Find Your Happy Place
Dan Brickey · Data Platform Architect — 25+ years in data engineering, Data Vault 2.0 + Kimball dimensional modeling, dbt / Snowflake / DuckDB. This is one of my portfolio projects. ↓
Find Your Happy Place
Where would you be happiest living in the US? It depends entirely on what you value. Gazetteer turns dozens of disagreeing public datasets into one place where that question can be played with — line up cost of living, income, housing, jobs, population, and health for the same place and the same period, and weigh them by what you care about.
What it proves. An integrated analytical platform built on a 4-layer medallion architecture — Data Vault 2.0 integration, a Kimball star consumption layer, ~10 reconciled public sources, generated documentation, an honest predictive-modeling experiment, and an interactive dashboard you can actually use. Architected by a human, built by directed AI. See how it was built →
Try it now
Turn the dials to say what you value, and the map re-shades and the ranked list re-orders live — it's a real interactive tool, not a screenshot. A good 10-second story: open Find your happy place, slide Real income (cost-adjusted) up and Cheaper homes to buy down, and watch high-cost coastal counties fall while interior-West counties climb — your dollar going farther, made visible.
The story
The face. Find Your Happy Place — the approachable door. Type in your county, compare it to a friend's, and immediately get something: "oh, my dollar goes 30% further there." Legibility is the point.
The substance. Underneath is a genuine analytical instrument for exploring place — comparison modes, criteria-matching, trends over time. The friendly face isn't a lie; it's how the substance invites you in.
The discipline. Staleness-tolerant, legible, exploratory — deliberately not operational, not actuarial, not an opaque recommender. It hands you the ingredients and the dials; you decide what "happy" means.
At a glance
This release covers 3,222 counties across 1969 – 2025 ( 180,533 place-years in all). Every measure carries its own coverage window, so a place-year appears whenever any measure has data for it.
The platform — as of v6 (2026-06-16)
- 31 modeled objects — 21 silver (Data Vault) · 8 gold (star + business vault) · 2 platinum marts
- 1 hub · 3 links · 2 bridges · 15 satellites · 4 computed satellites · 1 PIT · 2 dims · 1 fact
- ~10 reconciled public sources onto one geography spine, county + tract grain
The stack
DuckDB · dbt · AutomateDV (DV2.0) · Kimball star · Evidence.dev
· Python / scikit-learn (ML) · dual-target DuckDB + Snowflake
The contract. Everything you see is computed in the data vault; this dashboard only dials — it ranks and compares the ingredients (raw measures plus their national percentile), it never bakes a single "happy score." Weighting is yours to do. New here and want to understand and use it first? → Read the guide →. Want the engineering story? → How this was built: the architecture, the data model, and the design decisions →