The Twin Cities Living Quality Map is a data pipeline and interactive dashboard built to explore public safety, development, amenities and services, and housing affordability across St. Paul and Minneapolis districts, combining open government data, OpenStreetMap, Census, and MnDOT/CDC figures into a single comparable Living Quality Score per district. See below for the scoring methodology, Data Sources for where the data comes from and the tech stack, Score Drivers for what actually moves the score, Trends for the underlying data over time, and Home Price Prediction for a test of whether public Census data can predict home prices, built on the same pipeline.
The Living Quality Score is a comparative index, not an objective measure of quality of life: it reflects a specific choice of metrics, geography, normalization, and weighting, and is only as complete as the public data available for each city. Use it to compare districts, ZIP codes, or buildings against each other, not as a verdict on any one of them.
The score also tracks neighborhood demographics closely: across the 28 districts, poverty rate and racial/ethnic diversity alone explain about 60% of the variation in the overall score. Part of that reflects real differences in conditions, but part is how the inputs are built. CDC chronic-disease figures are modeled partly from demographics (they correlate −0.85 with bachelor's-degree share), poverty and unemployment each appear directly, and police-reported crime depends partly on where reporting and enforcement are concentrated. Read a low score as "these measures are less favorable here," not as a judgment of the people who live there.
Mainly because St. Paul and Minneapolis publish genuinely comparable free, open data: matching crime/permit feeds, the same Census geography, similar OSM/CDC/MnDOT coverage. Adding a third city means finding one whose open data holds up to that same standard, so every district stays on the same 0–100 scale.
Built by Kennedy Johnson. More projects at kennedyjohnson.github.io.
Every district gets an overall Living Quality Score (0–100) built from five component indices, each also scored 0–100:
All 28 districts across both cities (17 St. Paul District Councils + 11 Minneapolis Communities) are normalized together in one pool, not city-by-city, so a "72" means the same thing in either city. Crime and permit counts are restricted to a shared trailing recent-years window before becoming rates, so the two cities' differing data-history lengths (e.g. one dataset back to 2014, another only to 2019) don't skew which city looks safer or more active than it really is.
Raw counts (crime incidents, permits, etc.) first become a rate per 1,000 residents, so a small district isn't unfairly penalized against a large one. Each rate is then normalized against every other district with a z-score, squashed through a logistic curve into 0–100, deliberately not simple min-max scaling, which would force the single most extreme district to exactly 0 or 100 and overstate how unusual it really is; z-score + logistic compresses outliers asymptotically, landing near but never exactly at the floor or ceiling. Metrics where a higher raw number is worse (crime rate, traffic volume, housing cost) are inverted first, so a lower rate produces a higher score.
Within a component, each metric's normalized value is weight-averaged into that component's index, and the five indices are then blended into the overall Living Quality Score using the equal 20/20/20/20/20 weights above, chosen deliberately to avoid imposing a subjective preference for any one dimension of quality of life. A district-by-district breakdown of all five is always shown alongside the overall score, so you can weigh them differently yourself.
Clicking anywhere on the map or searching an address scores that point's 1-mile radius instead of the enclosing district, comparing it against a dense grid of thousands of other 1-mile circles sampled across the same city, not whole-district averages, which are diluted by sparse edges and parks a small circle would never actually cover. Each source is normalized independently before blending, the same z-score treatment district scores get. Safety blends the same four signals as the district-level index (crime, pedestrian/cyclist crashes, tract-level chronic disease burden, natural hazard risk); the one exclusion is the Walk/Bike Score's distance decay and any metric only available at the district level (its weight redistributed among the rest). A clicked/searched point gets its name from a local index of named OpenStreetMap places built at data-refresh time rather than a live lookup on every click, falling back to Nominatim only for addresses not already in that index.
The map's "Map view: District / ZIP Code" toggle scores every ZIP code the same way, but normalized in its own separate pool: ZIP vs. ZIP, never against districts or split per city, so a ZIP's score/grade is only meaningful relative to other ZIPs, the same way a district score only means something relative to other districts; the two pools aren't on a directly comparable 0–100 scale. Only ZIPs with at least 98% of their area inside St. Paul or Minneapolis's combined boundaries are included (27 of the ~49 ZCTAs touching the metro), since a ZIP that only clips a corner of a district would divide a real population by an artificially small incident count from a source that doesn't cover most of that ZIP's actual population: a large-majority-area test rather than exact containment, since real boundaries rarely align to the pixel.
Every metric that feeds the district score is recomputed at ZIP granularity from the same underlying sources (including household affordability figures and the Walk/Bike Score), rather than being interpolated or copied down from the enclosing district. The one gap: multi-year trend charts (crime, permits, affordability over time) aren't computed at ZIP granularity, so a ZIP's detail panel shows its current metrics and score but no historical trend line.