Cassie Kozyrkov

We're facing more and more questions about the use of synthetic data in market research. I'd recommend this blog post, which really resonated with me. So much of the conversation is "how closely can this replicate human data?" But there are so many other considerations researchers should be thinking about. https://lnkd.in/eSQfSTPj

31

3 Comments

Last week, I spoke at the Data Behind the Scenes conference #DataBS It's a conference run by Randy Au for data practitioners who are tired of polished case studies that skip over the actual work. Instead you get honest takes on project realities: the messy bits, the trial and error, the things that go wrong. I spoke about the 10 red flags that your multi-market research data went sideways. We've all had that moment, staring at survey results that look technically correct but feel fundamentally wrong, then having to explain why three countries' worth of data is unusable. Links to my talk, as well as the full playlist, are in the comments

16

5 Comments

We're thinking about semantic layers and AI completely backwards 🔄 The data industry narrative right now is "You NEED a semantic layer so AI can understand your data." It’s true, but the reality is we're still manually defining every metric, dimension, and relationship like it’s 2015. If AI can write SQL, debug code, and recommend business strategy... shouldn't it also know that "Monthly Recurring Revenue" and "MRR" are the same thing? I'm betting the next big breakthrough isn't just AI consuming our perfectly curated data models. It'll be in the AI building with us. What do you think? Are we one breakthrough away from self-improving semantic layers, or will this always require human intuition?

38

27 Comments

Building agentic workflows? There are actually only a few key concepts to be aware of. StateGraph orchestration: A StateGraph is the agent’s blueprint. It consists of nodes that handle tasks with edges (directions) to other tasks, defining the order. Together, they provide the agent with a flow to follow. Conditional routing: At decision points, the agent inspects results and follows “if-else” statements that skip failures, handle edge cases, etc. Memory and threads: Thread memory holds recent conversation turns. Essentially, it stores the short-term memory that you are John Doe and were interested in the weather. Long-term memory stores user facts, letting the agent stay context-aware, reduce repetition, and personalize replies, such as that you have been a customer since 2018. Human-in-the-loop: Strategic hand-offs let people review low-confidence outputs, catch critical paths or errors, and feed corrections back to the model, keeping quality and alignment tight. If this still feels too abstract, I’ve included a more detailed explanation in the article below. https://lnkd.in/eAr-6YAQ

14

1 Comment

It used to be accepted wisdom that high-quality causal marketing measurement was better left as an "artisan craft." Data scientists and analysts worked by hand, manually tuning models and dropping outlier data. The belief was that you can’t or shouldn’t automate this stuff - it’s too hard, and in doing so, you lose the nuance in the process. Haus Chief Scientist Joe Wyer explains how we turn that accepted wisdom on its head, automating hundreds of causal models weekly to intentionally remove that human-driven nuance. Why? Because when humans process noisy marketing data, they face a million seemingly reasonable choices with few objective principles to guide decisions. Over time, these choices steer results toward what feels comfortable, not necessarily what's true, and the value of data analysis evaporates. At the heart of any analysis at Haus is our Model Reliability Index (MRI): an objective scoring approach that stress-tests models through thousands of placebo simulations to measure how reliably a model can recover the truth. MRI powers two key systems: 1) The Gauntlet: Every model runs the gauntlet of real, messy customer datasets before it ships. If the MRI score doesn't beat a textbook baseline, it's dead. 2) Stone Turner: For every customer analysis, MRI automatically selects the optimal configuration based on reliability, not lift. Together, these systems enable Haus to scale causal marketing measurement almost entirely without human intervention - and remove the risk of unconscious bias. And that’s a very good thing for businesses.

54

3 Comments

Ben Lorica writes: The Multimodal Lakehouse has emerged ... utilizing formats like Lance to resolve a tension between BI and AI architecture. This, to "prevent GPU starvation without siloing the data stack." https://lnkd.in/eZQcey24

1

1 Comment

Humans don’t respond to data. They respond to meaning. So why do so many data presentations still fall flat? I loved hearing my teammate Kaitlin Henze featured on the Gut + Science podcast, talking with Nikki Lewallen Gregory about how to turn charts into conversations—and numbers into narratives that actually inspire action. They discuss storyboarding with stickies, the narrative arc, bringing problem-solving superpowers to little learners, and much more. 🎙️ Episode: From Numbers to Narratives: How to Turn Data into Action (link in comments below) 👇 If you listen, I’d love to hear your favorite takeaway or quote!

113

5 Comments