Ken Golonka

At South Shore Analytics, we've been building data stacks all year and we're pretty amazed with what's happening right now. The line between "data" and "operations" is basically disappearing. What used to be separate teams looking at dashboards after the fact is now real-time insights driving actual decisions in the moment. We're seeing companies where the data stack isn't just reporting what happened—it's actively shaping what happens next. Sales teams getting leads scored in real-time. Operations automatically adjusting based on demand patterns. Finance models that update themselves (huge shoutout to Omni's Sheets feature). It's wild how much changes when you stop treating data like a separate function of your business and start treating it like the nervous system of your business. The infrastructure we're building today isn't just about better reports. It's about creating businesses that think and adapt on their own. Pretty excited to see where this goes next. #DataInfrastructure #RealTimeOperations #SouthShoreAnalytics

33

Quick look at our in-progress planning cycle with some random thoughts throughout: the Zillow Rentals team has been deep in our Three-Year Planning process (3YP), carving out time in-person together to think bigger, look ahead, and imagine new ways to make the rental journey work better for renters and housing providers. The ideas coming out of this year’s planning cycle are some of the most creative I’ve seen - all grounded in the real challenges renters face today. Renters need more support than ever, and landlords, property managers, leasing agents, etc. are uniquely positioned to benefit from smarter, more modern solutions. Making time for long-term thinking is critical - and honestly, refreshing. Building a marketplace takes both tactical execution and space to dream. Dreaming too close to the present can get caught up in real constraints. Looking three years ahead gives just enough distance to unlock creativity (art) while staying grounded in reality (science). So many of the innovations that are now live on Zillow Rentals, like our free rent reporting option to help renters build credit, universal application, and new tools for multifamily owners/operators all started as rough concepts on a whiteboard, in a document, or during a heated discussion about what could be next. Can’t wait to see where this cycle takes us. One final thought from this 3YP cycle: ‘great ideas can come from anywhere (and at any time)’ really holds up. When it comes to long-term direction and strategy, that couldn’t be more true. While we set aside time for focused 3YP discussions, the reality is that it’s a living process that is constantly evolving in some draft form that should always be open for comments.  Curious how other teams approach this kind of planning — always open to new/different ideas.

95

5 Comments

For years, I’ve watched teams pour time and money into massive data platforms like BigQuery, only to hit a wall when it’s time to turn data into decisions. That’s why I’m pleased to share Glean’s new structured query agents that let anyone unlock real‑time insights from BigQuery. BigQuery power users can write and run SQL directly inside Glean. Everyone else can simply ask in plain English, and Glean – using Google Cloud’s Conversational Analytics API – converts it to SQL and returns answers quickly. You can even tap “Show work” to view the generated SQL for transparency. Ask things like: - “Which features have the highest weekly active usage among enterprise customers?” - “What’s our win rate for enterprise deals this quarter compared to last?” This is an important step in democratizing access to the data that drives our businesses. It makes it easy for anyone–with the right permissions–to analyze data, automate reporting, and make better decisions. If you’ve ever felt frustrated by the distance between your questions and your data, we built this for you. https://lnkd.in/g2mrskjc

860

24 Comments

"We can just plug in the historical data right?" My head: "No, because it's captured differently with different segmentation, at a detail level we don't support through two different systems with inconsistent naming conventions and saved in CSVs on someones hard drive but we don't remember who it is, except for the weeks they went on PTO which are just missing and can't be replaced." My mouth: "We'll look into it."

967

39 Comments

AI in mortgage isn’t optional anymore...it’s essential. I sat down with Melissa Langdale and she shared great insights on what is holding lenders back: 🔺 ROI confusion. You invest in AI, but can’t pin down real savings. 🔺 Regulatory worry. Generative AI making decisions? Fair-lending and compliance questions pop up immediately. So how do you move forward? ➡️ Rethink your workflow. Stop thinking in milestones. Start thinking in tasks. • Once income docs are in, let AI calculate and lock it—no extra hand-offs. • Break work into bite-sized steps and ask: where can AI slide in? ➡️ Nail your data. Your AI is only as good as your data. Clean it now so tomorrow’s tools snap right into place. ➡️ Engage regulators early. They’re already using AI to analyze HMDA data and flag risks. Talk to them. Know the guardrails. ➡️ Start small, win big. Automate the little things—voice reminders, document assembly, repeatable tasks. Track the time and dollars saved. ➡️ Own your mindset. Innovation isn’t a checkbox. It’s reimagining how you work. Curious how much faster and leaner your operation could run with AI Mortgage Manufacturing™? let's connect. 🎥 Full episode coming soon. Stay tuned! #MortgageTech #AIMortgage #Efficiency #LoanManufacturing #TRAiNEDAI _ _ _ 👋 I'm Mark Cunningham, Co-Founder and CEO at TRAiNED AI. We are on a mission to eliminate costly, manual processes with AI-driven automation, allowing mortgage lenders to scale without adding staff and cut costs without sacrificing efficiency. Curious how AI can scale your business? DM me and I'll show you.

41

7 Comments

📍 Data infrastructure startups 101 I've spent the last 10 years exclusively building early-stage data products. I learned these patterns through repeated failure and occasional success. 1. Your docs are your best form of marketing. They say everything about your customer commitment, taste, and user empathy. 2. To sell into mission-critical use cases, be patient and start by selling into non-mission-critical ones. Customer trust is built through small wins. 3. Prefer to be more public than private. Obscurity is more likely to kill your startup than copycats. 4. Pick co-founders with opposite skills. If you're an engineer, find a co-founder skilled in sales and marketing to balance you out. 5. As a rough rule of thumb, in the beginning, spend 2/3 of your time on sales and marketing and 1/3 on product/engineering. 6. If you want to build a community, make it easy to contribute. To docs. PRs. Patches. The harder it is, the fewer people will try. 7. Pick one and only one distribution platform. Trying to do open-source, cloud, and on-prem is a disaster for everyone. 8. Avoid raising money or hiring until you have customers or high confidence you can achieve product-market fit. 9. You can't outsource marketing. The founders are the marketers. 10. Sell your product based on the business value it provides—not on the underlying technology. 11. Work on usability from day 1. There is nothing more telling about your attitude towards customers. 12. It's okay to change your business model over time. No one has it figured out in the beginning. 13. Put a deadline on proof-of-concept projects to avoid the indefinite POC. 14. Technical marketing is king, castle, and moat. The best communicators win, not the best products.

75

7 Comments

The new State of Analytics Engineering Report from dbt Labs just dropped—and it's full of insight into how data teams are adapting to AI. 👀 70% are already using AI to assist with code 💰 AI tooling is the biggest investment area right now 🔍 But quality still leads—unreliable inputs = unreliable outputs The message is clear: AI is changing data work, but trust still comes first. Read the full report: https://lnkd.in/esaqPFX8

36

I was the first committer on dbt over 10 years ago. What Tristan and his team started was a movement. They recognized the need for a data engineer who understood the business and how to analyze it - an “analytics engineer.” Today, that role is evolving again. The real work isn't just building dashboards or modeling tables anymore; it’s about context engineering. In this next era, your leverage comes from curating a model that serves as a shared brain for the company. When you get the definitions and guardrails right, you are building the exact context an AI needs to speak your company's language without hallucinating. If data engineering owns ingestion and analytics engineering owns modeling, the next generation owns the semantics. But this isn't a solo task anymore; it’s a multiplayer game. You’re building a shared surface where humans and AI co-author the logic that turns raw data into a company truth. The ticket-taker era is over. If you don't own that collaborative context, you're just building a high-speed engine for hallucinations. Your leverage isn't in the SQL you write; it's in the multiplayer guardrails you build to make sure everyone—human or agent—is playing by the same rules. We’re building the next phase of that movement now. Are you in? 🚢

244

6 Comments

As we have been working with customers who are adopting Snowflake Semantic Views, one of the top requests has been a native integration with dbt Labs for materialization and referencing of Semantic Views within dbt models. I’m excited to announce that there is now a new dbt package, available on Package Hub, that does just this! You can read Yutsing Liu's blog here to learn more: https://lnkd.in/deXJtJBQ In summary, the dbt_semantic_view package introduces a new semantic_view materialization for dbt Core projects targeting Snowflake, enabling you to define business-centric semantic views (tables + relationships + dimensions + metrics) via dbt and automate their creation, renaming, and dropping. Link to package is in the comments.

365

17 Comments

Last week I spoke at Snowflake's World Tour event. The data ecosystem + career paths have changed quite a bit in the last four months since I presented at Snowflake Summit. A few changes I've seen on our data team since then: Today’s AI tools shorten "time-to-impact". We've seen new hires use Claude Code, Cursor, Notion AI, etc. to onboard faster to codebases and business domains. If you can search, reason, and ship, you can contribute far faster than before. You cannot measure agentic product experiences without full-stack data talent. Ramp's proportion of LLM products, questions, and decisions will only go up. AI-as-a-feature requires eval datasets, offline/online metrics, and drift monitoring. This is product analytics with semantic rigor vs. siloed DS research or SQL-only data modeling. Increasingly, the skill we're asking our team to develop / hiring for is owning the full product loop (metrics → modeling → experiment → decision → monitor) vs. deep specialization in one tool, language and ecosystem (whether "applied science" or "analytics engineering"). Big thanks to Nicole and Bobby for having me!

114

4 Comments

It's early days, but agent commission models are starting to change for homebuyers. New data from Lance Lambert shows that 1/5 buyside transactions now include either flat fee or negotiated commission rates. 3% really doesn't make sense in a world where ~80% of buyers are finding the home themselves.

40

4 Comments

Everyone has the same data. Only a few process 450,000 markets daily. Let’s be honest: CoStar, brokers, asset-specific data, census data...we’ve all got access. That’s not the problem. The real issue I spotted came down to bandwidth. Most teams can fully analyze maybe 2–3 markets a week. Even if you’re fast, that’s ~150 a year, amazing! BUT there are over 450,000 sub markets in the U.S. Last week, a prominent self-storage developer told me: “We spent 3 weeks on Charlotte and came up dry. PropRise showed us 12 solid pockets we hadn’t even considered in 30 seconds.” That’s not some AI buzzword. That’s just what happens when you stop guessing and start scanning everything. We flipped the question from: “Which 20 markets should we go after?” → “Why aren’t we looking at all of them?” And that unlocks stuff like: • Markets you’d never put on a target list • Trade areas with $2 rent per square foot 10 minutes from oversupplied zones • Actually having a *differentiated* pipeline on Monday’s IC call This isn’t about “secret data.” It’s about doing what humans simply aren't built to do: scanning the whole country daily, not just what fits in a spreadsheet. If you’re still doing top-down analysis manually… how many deals are slipping through? Would love to hear how many markets are you actively evaluating right now?

12

2 Comments

Data ≠ Health. Wearables are everywhere. Outcomes are not. Here’s why dashboards alone won’t move the needle. ☑ Adoption is up. Results are flat. 1. Nearly half of Americans wore a health tracker in 2023. 2. Obesity rates stayed stubborn—over 40% of adults, with severe obesity at 9.4%. 3. Life expectancy inched up to 78.4 years, but chronic disease still weighs heavy. ☑ Engagement is the real sinkhole. Most health apps lose users fast. Only 3–8% stick around after 30 days. ☑ More data does not mean better data. 1. Sleep tracking on wearables is not as accurate as gold-standard tests. 2. Even Apple’s own validation shows high sensitivity but lower specificity—meaning lots of false wake-ups. 3. Atrial fibrillation alerts are useful (PPV ~0.84), but still need clinical confirmation. ☑ Behavior change is real, but modest. 1. Meta-analyses show wearables boost steps by about 1,800 per day and help with about 1 kg of weight loss. 2. More steps do not always mean more intense activity. 3. Higher step counts link to lower chronic disease risk, but that’s correlation—not proof that wearables alone cut disease. ☑ The path forward: From dashboards to decisions. 1. Data alone is not enough. Most people need help turning numbers into action. 2. AI-powered coaching can bridge the gap—translating streams of data into real, daily decisions. 3. Real-world impact needs persistence, integration, and support—not just more metrics. (testing a new flow: personal research & topic ideation --> GPT post ideation --> GPT research/validation --> EasyGen + human editing)

208

59 Comments