Finance and banking data glossary

The Preql glossary defines the terms finance and data teams use when they work on shared numbers, from semantic layers and data lineage to core conversions and BCBS 239. Each entry gives a short definition and how it applies in finance and banking. The definitions are written to be quoted as they stand.

Editor's note: the original intro also says each term has its own page and links to an in-depth guide. Restore those lines once the pages and links are confirmed. ALCO, call report and CECL are new entries, because other articles use those terms.

Terms

ALCO. ALCO is a bank's asset-liability committee, which oversees interest rate risk, liquidity and the balance sheet. Its reporting package draws on figures from many systems, so consistent definitions matter.

BCBS 239. BCBS 239 is the Basel Committee's 2013 set of principles for how the largest banks aggregate and report risk data. Smaller banks often use it as a benchmark for showing that risk figures are accurate, complete, timely and traceable.

Business intelligence. Business intelligence is the set of tools and practices used to report on and analyze business data. It turns stored data into dashboards and reports people use to make decisions.

Call report. The call report is the quarterly report of condition and income that US banks file with their regulators. Its figures come from the GL and other source systems, so each line should be traceable.

CECL. CECL, the current expected credit loss standard, requires banks to estimate lifetime expected credit losses on loans and other assets. Its inputs come from loan, risk and accounting systems that often define terms differently.

Chart of accounts mapping. Chart of accounts mapping matches one entity's general ledger accounts to another's, so balances roll up the same way in consolidated reports. It is one of the first data tasks after a merger and a frequent source of reconciliation breaks.

Context layer. A context layer records what a company's numbers mean, including where each metric lives, how it is calculated and which version is approved. AI tools and people read from it, so their answers can use the same agreed definitions.

Core conversion. A core conversion moves a bank's accounts and transactions from one core banking system to another, often after an acquisition. Balances, product codes and branch hierarchies all have to be mapped from the old system to the new one.

Cost center mapping. Cost center mapping links each cost center in one system to its match in another, such as an acquired bank's cost centers to the acquirer's. Errors show up as expenses landing in the wrong branch or business line.

Data catalog. A data catalog is an inventory of an organization's data assets and their metadata. It helps people find data and understand where it came from.

Data cleaning. Data cleaning finds and fixes errors, duplicates and inconsistencies in data. It makes data reliable enough to report on and analyze.

Data governance. Data governance is the set of roles, rules and processes that decide how data is defined, owned, accessed and changed. In finance it determines who can say what a number means.

Data ingestion. Data ingestion moves data from source systems into a place where it can be stored and used. It is usually the first step in any reporting pipeline.

Data lineage. Data lineage traces a number back through the systems and transformations that produced it. Auditors and examiners may review it to confirm that reported figures come from approved sources.

Data storage. Data storage is where an organization keeps its data, such as databases, warehouses and lakehouses. Storage decides where data lives, not what it means.

Data transformation. Data transformation reshapes raw data into a form that is ready to analyze, such as joining, filtering and aggregating tables. The rules used in transformation often contain business definitions.

Dimensions. Dimensions are the categories used to group and filter metrics, such as region, product or month. The same metric can be sliced by many dimensions.

ETL. ETL stands for extract, transform and load, the process of pulling data from sources, reshaping it and loading it into a destination. Many teams now load first and transform later, called ELT.

Filtering. Filtering limits a dataset to the rows that meet a condition, such as one region or one quarter. Different filters on the same metric are a common reason reports disagree.

Metric layer. A metric layer stores how each metric is calculated, such as net interest margin or churn, and is often part of a semantic layer. Defining each metric once lets dashboards and queries that use it return the same number.

Metrics. Metrics are quantified measures of business performance, such as revenue, margin or deposit growth. A metric is only useful when everyone agrees on its definition.

Reconciliation load. Reconciliation load is the time a finance team spends making the same number agree across systems that each define it differently. It tends to grow after a merger or core conversion, when two sets of definitions have to be matched.

Semantic layer. A semantic layer translates raw database tables into business terms like revenue or active customers, so reporting tools query consistent definitions. It is usually built and maintained by a data team on top of a warehouse.

Single source of truth. A single source of truth is one agreed place where each metric is defined and calculated, so every report returns the same number. In practice it is a set of shared definitions more than a single database.

How we write definitions

1. Two sentences, in plain language.
2. One meaning per term, the one finance and banking teams use.
3. Written for the people who use the term at work.