Skip to main content
Version: 8.3.0

Smart ETL Advanced Calculation

Data Exploration

Feature Description: During Smart ETL development, users often need to inspect dataset statistics such as row count, the number of distinct values in a field, and the number of Null values. Traditionally, this often required manually combining operators such as Filter Data Rows and Group Aggregate. To simplify the workflow, Guandata recently added the Data Exploration operation to ETL so that dataset statistics can be generated at once and data processing efficiency can be improved.

How to use it:

  1. Open the ETL editor. In the left-side operation panel, locate Data Exploration under the Advanced Calculation category, drag it into the editor, connect the relevant data processing operators to Data Exploration, and then click it to configure the node.

  2. Select the dataset statistics you need. At the table level, you can count dataset rows and columns. At the field level, first select the target field, then compute its Null count, distinct value count, and enumerated values (up to 200 enumerated values can be displayed). Click Query at the bottom of the editing area to view the dataset statistics.

SQL Input

The SQL Input operator helps data analysts who are accustomed to SQL use SQL directly to output datasets, shortening the configuration process.

To make details easier for ETL visitors to review, field expression content in the SQL Input and Add Calculated Column operators can be displayed externally.

Why do some field types change in the SQL node preview?

Can the SQL operator use custom dynamic time macros?

Feature Description: Uses SQL to query, extract, and combine data.

Configuration Instructions: Drag the SQL Input operator onto the canvas, select the node, and then click Edit. You can introduce functions according to the examples, fill in field names, and click OK.

Correlation Mining

The Correlation Mining operator helps data analysts quickly perform association mining algorithms and identify highly related items among different items in the data. This is a paid module and is not enabled by default. If you want to try or purchase it, contact Guandata business staff.

Smart Benchmarking

The Smart Benchmarking operator allows users to specify category entities, related category features, and the number of neighboring entities, and then automatically obtain comparison items for each entity, such as comparison stores, in list form. This is a paid module and is not enabled by default. If you want to try or purchase it, contact Guandata business staff.

For other data processing operators used later, see Getting Started.

Intelligent Attribution

Overview

Attribution for key metrics can combine dimension analysis, intersecting dimensions, drill-down, metric decomposition, target-gap analysis, and significance analysis of related metrics. The Smart ETL attribution operator performs the attribution calculations. Other Smart ETL operators can then process and integrate the results into application data for front-end data products.

The operator follows three principles:

  • Atomic analysis: Compare two sets of data. Use other Smart ETL operators to prepare and filter the source data for year-over-year, period-over-period, target, or peer comparisons.
  • Standardized results: Use standard dimension-based and metric-based attribution. Users follow the attribution chain and obtain standardized outputs without needing to manage the calculation internals.
  • Batch processing: Analyze multiple periods and levels of granularity together and materialize the results as a dataset, rather than running only ad hoc front-end attribution.

Configuration

Drag the attribution operator onto the canvas, select it, and configure it in the left-side panel.

ConfigurationDescription
Comparison field and period conditionsSelect the comparison field and specify the conditions identifying the current and baseline periods.
Metric typeAggregate metrics use additive values with sum or count. Ratio metrics use calculated fields in the form sum(A) / sum(B).
Metric formulaDefine the main metric formula using functions and fields.
Iteration dimensionsSelect one or more dimensions over which to iterate.
Attribution dimensionsChoose single-dimension analysis, combinations of dimensions, or hierarchical drill-down. Attribution dimensions must not include iteration dimensions. The results show the impact of the selected dimensions or combinations.

Preview Results

Select the attribution result dataset and preview it to inspect the results.