Skip to main content

Create Atomic Metrics

1 Concept

Atomic metrics are metrics that cannot be further decomposed in business definitions. They are the most basic metrics, and subsequent composite metrics and derived metrics can be created based on atomic metrics. Atomic metrics are typically measurements under a specific business event behavior, used to quantitatively统计 products or business performance, usually through aggregation calculations, such as net profit = sum(order net profit), total transaction volume = count(distinct order number).

2 Operation Steps

  1. Select "Metrics Center > Metric Topics" in the navigation bar, click the "New Metric" button, and select "Atomic Metric".

  2. In the "New Atomic Metric" dialog, set the basic properties, calculation properties, and management properties of the atomic metric.

    |550

Basic Properties:

ParameterDescription
DatasetSelect a dataset with usage permissions to create metrics using the content of this dataset.
NameSet the Chinese name of the metric, which cannot be duplicated within a topic.
English NameSet the English name of the metric, which cannot be duplicated within a topic.

Calculation Properties:

Parameter
Description
Calculation Method
  • Source Field: Select a field from the dataset as the source field for calculating the metric value.

  • Aggregation Method: Set the aggregation method logic for this field, such as sum, minimum, maximum, etc.



  • Use Expression: Use custom expressions to define the aggregation result of the metric, which can reference fields from the currently selected dataset for detailed calculations first.

    Note: The outermost layer of the expression must be an aggregation function.

    • Data FormatSet the display format of the metric result.
    Applicable DimensionsSelect fields from the dataset as dimensions for subsequent metric analysis under different dimensions.

    When there are time fields in applicable dimensions, they can be selected as time dimensions for subsequent time-related derivative configurations.


    Filter ConditionsSupports adding multiple filter conditions. When calculating metric values, the final result will be calculated according to the filter conditions.
    Filter Conditions
    Business DefinitionMaintain the business meaning of the metric for subsequent metric applications.

    Management Properties:

    ParameterDescription
    Save LocationSet the save location for this metric.
    Responsible PersonThe person responsible for explaining the business definition, responsible for the reasonableness and accuracy of the business definition.

    3 Query Acceleration

    Query acceleration configuration is added to the atomic metric details page, supporting configuration of multiple groups of metrics' source datasets, source fields, and aggregation methods (only supports sum and count). It also supports priority sorting. When querying metric data, it will prioritize querying from datasets with higher priority.

    In principle, metrics are created on detail tables with sufficiently comprehensive applicable dimensions. The associated query acceleration tables are summary tables/dimension reduction tables of this table.

    Example: Atomic metric sum(sales revenue), the source dataset when defining the metric is Table 1, with dimensions of Province, City, District, Store. Associated tables are Table 2 and Table 3. Priority is set as Table 3 > Table 2 > Table 1.

    Table 1

    ProvinceCityDistrictStoreSales Revenue
    AAAa1
    AAAb2
    BBBc3
    CCCd4
    CDDf5

    Table 2

    ProvinceCityDistrictSales Revenue
    AAA1+2
    BBB3
    CCC4
    CDD5

    Table 3

    ProvinceSales Revenue
    A1+2
    B3
    C4+5

    • When dimension and filter fields all exist in Table 3, query from Table 3;

    • If applicable dimensions don't exist in Table 3, check if they exist in Table 2, query from Table 2;

    • If applicable dimensions don't exist in both Table 3 and Table 2, then directly query from Table 1.

    4 Examples

    （1）Inventory Market Value

    Calculation Logic: Within the statistical date, each store's inventory market value = product tag price * inventory quantity

    Table Structure: Date, Store, Category, SKC, Price, Inventory Quantity

    • Calculation Definition:
    sum(price * inventory_quantity)
    • Applicable Dimensions: Store, Category, Date

    （2）In-Hand Style Count

    Calculation Logic: Count the number of product styles with inventory in the store

    Table Structure: Date, Store, SKC, In-Hand Inventory Quantity

    • Calculation Definition:
    count(distinct concat(`store`, '_', `SKC`)
    • Applicable Dimensions: Date, Store
    • Filter: In-Hand Inventory Quantity > 0

    （3）R&D Person-Day Statistics

    Calculation Logic: R&D person-days = total work hours / standard work hours. Before 2025, the standard work hours per day is 7 hours, after 2025, the standard work hours per day is 8 hours. Calculate the average person-days within the statistical date

    Table Structure: Date, Total Work Hours

    • Calculation Definition:
    SUM(CASE
     WHEN ([date] < '2025-01-01') THEN ([total_work_hours])/7
     ELSE [total_work_hours]/8
    END
     )
    • Applicable Dimensions: Date