Before building a data warehouse in MaxCompute and DataWorks, familiarize yourself with the dimensional modeling concepts used throughout the workflow. These terms appear in steps such as defining metrics, configuring dimension tables, and building derived indicators.
The concepts below are organized into three groups:
-
Organizational terms — how business scope is divided (business unit)
-
Modeling terms — the building blocks of dimensional models (dimension, dimension attribute, measure, metric)
-
Constraint terms — how metrics are filtered, scoped, and aggregated (business filter, statistical period, statistic granularity)
Organizational terms
Business unit
A business unit is the division of business at a higher level than data domains. Use business units to organize large business systems that span multiple data domains.
Modeling terms
Dimension
A dimension is the entity object that serves as an environment for measures. Dimensional modeling, introduced by Ralph Kimball, organizes data models around business analysis and decision-making requirements. A dimension is a collection of business attributes that describe a business entity.
For example, when analyzing a transaction, you might describe the transaction environment using the buyer, seller, product, and time dimensions.
Dimension attribute
A dimension attribute is a column that makes up a dimension. Dimension attributes are the primary sources for query constraints, grouping, and report tags, and they determine how usable a dimension is for analysis.
Measure
A measure is a business performance value. In dimensional modeling, a measure is described as a fact and a dimension as an environment. A measure is typically numeric and is represented as a fact in a logical fact table.
Metric
A metric is a business measurement indicator. Metrics come in two types:
-
Atomic metric — an indivisible measure of a business process. An atomic metric has a clear statistical criterion and computational logic. For example, payment amount is an atomic metric.
-
Derived metric — an atomic metric within a business scope constrained by a time period and a modifier.
Constraint terms
Business filter
A business filter defines the business scope for a statistic. It identifies data that meets specific business rules.
SQL analogy: A business filter corresponds to the conditions in a WHERE clause, excluding any time-based conditions.
Statistical period
A statistical period is the time range for a statistic, such as the last day or the last 30 days.
SQL analogy: A statistical period corresponds to the time condition in a WHERE clause.
Statistic granularity
A statistic granularity is the object or perspective for statistical analysis. It defines the level of data aggregation and acts as the grouping condition for aggregation operations. A statistic granularity is a combination of dimensions that specifies the statistical scope.
SQL analogy: A statistic granularity corresponds to the object specified in a GROUP BY clause.
Example: To measure the turnover of a seller by province, the statistic granularity is the combination of the seller dimension and the region dimension. To collect statistics on the data of a table, the statistic granularity is the entire table.
The statistic granularity often serves as the modifier of a derived metric.