October 2021 Summaries

ClickHouse's materialized columns offer a significant optimization opportunity for query performance by creating new columns derived from existing data, which can expedite queries that involve parsing JSON objects. In an example from PostHog, a schema was optimized by adding a materialized column for URLs, which reduced the need for JSON parsing and decreased the data read from disk. This approach demonstrated a remarkable speed improvement in querying pageviews by reducing execution time and disk read significantly. Although this method increases storage requirements, ClickHouse's compression capabilities make it an efficient trade-off, with substantial performance gains evident, especially when multiple property filters are applied. To efficiently backfill existing data without rewriting entire tables, the article suggests using default type columns before finalizing them as materialized, thus optimizing performance without disrupting existing processes. PostHog utilizes this technique selectively, focusing on properties that most frequently slow down queries, leading to an average query time improvement of 55 percent. This strategic use of materialized columns highlights the potential of ClickHouse in enhancing analytics capabilities across extensive datasets.

PostHog's latest update, version 1.29.0, introduces a variety of new features and improvements aimed at enhancing user experience and data analytics capabilities. The release includes a revamped Paths feature to better analyze user actions, multivariate support for more complex feature flag testing, and the introduction of private projects to enhance privacy controls. Additionally, trailing DAU/WAU/MAU graphs are now available to help measure user engagement more effectively. The update also offers several UI and functional improvements such as custom graph series names, a column configurator for event tables, and a revamped password reset process. A deprecation notice highlights upcoming changes to API paths in preparation for the next version, 1.30.0. The community is actively involved, with recognition for contributions and open roles available for those interested in joining the team.

Product analytics involves collecting and analyzing data on how a product is used to make informed decisions about its improvement, and it encompasses both quantitative and qualitative data. Tools like PostHog are used for funnel analysis, session recording, and measuring user interactions to optimize product performance. Although nearly all modern businesses, from startups like Pry to established companies like Hasura, utilize product analytics, the tools are not limited to product teams; they are used by software engineers, UI/UX designers, and senior management to guide decisions on product roadmaps and interface design. Metrics such as the pirate funnel (AARRR) are essential for tracking user engagement across different stages, from acquisition to referral. While platforms like Google Analytics are popular, they are not ideal for product analytics due to their focus on web metrics rather than user behavior insights. PostHog offers a self-hosted solution with features like feature flags and session recording, allowing businesses to keep data on their own infrastructure and avoid third-party data sharing.

PostHog is removing its "Sessions" page due to user confusion and limited value, replacing it with a "Recordings" tab on the main menu to better address user needs. The Sessions page, which listed user sessions based on inactivity periods, was criticized for issues such as missing recordings and inadequate filtering, prompting a reevaluation of its purpose. PostHog identified two main use cases for session analytics, but found its existing features insufficient for comprehensive analysis, leading to an emphasis on Paths and Quantitative Analysis for deeper insights. The new Recordings page aims to streamline finding specific user recordings with enhanced filtering tools, while sessions-related insights are being reconsidered due to their limited utility. Additionally, improvements to recordings ingestion and playback are underway, with A/B testing in PostHog Cloud and considerations for future support of session-based behavioral analytics, including customizable session definitions and enhanced analysis capabilities.

Creating a successful mascot involves a complex and iterative process that goes beyond mere aesthetics to embody a company's brand identity, tone, and vibe, as illustrated by the creation of PostHog's hedgehog mascot. The journey involves careful consideration of what the company stands for and who it is selling to, as well as learning from previous mistakes, such as designs that inadvertently conveyed outdated or inhuman qualities. Market research and inspiration from companies like Sentry and Hasura help guide the creative direction, while the use of digital tools like ProCreate and Fresco allows for rapid iteration and experimentation with different designs, such as 3D models and various character traits. The hedgehog mascot underwent numerous iterations, exploring different styles and characteristics, to ultimately become a relatable and friendly figure that embodies the company's values and appeals to its audience.