording that it happened, is part of good practice.
Personal data. Information relating to an identifiable living person. Many jurisdictions have specific rules about how it may be collected and used, so organizations that handle it should take advice from privacy and legal professionals.
A note on using these terms
Using precise terms does not guarantee good data. Definitions are most helpful when they are written down once, stored where people can find them, and applied the same way across teams. Even then, some disagreement is normal: different parts of a business sometimes need data to mean slightly different things, and the aim is to make the differences explicit rather than hidden.