What is first party data?

First party data is a concept in personal epistemology: Direct observation produces higher-signal data than filtered accounts. Every layer of transmission between you and reality introduces distortion — compression, editorialization, selective emphasis, cultural normalization. First-party data is not just more convenient. It is structurally different from second-hand reports, and treating them as equivalent is a signal-processing error.

Example: A product manager reads a quarterly NPS summary that says customer satisfaction is "stable at 72." She presents this number at the leadership meeting. The VP asks if there are any concerns. She says no. Meanwhile, the support team has noticed that three enterprise accounts have filed nearly identical complaints about a workflow change shipped six weeks ago — complaints that were categorized as "UI feedback" in the ticketing system, averaged into the broader satisfaction number, and never surfaced as a pattern. The NPS summary is not wrong. It is compressed. The three complaints are first-party data: specific, timestamped, attributable to a causal event. The summary is a second-hand report: aggregated, abstracted, stripped of the contextual detail that would make the pattern visible. The product manager is making decisions based on a number that has been processed through two layers of abstraction — the survey instrument and the analytics pipeline — and each layer lost information that mattered.

This concept is part of Phase 7 (Signal vs Noise) in the How to Think curriculum, which builds the epistemic infrastructure for signal vs noise.