Quick BI offers the Q Chat feature, which combines the intelligent data assistant Q with the interactive capabilities of ChatBI. This feature enables you to retrieve data results through natural language interaction, providing instant answers to your data-related questions. It makes data analytics accessible to all users and introduces a new way to consume data. On a PC or mobile device, you can use the Q Chat dialog interface to preview and select datasets. You can also enter questions directly or use quick questions for multi-turn conversations. In the conversation list, you can view your historical Q Chat conversations.
This topic describes how to use Q Chat.
The Q Chat feature is a value-added module that requires an additional purchase.
This feature is currently available only on the China (Hong Kong) and Malaysia sites. Support for other sites will be available soon.
Prerequisites
You have created a dataset and completed the Q&A configuration.
You have obtained permissions for Q&A resources. For more information, see Permission Management.
PC entry point
On the Quick BI home page, click Intelligent Q to open the Q Chat dialog page.
Data sources
The data sources are datasets for which you have Q&A permission, located under All and within each analysis subject.
For information about how to configure permissions for Q&A datasets, see Q&A permission management.
For information about how to configure analysis subjects, see Analysis Subject Management.
Datasets marked with the Sample flag in the upper-left corner are sample datasets.
If you do not need the sample data, you can hide it. For more information, see Show or hide sample datasets.
Preview and select datasets
Select a single dataset
In the Q Chat dialog page, you can view datasets under All and within each analysis subject. You can hover over a target dataset to preview it or directly ask a question based on it.
Click Preview (①) to view the field details and data preview of the dataset. You can also use quick questions or click Ask in the upper-right corner to open the Q Chat dialog box.
Field details
Data preview
You can hover over a dimension field to view its details.
Quick questions
Ask
You can ask data-related questions based on the current dataset.
Click Ask (②) to open the data Q&A dialog interface and ask questions based on the current dataset.
After selecting a dataset, you can click the
icon to its right to preview data fields and view key indicators and analysis dimensions.
You can expand or collapse the data fields.
During the Q&A process, you can preview data and switch datasets.
Preview data
Click the
icon to the right of the selected dataset to preview the data.
Switch datasets
Click the name of the selected dataset to view available datasets.
Hover over the target dataset and click Ask to switch datasets.
Alternatively, you can first preview the target dataset, and then ask a question.
Select multiple datasets
If your question applies to multiple datasets, you can select all relevant datasets for your query.
This section describes only the procedure for selecting multiple datasets. Other operations are the same as those described in Select a single dataset.
Enter a question.
The system displays datasets that might be relevant to your question. You can select the datasets you need.
Select the checkboxes for multiple datasets and click Ask a Question Based on Selected Data.
View the analysis results.
Ask questions and get answers
Q Chat supports multiple inference and data interpretation methods. This example demonstrates how to use the default inference and data interpretation methods. For information about the analysis results of other methods, see Inference and interpretation methods.
Enter a question in the dialog box and send it. You can enter a question directly, use a quick question, or use speech input.
Enter a question directly
For example, enter 'What is the proportion of order quantities for each order level?' After you send the question, the system returns the data analysis results and displays the analysis process on the right.
If you do not need to view the analysis process, you can click Analysis process expanded to hide it.
After the process is hidden, the result shows only the chart.
Then, you can click Analysis complete to show the analysis process again.
When you enter content, the system may suggest recommended fields and recommended questions. For example, if you enter "2024", the following recommended fields and questions appear. You can select them as needed.
Favorite questions
NoteYou can add up to 30 questions to your favorites.
Click the
icon to the left of a question to add it to your favorites.
After you add a question to your favorites, it is available under Quick Questions.
If no dataset is selected, you can click a favorite question to ask it directly.
Click the
icon again to remove the question from your favorites.
Copy question
Click the
icon to the left of a question to copy the question.
Quick questions
Click the
icon in the lower-left corner of the question box to open the Quick Questions interface. Then, click a question from the Favorites, Recommended, or Recently Used tabs to ask it.
For questions in Favorites, you can select the Automatically switch dataset checkbox. This option is selected by default. If this checkbox is selected, clicking a favorite question switches to the attached dataset and asks the question. If this checkbox is not selected, the question is asked based on the current dataset.
Alternatively, you can click a recommended question above the question box to ask it.
Click the
icon to refresh the recommended questions.
After each round of conversation, you can also select one of the three recommended questions below the chart to ask another question.
Speech input
NoteAfter you enable the voice Q&A switch configuration, you can ask questions using speech input.
Click the Speech input icon to open the speech dialog interface.
In the speech dialog interface, ask your question.
The system recognizes your speech and converts it to text.
Click the
icon on the left to cancel the input, or click the 
icon on the right to switch to text input.Click Send or press Enter. The system returns the data analysis results.
While viewing data, the following features are available.
Data screening (①): The chart displays the query conditions for the current data retrieval. You can switch the conditions.
Chart switching (②): You can switch the visualization type to one that better suits your viewing habits. The available chart types vary based on the data.
View AI data retrieval process (③): You can verify that the data retrieval results meet your requirements. You can view the business logic SQL and the executed SQL.
Fluctuation attribution and trend prediction (④)
Fluctuation attribution
NoteMetric cards and line charts support fluctuation attribution. The fluctuation attribution feature is available only if the following conditions are met:
Supported chart types: Metric card and standard line charts (not prediction-based line charts).
The chart contains only one dimension.
The dimension is a date type and has no legend.
The chart that you want to analyze for fluctuation attribution must not have any errors.
Line chart conditions: One date dimension and one to three measures.
Metric card conditions: One date dimension and one measure.
Click the
icon in the upper-right corner to configure attribution report settings.
You can modify the attribution measure, data interpretation logic, and model selection. After you click OK, the system automatically updates the attribution report. For more information about related operations, see Key Driver Analysis.
Trend prediction
NoteLine charts and bar charts support trend prediction. The trend prediction feature is available only if the following conditions are met:
Supported chart types: Line charts and bar charts that contain a date and have no legend.
The chart is not a prediction result itself.
The chart that you want to use for prediction must not have any errors.
The chart that you want to use for prediction must have 12 or more dimension values.
Line chart conditions: One date dimension and one to three measures.
Click the
icon in the upper-right corner to share the trend prediction.
Full-screen view (⑤)
Rename (⑥)
You can rename the result. You can also click the area that is outlined in red to rename it directly.
Share data (⑦)
You can copy the link to share.
Export data (⑧)
You can export the data to a local Excel file with a custom export name.
NoteThe export feature is available on PCs only, not on mobile devices.
Data interpretation (⑨)
NoteHere, you can initiate a custom data interpretation. This lets you interpret the Q&A results based on your specified interpretation logic and model.
You can refer to the example interpretation logic, enter your own custom logic, select a model, and click Generate Interpretation. The system then automatically interprets the data for you.
You can stop the interpretation while the result is being generated.
You can adopt or modify the interpretation result. After you adopt the result, a data interpretation report is generated, as shown in the following figure.
If you want to modify the interpretation logic or model for another interpretation, you can click the
icon to the right of the data interpretation to make changes and generate a new report.
Append questions (⑩)
Rating (⑪)
Click
to like a result, or 
to dislike it. If you dislike a result, you can provide feedback.
Copy Trace_ID (⑫)
Inference and interpretation methods
Method description
By default, the inference and interpretation methods use built-in system models. You can select the system's built-in large models or your custom models. You can select the appropriate inference and interpretation methods to analyze your data based on your business needs.
Examples
The following examples use the question "What is the proportion of order quantities for each order level?" to demonstrate different inference and interpretation methods:
Example 1: Select the default inference method and do not perform data interpretation.
The system returns the analysis process and a chart.
Example 2: Select the built-in Qwen-Max model for both inference and data interpretation.
The system uses the Qwen-Max model to return the inference process, data interpretation results, and a chart.
Example 3: Select the built-in DeepSeek-R1 model for both inference and data interpretation.
The system uses the DeepSeek-R1 model to return the inference process, data interpretation results, and a chart.
Append questions
You can use the Append Question feature to ask follow-up questions about the previous result. This helps you understand the data from a more precise perspective. The following example shows how to perform this operation.
You can append only one follow-up question to a question.
In the dialog interface, you can click Append Question to ask a follow-up question to the previous one, such as "Which province has the highest sales?"
Enter a follow-up question as needed, such as "What is the product with the highest sales?"
Click the
icon in the lower-right corner of the dialog box or press Enter to send the follow-up question and view the analysis result.
NoteThe system then automatically exits the append question mode. To ask another follow-up question, you must click the Append Question button again.
Conversation list
Click the
icon in the upper-left corner to view your historical Q&A conversations.
Click the
icon to open a new Q Chat interface.
Display on mobile devices
Mobile micro-application configuration entry
Follow the steps in the figure to open the mobile editing interface.
In the editing interface, you can configure Q Chat for the mobile micro-application.
Suggestions for asking questions
Add a time frame to your question. For example:
Instead of asking 'Sales amount proportion for each product', ask 'Sales amount proportion for each product this year'.
When summarizing values across multiple dimensions, specify whether you want individual or total results to avoid ambiguity. You can use keywords such as 'individually', 'for each', 'total', or 'grand total'. For example:
The questions 'Sales in Zhejiang and Jiangsu' and 'Sales from 2020 to 2023' are ambiguous. You can rephrase them for clarity. For example: 'Sales in Zhejiang and Jiangsu individually', 'Sales for each year from 2020 to 2023', 'Sales for every year from 2020 to 2023', 'Total sum of sales in Zhejiang and Jiangsu', or 'Total sales of Zhejiang and Jiangsu'.
When asking for maximum or minimum values across multiple dimensions, be specific to avoid ambiguity between grouped and ungrouped results. You can rephrase the question to ask for the extreme value of a single dimension or use a window function. For example:
The question 'Which product sells best in Zhejiang and Jiangsu?' is ambiguous. You can rephrase it as 'Product with the highest sales in Zhejiang', 'Product with the highest sales in Jiangsu', or 'Product with the highest sales in each province'.
The question 'The 3 best-selling months last year in Zhejiang and Jiangsu' is ambiguous. You can rephrase it as 'The 3 best-selling months last year in Zhejiang', 'The 3 best-selling months last year in Jiangsu', or 'The top 3 order months by sales for each province'.
If a combination of dimension values is ambiguous, specify the dimension clearly in your question. For example:
The question 'Products purchased by corporate customers, ranked by sales' can be ambiguous. The dimension 'Customer Type' includes the value 'Corporate'. To avoid confusion, you can specify the dimension and value clearly. For example: 'For customer type Corporate, rank purchased products by sales'.