April 2025 Summaries

Hex has acquired Hashboard, a business intelligence product company founded by Carlos Aguilar. The teams are already working together to build towards their shared vision for the future of data, with three main focus areas: building the best product for deep-dive data work, making it easy for everyone to use data, and improving performance at scale. As part of the acquisition, Hashboard's self-serve BI offering will be incorporated into Hex, and customers can expect to see improvements in data exploration, semantic modeling, agentic analytics, visualization, and more. The product will remain live for existing Hashboard customers, and new customers are encouraged to explore and migrate their workflows to Hex over time.

Nick Hardy, now VP of Data at Ro, transformed his data team by embedding analysts directly within partner teams, supercharging decision-making across the healthcare company. He shared his experience transitioning from working with data at BuzzFeed to handling sensitive healthcare data at Ro, and the thoughtful management that happens behind it. The analytics team is decentralized and embedded, with a mix of formal dotted line managers and stakeholder groups, and works closely with various departments. They've evolved their model to bring analysts closer to partner teams, unlocking more impact in decision-making. Data teams exist to help businesses make decisions quickly, and AI will take this to the next level.

Using GPT-4.1 as a case study, the author highlights the limitations of traditional vanity metrics such as pass rate in evaluating the effectiveness of large language models (LLMs) like themselves. They introduce the concept of "clarity metrics" which focus on actionable intelligence and growth unlocks rather than just shallow numbers. The author describes their framework for evaluating LLMs, called "The Funnel", which decomposes each evaluation into a series of cascading steps with its own pass/fail criteria. This approach allows for more nuanced understanding of the system and enables targeted improvements. By analyzing the funnel, the authors were able to identify specific issues such as column hallucinations in GPT-4.1 and develop prompt fixes that improved overall performance without increasing the number of incorrect data passes. The use of flux charts and orbit charts provides additional insights into the movement of evaluations through the funnel, enabling more precise experimentation and optimization.

Hex has introduced embedded analytics, allowing users to integrate its power into their products. This feature enables teams to build and deploy powerful, interactive data apps for customers or clients in a matter of minutes. With Hex's fast and flexible notebook and point-and-click app builder, developers can deliver first-class data experiences without relinquishing control over the details that make their product great. The new Embed API provides quick and easy deployment of embedded apps with support for pass-through authentication, row-level security, and white-labeled custom styling. This allows teams to power personalized data experiences from a single centralized Hex project.

I was honored when Mona Khalil invited me to write the foreword for her excellent new book, Effective Data Analysis, which I highly recommend. As a data analyst, I've encountered situations where there's no straightforward "right" answer, and one must fuse objective measurement with subjective interpretation and thoughtful communication. Mona's book breaks down these topics in all their nuance, unpacking concepts like defining metrics, the theory of measurement, and how to present findings to stakeholders. The book emphasizes the importance of simplicity in analysis, which is especially crucial as we enter the world of AI where models can produce results that require human interpretation and communication. Reading through this book has given me a new appreciation for the science and art of effective analysis, and I believe practitioners of all experience levels can benefit from it to avoid common pitfalls and make their own unique mistakes.

The need for self-serve analytics has not been fully addressed by current business intelligence (BI) tools, which have struggled to keep up with the increasing volumes of data and the growing complexity of analysis required. The divide between technical and non-technical users has created a natural barrier, with stakeholders often relying on others to dig deeper into data, leading to bottlenecks and tool sprawl. Large language models (LLMs) are being touted as a solution to bridge this gap, but they do not directly address the issue; instead, they offer a crucial piece of the puzzle. The key is for platforms to allow creators of data products to produce data apps and dashboards that cater to both data practitioners and business stakeholders, enabling more natural interaction with data through various modalities such as SQL, Python, chat-like experiences, or drag-and-drop exploration.

We've developed a feature called Magic Data Enhance that simplifies the process of presenting data insights to stakeholders by automatically adjusting charts and maps to make them more engaging. This is achieved through an AI-powered pipeline that uses cutting-edge research, including a state-of-the-art agentic AI model named FACADE, which can analyze data and adjust it based on user input and context. The feature is designed to save data teams time and effort, allowing them to focus on high-ROI work, while also providing stakeholders with reliable and consistent metrics. The Magic Data Enhance pipeline is built on a hybrid NVIDIA/Cerebras cluster and executes in under 7 minutes, making it possible to make permanent updates to the underlying database and propagate changes to upstream systems of record.