MicroStrategy ONE

Auto Narratives

Starting in MicroStrategy ONE (September 2024), auto narratives is supported in MicroStrategy Cloud environments. You can use auto narratives in MicroStrategy Library, Library for iOS, Library for Android, and Workstation when you connect to a Library server.

Auto narratives simplifies data interpretation by transforming complex visualizations into easy-to-understand natural language summaries. This enhances data accessibility, reduces the time needed to identify key insights, and makes data comprehensible to users of all technical backgrounds. Users can generate customizable summaries for entire dashboard pages or specific visualizations, ensuring relevance by respecting any filters applied. Narratives are updated when underlying data changes, maintaining accuracy and context. This supports faster decision-making and caters to various user needs, from technical experts to non-technical stakeholders.

Prerequisites

Auto narratives is part of the MicroStrategy AI add-on bundle and is available for MicroStrategy Cloud Environment (MCE) customers starting in MicroStrategy ONE (September 2024). Only users and user groups with the Use Auto Assistant and ML Visualizations privileges can access auto narratives.

Create Auto Narratives

Dashboard authors can create auto narratives.

  1. Open a dashboard in edit mode.

  2. In the top right of an existing visualization, click Moreand choose Create Auto Narratives. By editing an existing visualization instead of creating a new one, auto narratives automatically generates a general summary of an existing visualization. This summary can be modified later to reflect an output that suits your business needs.

    If necessary, you can create a new auto narratives dashboard by clicking Visualization > Insight+ > Auto Narratives.

  3. On the Editor panel, enter @ in Instruction to select the current page or specific visualizations you want auto narratives to summarize. Describe your request in detail to identify key insights and display them in a formatted summary.

  4. Click Generate.

  5. Use the Format panel to modify the text.

Consume Auto Narratives

Business users can view auto narratives directly within dashboards, where they are presented as concise, natural language summaries of key insights. Once these narratives are created by dashboard authors, these summaries are tailored to provide you with the most relevant information based on the underlying data and filters applied. You'll be able to see these insights dynamically adjust as you interact with the dashboard, ensuring you always have access to the latest and most pertinent data.

  1. Click Refresh to update the summary.

  2. To copy the text with its formatting, hover over the top right and click Copy .

For every auto narrative, there is a corresponding cache containing summary data. The cache is invalidated when the dataset is refreshed or the document definition is modified.

Best Practices

Source Data

Auto narratives can only generate summaries based on the page it is currently located on or the visualizations within that same page. To summarize visualizations within a panel stack, ensure that the auto narratives is added to the same panel as the target visualization. It is not yet possible to use auto narratives to summarize content from other pages or visualizations located outside the current page.

If visualizations are controlled by attribute or metric selectors, any changes to the attributes or metrics of the target visualization may affect the summary's accuracy. In such cases, click Generate to regenerate the summary for the correct result.

Instructions

When setting up auto narratives, provide clear and specific instructions to get the best summary output. If you don’t provide specific instructions, auto narratives generally displays highlights, lowlights, key drivers, or trends if applicable. If the summary does not meet your expectations, modify the instructions and click Generate, as the structure of the outcomes can vary based on the data and visualizations used.

Formatting

Provide detailed instructions for custom formatting. Auto narratives supports rich HTML formatting, allowing you to customize font color, size, styles, images, bullets, titles, and paragraphs.

After generating and saving an auto narrative in the dashboard, it’s best to avoid drilling down on the data in the visualizations targeting auto narratives. Drilling could remove the attributes or metrics referenced in the instruction, potentially leading to incorrect results.

Managing Footnotes

A footnote for a single source visualization may appear in multiple paragraphs within an auto narratives visualization. If you prefer to not display the footnote, simply include instructions such as "don’t show the footnote."

Accuracy in Filtering Instructions

When providing instructions for filtering data, such as “display the revenue for region 'Northeast’,” ensure that 'Northeast' matches the exact capitalization and spelling as it appears in your dataset. This helps prevent errors and ensures the query retrieves the intended data accurately.

For more detailed information on tracking and analyzing the usage of auto narratives, see Auto Narratives Insights.