Ridge AI

Join me as I do a LIVE, unrehearsed, real-ass demo of Ridge AI, a new embedded analytics tool that I love. Two things really stand out about Ridge in my testing. One, the performance is ridiculous. It's the fastest BI interface I've ever used. Two, the architecture is very interesting. They bundle analytics, context and chat into a single unit that ships to the user's browser. Nobody does that, and for embedded it's really valuable IMO. So join me this Thursday, Wed 24th at 12 ET / 5 UK to see it live and get your questions answered.

So, what have we been cooking up at Ridge AI for the past year or so? A new visualization stack that is database-optimized and AI-friendly; agents to assist data transformation, visualization, and exploration; and an integrated platform for turn-key embedded analytics. New Visualization Tools Building on experience with D3, Vega, and Vega-Lite, we've developed a new layer cake of visualization tools to support scalability, interactivity, and usability by both people and AI agents. Ridge Plot is an interactive grammar of graphics that richly integrates with the Mosaic architecture for scalable, optimized database access and cross-filtering. Ridge Plot provides expressive visualizations, rich interaction, and responsive resizing – specified via either direct Svelte components or declarative JSON. Ridge Charts is a charting framework built on Ridge Plot: it provides a smaller surface area great for agentic development, while providing an array of customization options. We deploy these tools alongside a visualization knowledge base to power chart generation, refinement, and evaluation. The result is (we hope) an elegant coupling of agent flexibility with robust, deterministic tools for automatic design refinement and review. Agents Galore We've also been building a suite of AI agents to assist data preparation, visualization, and exploration. Our Transform Agent supports lightweight preparation to get data in shape prior to visualization: cleaning up data types and column names, adding derived columns, and so on. The Explore Agent enables open-ended Q&A with your data, translating analysis questions into SQL queries and charts, to both provide answers and prompt more refined questions. Our Build Agent guides people through dashboard authoring: eliciting context and driving questions; running preliminary analyses; suggesting filters, metrics, and charts to meet articulated goals; and supporting interactive refinement. Along the way, we've explored different approaches to agent harnesses, including an adaptive toolset that adjusts to the current phase of work, focusing agent attention where it matters most. …And Everything Else! Of course, we've also been busy building out a modern, cloud-based platform for embedded analytics. This includes data connectors, preparation and partitioning workflows, and secure publishing via standard web pages or direct embedding. All of this is in service of rich, in-browser analytics: dashboards with integrated data agents responsive to a long-tail of questions, while retaining what works for successful visual analytics. That said, much remains to do, including richer collaboration, more chart types, service integrations, and meta-analysis tools (to analyze analyses!) for continuous learning. As we continue the Ridge AI journey, we'll continue to share what we're up to at https://lnkd.in/guFetFZR Built with Fritz Lekschas, Andy Caley, Elias Mahfoud, and Ellie Fields

It's almost a year since I joined Ridge AI, and today we’re coming out of beta and opening up to anyone. A nice way to mark the occasion 🎉 It’s been quite a journey, from adding a data agent to the dashboard experience to building whole dashboards through conversation with our build agent. What excites me most is how this shifts your focus toward the questions you want to answer with data, while taking away much of the busywork of assembling charts. Last month we introduced our new build agent, which is much more flexible in how it helps you clarify your analytical goals and builds a dashboard around them. Bring a clear brief and it gets to work. Start with a broad goal and it asks focused questions grounded in your data. As the dashboard takes shape, it can revisit earlier choices and adjust what it builds as your direction evolves. Since then, we’ve continued testing and refining both the initial dashboards and how you work with the agent to improve them. A few updates I particularly like: • In-place prompting lets you ask questions or request changes directly on a chart or metric. Probably my favorite new feature, because I don’t need to take my eyes off the chart I’m working on. • Improved dashboard composition and layouts draw on research from the visualization community to give important views appropriate emphasis and make related charts easier to compare. • Expanded metric cards add comparisons and sparklines. Seeing that a metric increased is useful. Seeing whether that increase follows steady growth or a recent dip gives you much more context. • Direct editing and conversation work better together. You can resize chart rows, edit text, and make your own adjustments, and the agent now keeps track of those edits to better respect them as you continue. Attached are four of my favorite initial dashboards from our evaluations, along with the questions that started them. An initial dashboard is usually just a starting point for exploration to help you identify promising next steps and get ideas for refinements. Try it with your own data! Sign up for a free trial: https://lnkd.in/g6n3DDbK And if you’re in academia or a nonprofit on a tight budget, reach out and we can get you set up with a free account. It’s been fun building this with Ellie Fields, Jeffrey Heer, Andy Caley, and Elias Mahfoud! Happy visual data exploring! 📊

I'm glad to announce that Ridge AI is generally available! We're opening up to the world. Anyone can go to our website at ridgedata.ai, sign up, and get a trial of the product. By rethinking the problem from the ground up, we've been able to solve some of the gnarliest problems of embedded analytics: 𝗣𝗲𝗿𝗳𝗼𝗿𝗺𝗮𝗻𝗰𝗲 𝗶𝘀 𝗹𝗶𝗴𝗵𝘁𝗻𝗶𝗻𝗴 𝗳𝗮𝘀𝘁. Check out our gallery and see for yourself, or get a trial. 𝗬𝗼𝘂 𝗰𝗮𝗻 𝗮𝘀𝗸 𝘁𝗵𝗲 𝗻𝗲𝘅𝘁 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻. The dashboard is there for the big picture, but a data agent is there too-- with multiple rounds of validation, so the answers are good. Link the dashboard and data agent for richer interactivity. 𝗕𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗶𝘀 𝘁𝗿𝘂𝗹𝘆 𝗳𝗼𝗿 𝗮𝗻𝘆𝗼𝗻𝗲. No learning curve. You just need data and some questions (really!) The Build Agent translates business speak into best-practice dashboards. Most analytics platforms are server-based monoliths, leading to slow performance and high charges for compute. Ridge takes a different approach. We deploy standalone analytical apps to the user's browser, making the whole system more lightweight. Cost-wise, this approach saves our customers money: Local computation for analytics means lower compute charges in the data warehouse, and pricing that makes it easy for our customers to start on one use case and expand as needed. Our mission is to help anyone share, understand and act on data. Great to be building toward this mission with Jeffrey Heer, Andy Caley, Fritz Lekschas and Elias Mahfoud. We're on it! Read more in the blog: https://lnkd.in/gSw5dmhE

I got my hands on Ridge AI! Video below 👇 Three things I love and talk about in the video: 1️⃣ "Design For Steering, Not Just Doing" is a great mantra for AI-app UX from Jakob Nielsen. One-sentence prompts don't work. Ridge encourages you to upload reference docs and more, nudging you to provide as much context as you can. Really helpful. 2️⃣Direct interaction between Chat Window and Dashboard. Ridge's fast interactivity tech really brings the connection between the charts and the chats to life. 3️⃣"AI Where Your Eyes Are." I wrote about this recently: I want an AI chat box to be where my mouse cursor is, not shoved off in some side-pane. Great to see Ridge enabling this. Things up for debate: In my closing thoughts, I point out that, coming from Tableau, Ridge feels light on ways to tweak and customise the dashboard. Is this good? The longer ago my Tableau days are, the more I realise most people just want basic stuff: KPIs, bars and lines. The smaller feature list is by design from the Ridge team, but is it what the industry wants? What do you think? Thanks to Ellie Fields and the team for the early access.

In all product design, there's a frontier between power and ease-of-use. More power, harder to use. Easier to use, more of a toy product. That was true when every capability had a unique affordance for the user. AI chat bends this curve. Instead of adding an affordance for every new capability, someone can just ask AI to do it. This means every new capability doesn’t have to be a new affordance that clutters the user interface.  Ridge AI’s Build Agent uses AI in two ways:  1. To let the builder to give general instructions in natural language. 2. To guide the conversation to the business-value level. This means the builder doesn’t have to translate their business intent into the language of axes and fields. This is why you can tell the Build Agent that you want to understand the link between value and data. For example, “𝘋𝘰𝘦𝘴 𝘮𝘰𝘳𝘦 𝘵𝘳𝘢𝘪𝘯𝘪𝘯𝘨, 𝘦𝘢𝘳𝘭𝘪𝘦𝘳 𝘪𝘯 𝘰𝘯𝘣𝘰𝘢𝘳𝘥𝘪𝘯𝘨, 𝘳𝘦𝘴𝘶𝘭𝘵 𝘪𝘯 𝘩𝘪𝘨𝘩𝘦𝘳 𝘳𝘢𝘵𝘦𝘴 𝘰𝘧 𝘤𝘦𝘳𝘵𝘪𝘧𝘪𝘤𝘢𝘵𝘪𝘰𝘯?” The Build Agent guides you through the process and translates your goals into related charts on a dashboard. You don’t have to design the dashboard then click an endless number of buttons to build the dashboard you want to see. But chatting only gets you so far. Shout out to Industry Analyst Andy Cotgreave, among others, who has argued that a chat bot is a terrible interface for a lot of data work: https://lnkd.in/gfsPYMPj We agree. Sometimes you don’t want to type a whole thing. Sometimes you just want to change something directly. This makes the product feel much more fluid and gratifying. It’s also easier to learn and use– just touch and change the thing you want to change. We've added a host of direct interactions in Ridge, being thoughtful about making the product discoverable and intuitive. They're all listed in the blog post linked below. My favorite is reordering of charts, shown in the image for this post (but one-click delete is right behind it). We'll continue to look for ways to bend the curve in analytics interaction and design. Blog: https://lnkd.in/ga-U3ary

"Visual data helps streamline the value conversation with customers, which is always a challenge." A customer success leader in our conversations last week. Why don't more companies use data to prove value? 1. No data about value. Product team, looking at you for this one. 👀 2. Team has no access to data. Highly fixable. 🔨 3. Unvetted analytics. This puts the burden on the field team to parse different results each time and figure out what's right. 🤷♀️ 4. Decent analytics, but not customer-facing. Again, burden is on the team, this time to read the customers' mind and engage when they're considering churning. 🤔 5. Sloppy, vibe-coded numbers that no one owns. This is when the data team acts as the "human harness," 🐴 chasing around numbers from Clauded dashboards and managing data updates, performance, permissions and a dozen other issues. The right way to do this: shared, vetted, visual analytics that provide long-tail Q&A. That's what we're building at Ridge AI.  When you've got proof of value, you've got a conversation. You're identifying risk by showcasing gaps in adoption or engagement levels. You're finding opportunities in the business, making the field team partners to the customer, rather than sellers. You've got shareable proof of value, and confidence in it. Data changes the meeting with customers from "YOLO, buy more!" to a real conversation.

Great to see some of the OGs in person in Seattle at the Ridge AI office! What a range of perspectives on the most interesting time in data in decades. Steve Wexler Keith Helfrich Jonathan Drummey Mark T. Nelson Taha Ebrahimi Bo McCready Ann Jackson Carl Allchin Jim Dehner MS Kirk Munroe Lorna Brown Pablo Gomez Zach Bowders, MBA Andy Caley Elias Mahfoud Neal Baron