May 2023 Summaries

In 1801, William Playfair introduced the pie chart as a means to depict the territorial proportions of empires, a method that has since become one of the most debated visual tools in data representation. Pie charts are favored for their simplicity and visual appeal, especially when representing a small number of categories, but they are criticized for their inaccuracy and potential to mislead due to human difficulty in accurately comparing angles and areas. Alternatives such as bar charts, stacked bar charts, treemaps, sunburst charts, and radar charts offer various advantages, such as better accuracy in comparison and suitability for hierarchical or multivariable data, though they also have their own limitations. The blog discusses these alternatives in the context of Apache Superset, demonstrating how each can be implemented to enhance data visualization and urging users to explore these tools to diversify their analytical presentations.

The May 2023 edition of the Superset Community Newsletter, authored by Evan Rusackas, encourages readers to explore visualization tools by trying Preset, offering a free start option. It also invites readers to subscribe to blog updates to receive a weekly digest of new blog posts, aiming to keep the community engaged and informed about recent developments and insights.

Data drilling is a technique that allows users to explore detailed data behind summaries and understand underlying trends, and Preset has introduced a new feature called Drill By to enhance this capability. Drill By enables users to delve into charts and tables by any dataset dimension within Preset dashboards, facilitating more flexible and intuitive data analysis compared to the traditional Drill Down method, which relies on a predefined hierarchy. This feature is accessible to all Preset users, regardless of their role or subscription plan, and allows for dynamic exploration by selecting dimensions directly from a chart's contextual menu. Users can perform multiple Drill By operations, view data in table format, and even save drilled results as new charts for future reference. Supported by various Preset chart types, Drill By aims to simplify self-serve analytics, making it easier for users to uncover insights across product, business, and operational domains.

Data modeling is an essential process that structures raw data into a useful format, forming the foundation of successful business intelligence strategies. It includes three main types: conceptual, logical, and physical data models, each with distinct characteristics and suitable scenarios. Conceptual models offer a high-level overview focusing on relationships between entities, while logical models provide more detail using tools like Entity-Relationship diagrams without focusing on data storage specifics. Physical models delve into the technical specifics of data storage, including table structures and data types. Various data modeling infrastructures, such as hierarchical, relational, entity-relationship, and object-oriented models, cater to different organizational needs. Effective data modeling can improve data quality, enhance understanding and communication, facilitate data integration, and support business goals. However, it faces challenges like complexity, time consumption, and the need for expertise. Tools like Preset aid in data exploration and visualization, offering intuitive interfaces and collaboration features to enhance the data modeling process. As data modeling continues to evolve with trends in AI and machine learning, it remains pivotal in helping organizations unlock the potential of their data for strategic decision-making.

The article provides an in-depth exploration of using language models like GPT and Bard in conjunction with Apache Superset, focusing on practical applications, limitations, and the author's personal experiments. It emphasizes the promise and pitfalls of these tools, noting that while GPT can handle tasks such as generating SQL queries from natural language and explaining SQL code in plain English, it struggles with tasks requiring precise technical accuracy, such as troubleshooting query errors and converting SQL dialects. Despite the challenges, there's optimism about the potential benefits, like integrating GPT for feature ideas and automating documentation tasks. The author expresses a cautious enthusiasm for the future of AI in analytics, urging users to validate AI-generated outputs and participate in further discussions and testing to refine these technologies in real-world applications.

Bureau, a modern no-code decisioning platform, transitioned from using Metabase to Preset, a hosted solution for Apache Superset, to better meet its growing business intelligence needs. The switch was driven by the need for more precise forecasting and enhanced visualization to support internal stakeholders, like the customer success and sales teams, in performing their roles effectively. Bureau found Preset to be cost-effective, user-friendly, and versatile, thanks to its semantic layer, drag-and-drop chart builder, and affordable pricing model. This new tool enabled Bureau to democratize data access across the organization, with over 90% of the company utilizing Preset for insights and decision-making. Additionally, Bureau integrated Preset's dashboards into its core offerings, providing customers with real-time visibility into their usage data, which helped build trust and streamline invoice processes for the finance team.

Data democratization is a transformative approach that aims to break down traditional barriers to data access, allowing all employees to utilize data for improved decision-making, innovation, and efficiency. By decentralizing data control and promoting a data-driven culture, organizations can enhance decision-making, accelerate innovation, and improve operational efficiency. Effective implementation requires establishing data governance, fostering data literacy, and embracing self-service analytics, often employing architectures like data mesh or data fabric to balance centralized integration with distributed ownership. Challenges such as data quality, security, and cultural shifts must be managed with robust governance frameworks and continuous training. Real-world examples from financial institutions demonstrate the tangible benefits of data democratization, including cost savings and improved compliance. By integrating inclusive data cultures and aligning technology with business goals, companies can unlock hidden potential, gaining a competitive advantage in an increasingly analytics-driven market.

Apache Superset 2.1.0 introduces various new features, enhancements, and security improvements, addressing a significant security issue by requiring users to change the default SECRET_KEY to prevent production mode booting with default settings. The update includes new visualization capabilities, such as a generic non-temporal X-axis for ECharts and horizontal filter bars for dashboards, while also enabling users to drill down into chart details and export data in XLSX format. Performance improvements have been made, with more components moved to a React single page application and support for Python 3.11, which enhances speed. Database connectivity sees enhancements with added support for ClickHouse, Risingwave, and improved authentication options for Snowflake and Presto/Trino. Internationalization efforts have brought improved translations in several languages, and further strides have been made in API V1 migrations. The release also includes numerous bug fixes and general improvements aimed at enhancing user experience and security, with ongoing community efforts toward the development of Superset 3.0 anticipated to bring further advancements.

Preset has introduced a range of new product features and enhancements, including the availability of cross-filtering for all users, which allows for more dynamic and self-serve data exploration across dashboards. The platform also now supports non-temporal X-axis visualization, expanding the ability to analyze trends beyond time-series data, and includes new visualization options like Apache EChart’s Sunburst Chart and stream graphs. Users can now connect to databases via SSH tunnels more easily, upload CSV files without needing an immediate database connection, and disable caching on a per-chart basis. The introduction of a "Secondary Contributor" role allows for broader dataset editing permissions within workspace-level roles. Additionally, workflow enhancements make dataset creation more intuitive, and updates to the Chart Builder view and dashboard filters offer improved flexibility and usability. These updates are designed to enhance user experience and provide more robust tools for data analysis and visualization.

Apache Superset 6.0 introduces a comprehensive theming system with design tokens and dark mode, marking a significant advancement in customization capabilities. The transition to this new theming approach involves moving away from LESS to the Emotion CSS-in-JS library, integrating Ant Design for a shared component library, and adopting Apache ECharts for core visualizations. This shift requires extensive migration efforts, including converting pages to a React Single-Page Application and consolidating components to ensure a cohesive design system. While the Superset community has made considerable progress, there remains a substantial amount of work to fully implement theming across the platform. Users can begin experimenting with theme tweaks through the THEME_OVERRIDES configuration, although full themability is still in development. The community is encouraged to contribute to these ongoing efforts, participate in discussions on Superset Slack, and attend meetings to advance the project's goals, with the ultimate aim of achieving features like a fully functional dark mode.