September 2022 Summaries

The digital age is characterized by the ongoing story of bundling, unbundling, and rebundling in the data space. Emerging tools provide flexibility and a lower entry barrier for practitioners, while established solutions offer a safe harbor with well-known integrated products promising big returns on investments. Cube and Apache Superset are alternative tools to Looker that provide a finer solution in many use cases, offering open-source licensing, separation of concerns, compatibility, and a headless approach. These tools complement each other by providing a unified semantic layer for data modeling and visualization, allowing for flexibility and reducing vendor lock-in. They offer free as in "open source" and ready-to-go solutions with freemium pricing models, making them an attractive alternative to Looker.

Demand for real-time and operational data is growing as consumers and enterprises expect applications to react quickly to changes, providing intelligent notifications and alerts. Cube aims to address this by creating a seamless experience for data engineers to work with streaming data alongside batched data, allowing for a single, headless data access layer regardless of the data type. Historically, working with real-time data has been challenging, but recent innovations in streaming SQL technologies have made it easier to process and analyze streaming data using SQL. Cube now supports streaming SQL engines and exposes streaming data via its REST, GraphQL, and SQL APIs, enabling developers to build real-time data apps without needing a new language or specific integrations. Additionally, Cube is introducing lambda pre-aggregations to address the complex problem of merging batch and streaming data, providing a single interface for unioned data and simplifying data consumption. The company plans to expand its support for Materialize, Flink SQL, and Spark Streaming in the near future, making it easier for users to build powerful real-time applications.

Cuboh, a restaurant-tech company, was facing "delivery-processing hell" due to its SaaS platform's inability to handle large volumes of orders efficiently. To address this challenge, the team sought a robust solution for ledger-based reporting, exploring options like Kafka before settling on Cube, an open-sourced and self-managed platform that provided low latency requests, near-real-time reporting, and caching capabilities. Cuboh successfully implemented Cube, reducing report generation time from minutes to seconds and lowering hosting costs by 80%. The company's experience highlights the benefits of using Cube for data modeling, access control, and caching needs, with a user-friendly configuration process and opportunities for community support.

There's a growing demand for customer-facing analytics, driven by consumers' increasing curiosity about their businesses and investments. Embedded analytics is becoming increasingly popular as companies integrate data analytics and visualization capabilities into user workflows. Tableau is a leading platform for data visualization and dashboarding, but it has limitations when used with embedded analytics, such as difficulties with customizability, duplicated effort, and weaker analytics performance. To overcome these issues, Cube offers a headless BI platform that streamlines building embedded analytics with Tableau and custom presentation layers by centralizing data modeling, security, and caching orchestration, ensuring uniformly performant and efficient embedded analytics.

Observable is a collaborative data canvas that offers interactive and reactive data exploration capabilities, built using JavaScript, which complements its native integration with Observable Plot, an open-source visualization library. It provides a rich set of cell types, tutorials, examples, templates, and notebooks contributed by a community of creators. Observable notebooks run in the cloud and can be embedded as whole or individual blocks. Cube is a headless BI platform that sits between data sources and downstream applications, providing consistent metrics and performance for all data flowing downstream. It works with modern data stores, creates a metrics layer with consistent metrics definitions, and delivers them to any data notebook, application, or BI tool via its SQL, REST, or GraphQL APIs. By using Observable and Cube together, teams can provide a common ground for all teams working together within an organization, eliminating inconsistent data usage across different teams. Observable complements Cube by providing a customized data model that can be used to define consistent metrics definitions. Users can connect their data to Cube through SQL API or REST API, and embed the notebook into various environments, such as HTML documents, React apps, or content management systems like Gatsby.

Data applications, encompassing tools like recommendation engines and customized reporting tools, provide interactive insights and actions beyond traditional embedded analytics. While embedded analytics integrate exploratory tools into existing applications, data applications focus on data explanation by offering dynamic, purpose-built user experiences. These applications include embedded data apps that evolve static dashboards into highly customized, dynamic features, internal data products tailored for specific business units, and consumer-facing apps requiring high performance and design customization. Built on cloud data warehouses, data applications utilize a headless BI layer for data modeling, access control, and caching, and are developed by integrating modern engineering workflows. While the foundational layers of the data stack remain consistent across application types, no code/low code tools and frameworks like Appsmith, Retool, Plotly Dash, and Streamlit facilitate the development of analytics interfaces and shareable web applications. As the technology stack advances, data applications will expand in number and variety, offering new opportunities for handling larger and more complex datasets, although traditional dashboard-centric experiences will continue to fulfill specific needs.

Here's a neutral and interesting summary of the text in one paragraph: Tree-shaking is the removal of unused code from JavaScript bundles, but not all code that looks unused is tree-shakeable. This concept is relevant when using ES6 imports/exports with JavaScript bundlers, as it allows for static analysis of used and unused exports. Tree-shakable design refers to writing code in a way that minimizes the amount of code included in bundles, making it smaller and more efficient. To achieve this, developers should follow best practices such as marking pure functions with /* #__PURE__ */ comments, using the sideEffects property in package.json files, and designing libraries with caution to avoid side effects. By doing so, developers can take advantage of tree-shaking and dead code elimination features in bundlers like Webpack and Rollup.