AI glossary · Basic · Data engineering
OLAP
OLAP (online analytical processing) is a way of processing data that groups and aggregates stored data by multiple criteria such as period, region and product so it can be analyzed quickly. Rather than recording individual transactions, it is designed to answer aggregate questions such as 'What were sales by region last quarter?'
About this term
- Level
- Basic · Technical terms you often meet in AI news
- Category
- Data, ops & platforms
- Fields
- Data engineering
- Added
- 2026-10-09
How it works
OLAP divides data into 'measures' and 'criteria'. Measures are numbers to be summed or averaged, such as sales and order counts, and criteria (dimensions) are what you group by, such as date, region and product. Users look at the same data from several angles, for example drilling down from years to months (drill-down) or slicing out a specific region. To handle such questions quickly, OLAP systems pre-aggregate data or use storage formats suited to analysis, such as columnar storage.
Why it matters
Most company dashboards, monthly reports and sales analyses are OLAP-style questions. Sending these heavy aggregations straight to the operational database that processes transactions (OLTP) can slow down the service, so a separate system is kept for analysis. Data warehouses and data lakehouses are exactly where this analytical work happens. Recently, Text-to-SQL use cases, where an LLM writes SQL in place of a person to query OLAP systems, are also increasing.
OLTP and OLAP
OLTPTransaction processing
- Records each order and payment
- Frequent writes and updates
- Service operational database
OLAPAnalytical processing
- Groups and aggregates many rows
- Mostly reads
- For dashboards and reports
공통 Both are database technologies, but the shape of the questions differs
Things to know
OLAP is strong at 'reading and aggregating' but is not suited to frequently modifying records one at a time. So operational data has to be moved into the analytical system using methods such as ETL/ELT or CDC, and this can introduce delays of minutes to hours. In the past, it was common to build predefined aggregation structures called 'cubes', but today it is common for columnar databases to aggregate data close to the original source directly. Because even the same 'sales' figure can be calculated differently by different teams, a semantic layer is sometimes used to align definitions.
Example
When a shopping site's operations team checks 'order counts by category and region for the last 3 months' on a BI dashboard, behind the dashboard an OLAP system such as a data warehouse groups and aggregates a large number of order records and returns the result. By contrast, saving a single order when a customer clicks the payment button is handled by an OLTP database.