I Use EDA and Local LLMs to Make Better Product Decisions

4 points

a year ago

How product managers supercharge AI/ML product decision-making with smarter EDA. Good decisions start with good questions. Before product managers decide what features to build, one of the most overlooked yet critical steps is: Exploratory Data Analysis (EDA)

Why? Because good EDA accelerates hypothesis validation, helps you ask the right questions, test assumptions, apply statistical techniques, uncover insights about the user journey, clarify KPIs — and ultimately helps you make smarter, more strategic, and user-centric decisions.

Over the years, I’ve developed a lightweight but effective process to help myself and teams move faster — especially when: dealing with sensitive or PII data that can’t leave local networks working with very large datasets streamlining collaboration with analysts and data scientists

Here’s my go-to setup that has saved us days (if not weeks): Data analysis & visualization — Jupyter Notebook + Pygwalker No more exporting CSVs or bouncing between BI tools & raw data. Local LLM with LMStudio When I need help exploring hypotheses, drafting SQL, or summarizing findings

Here is my LinkedIn post about this topic. https://www.linkedin.com/posts/peekay-chan-453102_pygwalker-lmstudio-productmanagement-activity-7352780192395792384-AFDe?utm_source=share&utm_medium=member_desktop&rcm=ACoAAAAD2_8BA9f2IpYDPZmflw9ziUIVH_mw7V8

Would love to hear how other PMs/Analysts/Data Scientist go about this.