From Tables to Decisions: What Your Database Class Teaches You About Product Management
Three stories about grain, joins and messy data, and why SQL is a head start for product management, not a detour from it.
Read post →// Turning messy data into real value.
Data, AI, retail & product — from null to value.
I'm Darsana — a product leader who started out as a software developer and has spent about twenty years across software engineering, data platforms, analytics, and retail. These days I build GenAI platforms. NullToValue is where I write up what I'm learning: original research, practical frameworks, and lessons from turning messy data into something useful.
Latest posts
From Tables to Decisions: Why Dashboards Don't Change What Teams Do
Most analytics stacks end at a query result. Here's an idea for the missing layer — one that turns tables into decisions, not just charts.
Read post →Silent Anomalies: When AI Agents Keep Working — and Keep Getting It Wrong
AI agents depend on tools that change underneath them — renamed fields, switched currencies, new status codes. Nothing crashes, so nothing alerts. Here's an idea for catching these silent failures before customers do.
Read post →Semantic Freshness: Your Vector Search Index Is Going Stale — What If We Could Measure It?
Retail catalogues change thousands of times a day, but the vector indexes behind product search are built as if they never do. Here's an idea for measuring that hidden decay — and rebuilding only the parts that need it.
Read post →Seeing the Shopper: A Retail Digital Twin Built on What Customers Say, Feel, and Buy
Surveys tell us what customers say. Computer vision can tell us how they feel. Retail already knows what they buy. Put the three together and you get a digital twin of the shopper that can actually be checked against reality.
Read post →Untangling Retail Payment Data with AI — and Why Fraud Detection Is Next
Modern checkout systems accept 30+ payment types — but the data they produce is a mess. Here's how I approached synthetic data generation to make sense of multi-tender transactions, and where I'm heading next with fraud detection.
Read post →