Get started with Quick BI ChatBI

Step 1: Initialize your ChatBI account

Before you can use ChatBI for analysis, a Quick BI organization administrator must handle permission management. This includes creating ChatBI-related organization roles and assigning them to you.

1. The organization administrator logs in to the Quick BI console.

2. In the Quick BI console, the organization administrator creates the required ChatBI roles, as shown below.

   1. Navigate to System Settings > User Management > Role Management > Function Permissions to access the role creation page.

   2. Click Create Role, name the new ChatBI-related organization role, and save it.

   3. Select the newly created organization role and configure its ChatBI permissions.

   4. The ChatBI-related organization roles are as follows:

      Organization role	Created by	ChatBI-related permissions
      organization administrator	System-defined	Adds users to the Quick BI organization, creates ChatBI-related organization roles, and assigns users to these roles.
      Standard user	System-defined	Can use the ChatBI feature on datasets for which they have permissions.
      Q&A configurator	Custom role created by an organization administrator	Configures ChatBI for owned datasets. By default, can use ChatBI on datasets for which they have permissions.
      Q&A administrator	Custom role created by an organization administrator	Manages all ChatBI resources and authorizations, including resource management and data permission management. By default, has permissions for dataset configuration and analysis.

3. After creating the roles, the organization administrator assigns the appropriate ChatBI role to you.

   1. If you are a new Quick BI user, the organization administrator assigns the role when adding you to the organization under **User Management** > **Member Management** > **Add User**.

2. If you are an existing Quick BI user, the organization administrator adds the ChatBI role to your account under **User Management** > **Role Management** > **Role User**.

Step 2: Set data permissions for ChatBI

A workspace developer must first connect to a data source. Then, a user with both workspace developer and Q&A configurator roles needs to create a dataset, enable data preparation for it, and grant you permissions to query it.

1. The workspace developer for the ChatBI workspace logs in to the Quick BI console.

2. In the workspace where the ChatBI dataset will be created, the developer creates a data source.

   1. Method 1: If you already have analysis data, create a data source in Quick BI. For more information, see Create Database Data Source Overview or Create File Data Source.

   2. Method 2: If you do not have analysis data, you can use the demo data by uploading DEMO_Sales_Revenue_and_Profit.xlsx to the Exploration Space. The workspace developer navigates to the **Data Sources** page, selects the Exploration Space, and clicks **Upload File** to import the Excel file. Before clicking **Confirm and Upload**, verify that the field types are correct, especially for date and ID fields.

      1. Click **Data Sources** > **Exploration Space** > **Upload File**.

         

      2. Upload the file.

         

      3. Click **Confirm and Upload**.

         

3. The user with both workspace developer and Q&A configurator roles creates and saves a dataset, then enables its ChatBI configuration.

   1. From the **Data Sources** or **Datasets** page, the user selects the created data source to create, name, and save a new dataset. The following example starts from the **Data Sources** page:

      1. Click **Create Dataset**.

         

      2. On the dataset editing page, click Done, Start Data Processing.

         

      3. Click Save, enter a Dataset Name, select a Location, and click OK.

         

   2. On the dataset editing page, go to advanced settings > Intelligent Q > ChatBI to enable the feature.

      

   3. The ChatBI configuration involves three steps: entering basic information, assessing field quality, and setting quick questions.

      Step 1: Enter basic information		Fill in basic information Display Name: The dataset name is used by default but can be customized. This name appears to users in the ChatBI interface, for example, "Regional Category Sales Data". Description: Add a brief description to make the dataset easier to find. Dataset Type: The default is Detail Table. You can also choose other types like multi-metric periodic tables or key-value tables. The dataset type helps Intelligent Q understand the data structure and improves query accuracy. For this example's detail data, select Detail Table. Click Start Learning. Click Next to continue to the field quality assessment.
      Step 2: Assess field quality		Field quality assessment This step is optional. If you wish to skip this assessment or not apply the suggestions, click Next. Field Quality: The system automatically assesses the quality of the dataset fields and provides intelligent suggestions for improvements to enhance query performance. You can choose to accept the recommendations and then click Apply and Relearn. Assessment Time: The field quality assessment may take 1 to 2 minutes. You can proceed to the next steps, and a notification will appear when the assessment is complete. Click Next to proceed to the quick questions page.
      Step 3: Set quick questions		Quick Question: Display Location: To help users get started, quick questions are displayed in the data preview window of the ChatBI interface. They can also be triggered by typing a forward slash (/) in the input box. System Recommended Mode: This is the default mode, where the system automatically generates a list of suggested questions for the dataset. Expert Custom Mode: This mode allows you to define a custom list of questions for the dataset. Click Confirm and Enable to complete the ChatBI configuration for the dataset.

4. The user with both workspace developer and Q&A configurator roles grants you ChatBI permissions for the dataset.

   1. On the dataset editing page, navigate to advanced settings > Intelligent Q > ChatBI Permissions and grant query permission for this dataset, setting an expiration date if necessary.

      

   2. Different users have different needs, as shown in the table below:

      User permission requirement	Required roles	Dataset configuration
      Business operations staff only need to perform ChatBI analysis on a dataset.	(Standard user, Workspace analyst)	Enable ChatBI for the dataset and grant ChatBI permissions to the business operations staff.
      Data analysts need to configure and analyze datasets they own.	(Q&A configurator, workspace developer)	Enable ChatBI configuration for the dataset and grant ChatBI permissions to themselves.
      Data analysts need to configure and analyze datasets they do not own.	(organization administrator, workspace developer)	Enable ChatBI configuration for the dataset and grant ChatBI permissions to themselves.
      Data administrators need to manage ChatBI resources and permissions for datasets.	(Q&A administrator, Workspace administrator)	Enable ChatBI configuration for the dataset.

Step 3: Analyze business data with ChatBI

After an organization administrator assigns you a ChatBI role and a developer grants you dataset permissions, you can start a new conversation in the ChatBI interface to perform Q&A analysis.

1. Log in to the Quick BI console.

2. In the Quick BI console, navigate to **Intelligent Q** > **ChatBI**, select the dataset you want to analyze, and start a new conversation.

   

3. In the conversation, analyze the net profit for the China division using the "Regional Category Sales Data" dataset.

   Ad-hoc questions	Instant answers
   Examine the growth performance of the China division Ask: net profit of China division in December 2024 Follow-up: month-over-month growth Follow-up: year-over-year growth Business Conclusion In December 2024, the net profit for the China division was 192.2 million, a decrease of 45.62% month-over-month and 2.58% year-over-year. Next Analytical Step Compare this performance to the entire company.
   Compare against the company's overall growth Ask: company net profit in December 2024 Follow-up: month-over-month growth Follow-up: year-over-year growth Business Conclusion In December 2024, the company's total net profit was 1.413 billion, a decrease of 5.11% year-over-year but an increase of 23.01% month-over-month. In contrast, the China division's profit decreased by 45.62% month-over-month. Next Analytical Step Determine if the decline in the China division is an anomaly or part of a normal cyclical trend by examining the division's data trends.
   Analyze the growth trend of the China division Ask: monthly net profit trend for China division from 2023 to 2024 Business Conclusion The trend chart and data interpretation reveal a cyclical pattern in the China division's net profit. The profit figures for December 2024 are comparable to December 2023, with both months showing a significant month-over-month decline. The trend forecast suggests that net profit will recover in early 2025. Therefore, the decline in December 2024 is consistent with a cyclical fluctuation. Next Analytical Step Analyze the reasons for this cyclical fluctuation. Click the fluctuation analysis feature on the trend chart to discover that the cycle can be attributed to specific product sub-categories.
   Attribute the net profit decline in the China division Ask: net profit share of each product sub-category in China division in December 2024 Ask: month-over-month net profit ranking of product sub-categories in China division in December 2024 Ask: details of Brand A Enjoy series in China division in December 2024 Business Conclusion The analysis of product sub-categories shows that the net profit decline in the China division in December 2024 was mainly caused by the Brand A Enjoy series. This series accounted for 59.05% of the net profit and decreased by 49.02% month-over-month, corresponding to a drop of 110 million in net profit.

4. ChatBI feature highlights.

   Feature	Description	Icon
   Voice input	Supports using voice input to ask questions.
   Natural language parsing	When providing an answer, ChatBI can show or hide the analysis process, including the dataset, relevant fields, knowledge base entries, and user intent.
   SQL self-service query	When providing an answer, you can view the SQL pseudocode used by ChatBI to understand the data processing logic.
   Chart-based results	When providing an answer, ChatBI displays dimension and measure filters, selects appropriate charts to visualize the results, and allows you to customize the chart type.
   Follow-up questions	After receiving an answer, you can ask follow-up questions. ChatBI automatically applies the context from the previous query, including dimensions and measures, to parse the new query, retrieve data, and visualize the result.
   Trend analysis	When answering, ChatBI can recognize a user's request to see a trend and displays the data in a trend chart by default.
   Trend forecasting	After a trend analysis, you can request a forecast. ChatBI uses existing data and statistical models to predict short-term data trends with high accuracy.
   Fluctuation analysis	After a trend analysis, you can perform fluctuation analysis. You can configure the comparison period and business dimensions to analyze the factors contributing to changes in net sales profit.
   Data interpretation	When providing an answer, you can choose from different large language models for data interpretation. ChatBI can interpret the results and provide data findings, conclusions, and suggestions.
   Proportion analysis	When answering, ChatBI can recognize a request to see a proportional distribution and displays the data in a pie chart by default.
   Ranking analysis	When answering, ChatBI can recognize a request to see a ranking and displays the data in a leaderboard chart by default.
   View data details	When answering, ChatBI can recognize a request to see detailed data and displays it in a detail table by default.

Step 4: (Optional) Tune ChatBI performance

If you receive inaccurate results, this may not be a product issue. You can often improve accuracy by tuning.

1. Tuning scenarios, benefits, and methods:

   1. Common inaccuracies include deviations in intent parsing, errors in intent parsing, and incorrect metric retrieval logic.

   2. Tuning can improve semantic parsing accuracy, enhance complex query handling, and reduce response times.

   3. Tuning methods include question tuning, dataset tuning, and knowledge base tuning.

2. Examples of tuning methods: