Fluctuation analysis is a core feature of the intelligent module. This feature uses automated attribution algorithms to support multi-dimensional and multi-step root cause analysis of metric changes within a single dataset. It generates visual key driver analysis reports to help you quickly identify the core drivers of business fluctuations. This topic describes how to configure and use the fluctuation analysis feature.
Key driver analysis is a value-added module that requires a separate purchase.
This feature is currently available only in the China (Hong Kong) and Malaysia sites.
Core capabilities
Supports single-step and multi-step key driver analysis.
Provides multiple attribution methods, such as dimension attribution and metric attribution.
Scenarios
Business fluctuation analysis
Example: "Why did yesterday's sales drop by 30%?"
Cross-period comparative attribution
Example: Analyze the drivers of change for a metric by comparing the "current period" with the "comparison period".
Configuration entry point
On the dataset edit page, click Advanced Configuration and then select Fluctuation Analysis to open the configuration page.

In the fluctuation analysis configuration interface, you can configure the analysis logic and report template for your metrics.
Analysis logic configuration
Global settings
If no key driver analysis is configured for a dataset, the system performs attribution using single-dimension analysis by default. In this section, you can set the priority attribution dimensions.
In the global settings, select dimension fields from the secondary attribution dimensions, click the
icon, and then click Save to set them as priority attribution dimensions.

To remove a priority attribution dimension, select the dimension and click the
icon.

Configuring analysis logic for attributable metrics
Select a field, click New, and choose an attribution method. Single-step attribution and multi-step attribution are supported.
Only attributable fields are displayed in the calculated fields list. The available aggregation methods include sum, count, average, and distinct count.

If analysis logic is already configured for the metric, you can also click the New button on the right to create a new analysis logic.

Single-step attribution
You can choose standard attribution methods, such as dimension attribution and metric attribution.

The configuration items are as follows:
In the Attribution Metric section, you can configure the aggregation method, metric unit, and filter conditions.
Set Aggregation Method to Sum, Average, Count, or Distinct Count.
Customize the Metric Unit, for example, "Ten Thousand".
Set the Filter Condition for the metric.

In the Attribution Date section, you can select a Date Field, a filter method, and the corresponding attribution and comparison dates.
The supported filter methods are Single Date Comparison and Date Range Comparison:
If Filter Method is set to Single Date Comparison, set the Attribution Date and Comparison Date.

If Filter Method is set to Date Range Comparison, set the Attribution Period and Comparison Period.

In the Attribution Configuration section, you can configure the attribution type and attribution method.
Set Attribution Type to Dimension Attribution or Metric Attribution.
Dimension Attribution shows the contribution of each dimension value to the change in the attribution metric. When Attribution Type is set to Dimension Attribution, set Attribution Method to Single-Dimension Analysis, Dimension Intersection, or Dimension Layering.
Single-Dimension Analysis
To analyze the impact of a single dimension (such as region) on the results, you can select multiple values for that dimension.
You can set the Priority Attribution Dimensions and Secondary Attribution Dimensions.

Dimension Intersection
Analyzes the impact of a combination of two to three dimensions, such as "Region + Channel". You can select up to three dimensions.
You can select the Attribution Dimensions.
NoteYou can add up to three dimensions.
Dimension Layering
Breaks down dimensions by level to find the cause, such as "Total Performance → Department → Employee". Supports multiple levels.
You can set the Dimension Layering.
NoteYou can add up to three dimensions.
Metric Attribution shows the contribution of related metrics to the change in the attribution metric. When Attribution Type is set to Metric Attribution, set Attribution Method to Metric Factor or Metric Correlation.
Metric Factor
Calculates the impact of multiple metrics that use basic operations, such as "Number of Users × Customer Spend".
You can edit the Metric Factor Relationship.

Metric Correlation
The correlation weight for metrics that are not related by basic operations (such as the correlation between Customer Satisfaction and GMV).
You can add Related Metrics and set their aggregation methods.
NoteYou can add up to six attribution metrics.
Multi-step attribution
You can organize and chain different types of attribution nodes. These nodes include Attribution Metric, dimension attribution (Single-Dimension Analysis, Dimension Intersection), and metric attribution (Metric Correlation Attribution, Metric Factor Attribution).

Attribution Metric
You can configure the aggregation method, metric unit, date field, filter method, and the corresponding attribution and comparison dates. The configuration is the same as that for single-step attribution.

Single-Dimension Analysis
Analyzes the impact of a single dimension on the result, such as region. You can select multiple dimensions at a time.

Dimension Intersection
Analyzes the impact of a combination of two to three dimensions, such as "Region + Channel". You can select up to three dimensions.

Metric Correlation Attribution
Correlation weight for metrics not related by basic operations (such as the correlation between customer satisfaction and GMV).

Metric Factor Attribution
Calculates the impact of multiple metrics that use basic operations, such as "Number of Users × Customer Spend".

Managing analysis logic
Edit
Click the
icon in the Actions column to edit the analysis logic.

Set as default
Click the
icon in the Actions column and select Set as Default. This sets the current analysis logic as the default, which will be automatically selected for future key driver analyses.

Delete
Click the
icon in the Actions column and select Delete.

In the confirmation dialog box that appears, click OK to delete the current analysis logic.

Report template configuration
After you configure the analysis logic, you can configure a report template for it.
Create a report template
On the report template configuration page, click New.

You can also click Create Report Template directly while configuring the analysis logic.

Select analysis logic
You can select an existing analysis logic.

Generate an analysis report
The system automatically generates an analysis report.

Configure report content
You can configure the attribution report content.

You can show or hide sections, such as the Title Area, Metric Overview, Attribution Summary, Key Influencing Factors (under Attribution Conclusion), and Detailed Analysis.
Click Update Data to view the latest data.

Click Save to create an attribution report template.

You can switch to view the mobile-friendly version.

Manage report templates
Edit
Click the
icon in the Actions column to edit the key driver analysis report template.

Delete
Click the
icon in the Actions column and then click Confirm to delete the attribution report template.

Use Fluctuation Attribution
In Quick BI Q&A, you can use the fluctuation analysis feature to gain insights into unexpected changes by performing an in-depth analysis of various metrics.
Using fluctuation attribution in Quick Query
Follow these steps:
On the Quick BI Q&A page, click the
icon to open the Quick Question interface. Then, click Smart Insights and click Fluctuation Analysis.
Enter the current date and comparison date, select an attribution metric, and send the query.

The system returns a key driver analysis report.

The Quick BI analysis procedure is displayed on the right side of the analysis report. Click the
icon in the upper-right corner to exit the analysis procedure.
After you exit, you can expand the analysis procedure again.

Click the
icon in the upper-right corner of the key driver analysis report to exit the preview interface.
After you exit, you can click "Preview" in the conversation interface to re-enter the report preview interface.

You can configure attribution report settings.
NoteFor more information, see Analysis logic configuration.
You can modify the Attribution Metric, set the Data Interpretation logic, and perform Model Selection as needed.

Click More Options to configure the attribution metric, attribution date, attribution configuration, and data interpretation.

After the configuration is complete, click OK. The system regenerates a key driver analysis report based on the modified configuration.
Return key driver analysis based on semantics in Q&A
Follow these steps:
In the Quick BI Q&A interface, you can enter a question, such as "Help me analyze the reasons for the fluctuation in the order amount". The system then returns the analysis results.

Click the
icon in the top-right corner to configure the Attribution Report Settings.NoteFor more information, see Analysis logic configuration.

View fluctuation analysis in charts from Q&A results
Follow these steps:
In the Quick BI Q&A interface, you can enter a question, such as "What is the monthly sales trend from January to December 2023?" The system then returns the analysis results.

In the returned chart, click the
icon and select Fluctuation Analysis.
The system automatically generates a key driver analysis report.
After the report is generated, you can view it.

To configure Attribution Report Settings, click the
icon in the top-right corner.NoteFor more information, see Analysis logic configuration.
You can modify the Attribution Metric, set the Data Interpretation logic, and perform Model Selection as needed.

Click More Options to configure the attribution metric, attribution date, attribution configuration, and data interpretation.

After the configuration is complete, click OK. The system regenerates a key driver analysis report based on the modified configuration.