Skip to main content
Let’s Live There

Find your fit. Know the trade-offs.

Methodology

How we score places

Every number on Let's Live There comes from a public dataset or is labeled as an estimate. This page lists each one: where it comes from, how it becomes a 0 to 100 score, and how sure we are of it.

Overview

What this is

Let's Live There helps you choose where to live. You say how much each thing matters to you, from 0 (ignore it) to 5 (essential), across 42 dimensions in nine categories. We score every one of 6,398 U.S. places from 0 to 100 on each dimension, and rank the places by how well they fit your weights.

Match score = sum of (your weight × the place’s score) ÷ sum of your weights

It is a weighted average over only the dimensions you weighted. Nothing else decides the order. Four things adjust it, and the first two are your own choices:

  • Hard filters (states, regions, city size, must-haves, “Only places with strong data”, and “Avoid places with elevated natural-hazard risk”) remove places entirely; they never reorder the rest. The hazard filter covers nine FEMA hazards: a place is excluded when a hazard you tick is rated Relatively High or Very High for its county, and a hazard FEMA does not rate for a place never excludes it.
  • A region you mark as preferred adds 15% to the match score of its places, and it is shown as “Region boost” in the score breakdown.
  • Places with limited data have their scores pulled toward the middle before weighting (see Data confidence).
  • If you add your own criteria, their AI ratings join the weighted average (see AI features).

Advertising never affects scores, rankings or search. The scoring code has no access to ads, and ads cannot pay to move a place up.

Scores

How scores are built

All scores run from 0 to 100 and higher is always better, so “Low crime” and “Low humidity” are the good direction. Most are a figure placed on a national range: the best realistic value scores near 100, the worst near 0. Some are percentiles among U.S. counties, which stops one extreme value from stretching the scale.

Measured30 of 42
Computed from a public dataset that covers essentially every place.
Partly measured11 of 42
Measured where a source reaches the place; a hand estimate or an unknown elsewhere, or a mix with a hand-assembled part.
Estimated1 of 42
Hand-assembled for every original place, with no automated source behind it yet.

Each entry also says what a place added from the Census list (the 25,000+ expansion, the 10,000 to 24,999 second tier and the optional 2,500 to 9,999 small-town tier) gets: a measured score, a partly measured one, or “unknown”. 1 dimensions are unknown for those places because no automated source exists yet; they are never guessed.

Cost & Taxes

How far everyday prices (goods, housing rents, utilities and services) sit below the U.S. average.

Source
Bureau of Economic Analysis, Regional Price Parities (all items), by metropolitan area and by state nonmetropolitan portion (2024, released February 2026); Census Bureau CBSA delineation (which county belongs to which metro area) (July 2023)
How
100 minus the all-items price level placed on an 80 to 120 range, rounded: 100 on the price level is the U.S. average, so an average place scores 50 (San Francisco-Oakland-Fremont at 115.6 scores 11, Monroe, Louisiana at 83.6 scores 91). A place in a metropolitan statistical area takes its metro area’s figure; any other place, including every micropolitan county, takes its state’s nonmetropolitan figure.
Level
Metro area, or the state’s non-metro portion
Note
Prices are measured for the whole metro area, so places in the same metro share the value; the housing cost dimension carries the difference between them. 6 original places (Sussex County, Delaware: Milford, Seaford, Lewes, Rehoboth Beach, Georgetown and Laurel) have no area in the BEA tables and keep a hand estimate.
Newer places
Measured. Except 1 new place with no area in the BEA tables, which is unknown.

How low the typical home price is.

Source
Census ACS table B25077, median home value (2020-2024 5-year)
How
On a log scale: $150,000 or less scores 100, about $360,000 (the U.S. median) scores 62, $1.5 million or more scores 0.
Level
City, or the county’s figure where the Census publishes no place value
Note
Owner-reported values, so lower than asking prices where the market moved fast. Where the Census publishes no value for the place itself, the county’s ACS figure stands in and the place page says so: 10 places (1 original, 9 new).
Newer places
Measured.

How low the state taxes you pay as a resident are: the state income tax (two-thirds of the score) and the sales tax, state plus the average local add-on (one-third).

Source
Each state’s own revenue department (top marginal single-filer rate), one primary document per state; the Tax Foundation’s State Individual Income Tax Rates and Brackets is kept as a cross-check only (tax year 2025); Tax Foundation, State and Local Sales Tax Rates (state rate plus population-weighted average local rate) (rates as of January 1, 2025)
How
Two-thirds of the income tax score plus one-third of the sales tax score, rounded. The income tax score is 100 minus the state top marginal rate placed on a 0% to 13.3% range, so no income tax scores 100. The sales tax score is 100 minus the state rate plus the average local rate placed on a 0% to 10% range, so no sales tax scores 100 and 10% or more scores 0. A state is marked none, flat or graduated for income tax.
Level
State
Note
The same two state figures for every place in a state, taken from the published tables, so two cities in one state always tie. The local part of the sales tax is a state-wide population-weighted average, not your city or county rate. Local income taxes and property taxes are not in this score (property tax has its own), and Washington counts as no income tax because it taxes only capital gains.
Newer places
Measured.

How little property tax a homeowner pays relative to home value.

Source
Census ACS tables B25103 and B25077 (2020-2024 5-year)
How
Median real-estate tax over median home value, placed on a 0.3% to 2.5% range and inverted.
Level
City
Note
A ratio of two medians, so approximate.
Newer places
Measured.

Work & Economy

How low unemployment is.

Source
BLS Local Area Unemployment Statistics (via USDA ERS) (2023 annual average)
How
100 minus the county unemployment rate placed on a 1.8% to 9% range.
Level
County
Note
A county rate describes the county labor market, not the city.
Newer places
Measured.

How well connected a place is for working from home.

Source
Census ACS table B28002, household broadband subscriptions (2020-2024 5-year); Coworking space density (hand-assembled)
How
60% the broadband subscription rate placed on a 70% to 95% range, plus 40% the percentile of coworking spaces per capita.
Level
City
Note
The broadband half is measured and is a take-up rate, not speed; the coworking half is by hand.
Newer places
Partly measured. The coworking half is unknown, so it takes the median of the original places (48.7).

How fast the population has grown.

Source
Census Population Estimates (April 2020 to July 2024)
How
Percent change in population since 2020, placed on a -4% to +12% range.
Level
City (county for census-designated places)
Newer places
Measured.

Climate & Weather

Warm climate

Measured

How warm the year is, with winter weighted above summer.

Source
NOAA U.S. Climate Normals, monthly, nearest station (1991-2020)
How
60% the January average high plus 40% the July average high, placed on a 35 to 80 degree F range.
Level
Nearest weather station (median 4 miles, at most 40)
Note
A station describes the station: valleys, coasts and hills can differ by several degrees.
Newer places
Measured.

Cool climate

Measured

How cool the year is (the mirror of warm climate).

Source
NOAA U.S. Climate Normals (1991-2020)
How
100 minus the warm climate score, exactly.
Level
Nearest weather station
Newer places
Measured.

How distinct summer and winter are.

Source
NOAA U.S. Climate Normals (1991-2020)
How
100 when July is 45 to 55 degrees F warmer than January, falling in a straight line to 0 at 30 degrees outside that band.
Level
Nearest weather station
Note
Counts temperature spread only, not snow, humidity or shoulder seasons.
Newer places
Measured.

Abundant sunshine

Partly measured

How much of the daylight the sun is out.

Source
NOAA Comparative Climatic Data, percent of possible sunshine (the measured source) (period of record, mostly 1965-1983); NOAA Comparative Climatic Data, mean number of clear, partly cloudy and cloudy days (the fallback, converted to an estimated percent of possible sunshine) (period of record, mostly 1965-2017)
How
Annual percent of possible sunshine placed on a 40% to 85% range. Where no sunshine station is within 50 miles, the nearest cloudiness station within 50 miles is used instead and its clear, partly cloudy and cloudy days are converted by one fixed rule, 111% × clear + 77% × partly cloudy + 13% × cloudy (weighted by days), fitted by least squares on the 138 stations that are in both tables (correlation 0.94, typical error about 3 points). A place on the fallback shows its value as an estimate.
Level
Nearest station within 50 miles (about 140 sunshine stations, about 250 cloudiness stations)
Note
About 85% of places have a value: 4,015 from a sunshine station and 1,410 estimated from cloudiness (for example every place around New York, which has no sunshine station). The 260 original places beyond 50 miles of both keep a hand score and the 713 new ones are unknown.
Newer places
Partly measured. Measured for 3,658 of the 5,649 new places and estimated from cloudiness for 1,278; the other 713, more than 50 miles from any weather station in either table, are unknown.

Low humidity

Partly measured

How dry the summer air feels.

Source
NOAA hourly normals, July dew point (1991-2020)
How
100 minus the July mean dew point placed on a 40 to 72 degree F range.
Level
Nearest hourly station within 60 miles (median 21 miles)
Note
Winter dryness is not in it; 82 original places without a station keep a hand score.
Newer places
Partly measured. Unknown for 175 new places.

How long the shortest days are.

Source
Computed from latitude (December solstice day length) (astronomical)
How
Sunrise-to-sunset hours at the December solstice placed on an 8.5 to 10.6 hour range.
Level
The place’s latitude
Note
Day length only, not cloud cover.
Newer places
Measured.

Nature & Environment

Outdoor recreation

Partly measured

How much public land, mountains, coast and sun are in reach.

Source
USGS Protected Areas Database (PAD-US) (version 4.0); Natural Earth 1:10m coastline and Great Lakes (distance to the coast) (fetched 2026-10-25); NOAA Comparative Climatic Data, percent of possible sunshine, or its cloudiness-based estimate (the Sunshine score) (period of record, mostly 1965-1983); USGS 3D Elevation Program (3DEP) Elevation Point Query Service (terrain relief: the mountain term) (fetched 2026-10-25)
How
40% public land access, 20% the mountain term, 20% near the coast (100 if the ocean or a Great Lakes shore is within 30 miles, else 40), 20% the Sunshine score (percent of possible sunshine placed on a 40% to 85% range). The mountain term is terrain relief from USGS 3DEP elevations: the highest of eight points on a ring 20 miles out, or 60% of the highest of eight points on a ring 40 miles out if that is larger, minus the place’s own elevation, placed on a 0 to 4,000 ft range (0 to 100). A place counts as near mountains when the land rises at least 2,000 ft within 20 miles or at least 3,500 ft within 40 miles.
Level
County, place coordinates, nearest sunshine station, terrain around the place
Note
The public-land, coast, sunshine and mountain parts are measured (sunshine is estimated from cloudiness for places with no sunshine station within 50 miles, and a place beyond both keeps its hand sunshine score; a place whose terrain the service could not measure keeps the hand mountain flag). The relief is a straight-line sample of 16 points around the place, not a survey of the ranges.
Newer places
Partly measured. Public land, the coast and terrain relief are measured, and so is sunshine (estimated from cloudiness for 1,278) except for 713 new places more than 50 miles from any weather station, which take the median Sunshine score (40); a place the terrain service could not measure takes the median mountain term (30).

How much of the county is open public land.

Source
USGS Protected Areas Database (PAD-US), open-access acres (version 4.0)
How
On a log scale: 2% of the county or less scores 0, 60% or more scores 100.
Level
County
Note
States that record access incompletely read low (much of New England).
Newer places
Measured.

How close the ocean, the Gulf or a Great Lakes shore is.

Source
Natural Earth 1:10m coastline and Great Lakes shorelines (public domain) (fetched 2026-10-25)
How
100 × (1 − √(distance ÷ 300 miles)), rounded: the straight-line distance from the place’s coordinates to the nearest ocean, Gulf or Great Lakes shore, on a square-root scale that reaches 0 at 300 miles. A day trip to the shore matters up to a few hours’ drive, and the first miles matter most: 10 miles scores 82, 50 miles 59, 150 miles 29, 300 miles or more 0. “Must be near the ocean or a Great Lake” in the survey’s must-haves means within 30 miles. The ocean, the Gulf and the Great Lakes count alike. When you choose “Ocean” or “Great Lake” under the question, the same formula is applied to the distance to that kind of water only (the nearest ocean or Gulf shore, or the nearest Great Lakes shore), so a place on Lake Erie scores high for “Great Lake” and low for “Ocean” unless the sea is within a few hours’ drive. A place with no distance for the kind you chose counts as unknown (a neutral 50), never a guess.
Level
Place coordinates
Note
The coastline is generalized (a bay mouth can be cut across by a few miles). Bays and sounds count as coast; tidal river reaches (the St. Lawrence, the Hudson above Yonkers, the Sacramento-San Joaquin Delta and a few more) and lakes other than the five Great Lakes do not. The score does not say which water body, or whether there is a beach.
Newer places
Measured.

How clean the air is across the year.

Source
EPA Air Quality System, annual AQI by county, median (2024 (else 2023))
How
100 minus the median AQI placed on a 20 to 90 range.
Level
County, else the nearest monitored county within 75 miles
Note
A place with no monitor in its own county borrows the nearest one within 75 miles; the city page says which. 9 original places with no monitored county in range keep a hand value.
Newer places
Measured. Except 43 new places with no monitored county in range, which are unknown.

How many county residents are on water systems with no recent health-based violation (a higher score means more residents on clean systems).

Source
EPA Safe Drinking Water Information System (violations 2023-2025)
How
A concern index is 70% the share of residents on a system with a recent health-based violation plus 30% the violation rate per 10,000 served. The score is the share of counties with a worse index, so a county with no violations scores 100 and a higher score means cleaner water.
Level
County
Note
Counts regulatory violations, not taste, hardness or lead in one home; small systems report less. 2 original places keep a hand score.
Newer places
Measured. Except 12 new places, which are unknown.

Safety & Health

Low crime rate

Partly measured

How low reported crime is.

Source
FBI Crime Data API (summarized agency offense counts) (2025)
How
60% (100 minus violent crimes per 100,000 residents, placed on a 0 to 1,200 range) plus 40% (100 minus property crimes per 100,000, placed on a 0 to 5,000 range). Violent crime is murder, rape, robbery and aggravated assault; property crime is burglary, larceny-theft, motor vehicle theft and arson. The two ends are the 99th percentile of the measured rates, so a rate at or above one scores 0 on that part.
Level
Police agency (city) or county sheriff
Note
Measured for 5,458 of 6,398 places: 3,862 from the place’s own police agency and 1,596 at county level (the county sheriff stands in where the FBI lists no police agency of the place’s name, and a metropolitan department that covers far more than the place counts as county level too; the city page says which). An agency that did not report all 12 months of 2025 to the FBI has no figure, never a zero. 940 places have none (no matching agency, an agency that did not report the whole year, or one not fetched yet): the 49 original places among them keep the older hand score and the newer ones are unknown.
Newer places
Partly measured. Unknown for 891 places whose police agency did not report the whole year, could not be matched, or has not been fetched yet.

How little of what is built there natural hazards destroy each year.

Source
FEMA National Risk Index v1.20, expected annual building loss (Dec 2025)
How
100 minus the percentile of expected annual building loss per dollar of building value among every U.S. county.
Level
County
Note
Buildings only (not crops or lives), so it understates drought.
Newer places
Measured.

How many doctors and hospital beds there are, and how near a well-rated hospital is.

Source
HRSA Area Health Resources File (physicians, hospital beds) (2024-2025); CMS Hospital General Information (emergency hospitals, star ratings) (current)
How
40% physicians per 100,000 (percentile), 20% hospital beds per 100,000 (percentile), 40% distance to the nearest 3-star-or-better emergency hospital (5 miles or less is 100, 60 or more is 0).
Level
County rates, straight-line distance
Note
County rates flatter a county with a medical center and everyone around it.
Newer places
Measured.

Getting Around

Walkability

Measured

How walkable the built environment is.

Source
EPA Smart Location Database v3, National Walkability Index (2018 block groups)
How
The index (1 to 20) population-averaged over the place’s block groups and placed on a 4 to 17 range.
Level
Block groups within the city
Note
4 original places (the small western towns Eagle and Silverton, Colorado; Joseph, Oregon; and Pinedale, Wyoming) are unknown: the block groups that cover them lie mostly outside their limits, and we no longer substitute an estimate.
Newer places
Measured. Except 37 places, which are unknown: the block groups that cover them lie mostly outside their limits.

How often transit runs near homes.

Source
EPA Smart Location Database v3 (2018 block groups)
How
Evening-peak trips per hour within a quarter mile on a log scale: 10 trips an hour scores 50, 120 or more scores 100, none scores 0.
Level
Block groups within the city
Note
The same 4 original places as walkability hold no value and read unknown (a neutral 50 in the score); a place with no service near homes scores 0 (417 of the 749 original places).
Newer places
Measured. Except the same 37 places as walkability, which are unknown.

How short the daily commute is.

Source
Census ACS tables B08013 and B08301, mean travel time to work (2020-2024 5-year); Texas A&M Urban Mobility Report, annual hours of delay per commuter (2025 (year 2024))
How
Commute time placed on a 15 to 40 minute range and inverted; where the urban area is covered, averaged half and half with delay hours on a 10 to 80 range.
Level
City commute; urban-area congestion
Note
Congestion is measured for the urban area and only where covered (313 of the 749 original places); elsewhere commute time alone.
Newer places
Measured.

How near scheduled commercial air service is.

Source
OurAirports, scheduled-service large and medium airports (current)
How
40% distance to the nearest scheduled airport (10 miles or less is best, 100 or more worst), 40% distance to the nearest large airport (15 to 150 miles), 20% large plus half the medium airports within 60 miles.
Level
Straight-line miles from the place
Note
The large and medium split is OurAirports’ own, not the FAA’s hub size.
Newer places
Measured.

How near the core of a metro of about a million people is.

Source
Census metro area estimates, 55 metro cores (hand-compiled list) (Vintage 2024)
How
100 minus the straight-line miles to the nearest of those cores placed on a 15 to 250 mile range.
Level
Straight-line miles from the place
Note
Driving is typically 20 to 30 percent longer.
Newer places
Measured.

Amenities & Entertainment

Arts & dining

Partly measured

How many arts venues, museums, performing-arts companies and restaurants are within reach, with a nod to major-league and big-time college sports.

Source
Census County Business Patterns, NAICS 71 and 722 (arts and restaurant counts), plus the venue counts behind “Performing arts & museums” (2022); Curated list of major-league teams (original places: hand flags) (hand-curated, 2025 seasons); NCES IPEDS campus coordinates of the 68 Power-4 schools (ACC, Big Ten, Big 12, SEC), 2025 membership (IPEDS 2024, membership as of the 2025 season)
How
50% arts-venue count percentile plus 30% restaurant count percentile (log scale among counties) plus 15% the Performing arts & museums score plus 5% for having a major-league team or a Power-4 college program (a school of the ACC, Big Ten, Big 12 or SEC within 30 miles). Where the Census withholds a county’s venue count, the 15% takes the median Performing arts & museums score of the places that have one.
Level
County
Note
The counts are measured; the 5% sports term is measured for Power-4 college programs (campus within 30 miles) and comes from team lists for major-league teams; the Performing arts & museums part is the median where its count is withheld. Counts include gyms, golf courses and marinas as well as theaters.
Newer places
Partly measured. The sports term uses a curated list of league teams within 25 miles and the measured Power-4 flag, the same as for the original places.

How many performing-arts companies, event promoters, museums and historic sites there are in the county.

Source
Census County Business Patterns, NAICS groups 7111 (performing arts companies: theater, dance, musical groups and artists), 7113 (promoters of performing arts, sports and similar events) and 7121 (museums, historical sites, zoos and nature parks) (2022)
How
The county’s establishments in those three groups, added together, on a log scale from none (0) to about 250 (100, roughly the 99th percentile of counties): a single establishment scores 13, ten score 43. Absolute counts, not per resident, so a small tourist town is not flattered.
Level
County
Note
Counts employer establishments, not seats or quality, and government-run museums are not counted; a festival orchestra with no paid staff does not appear. The Census lists some establishments in its broader rows without placing them in a group; where those outnumber the counted ones, the county’s count is withheld (not zero) and the place shows “Unknown”, or keeps its older hand estimate: 528 of 6,398 places. Elsewhere a small county’s count can still be a little low.
Newer places
Partly measured. The count is withheld for 436 of the 5,649 new places: unknown.

Golf course access

Partly measured

How many golf courses there are per resident of the county.

Source
Census County Business Patterns, NAICS 713910 (golf courses and country clubs), with county population from the Census Population Estimates (2022)
How
Golf courses and country clubs per 100,000 county residents, on a log scale from none (0) to 20 or more (100): about 5 per 100,000 scores 59.
Level
County
Note
Counts establishments, not holes, quality or public access, per resident, so a big county with many courses and many people (Maricopa, Arizona) scores modestly. The Census withholds small counts: where it lists no golf row but its recreation group has establishments it does not itemize, the county’s golf count is withheld (not zero), and the place shows “Unknown”, or keeps its older hand estimate. That is 854 of 6,398 places.
Newer places
Partly measured. The count is withheld for 634 of the 5,649 new places: unknown.

Community & People

Small town feel

Partly measured

How small the place is.

Source
Census Population Estimates (July 2024)
How
100 minus population placed on a national range of roughly 30,000 to 8.5 million.
Level
City
Note
Hand-calibrated from an older population for the original places.
Newer places
Measured. A fitted line on the log of the 2024 population, clamped to 0-100.

Big city feel

Partly measured

How big the place is.

Source
Census Population Estimates (July 2024)
How
Population placed on the same range, the exact inverse of small-town feel.
Level
City
Note
Hand-calibrated from an older population for the original places.
Newer places
Measured. A fitted line on the log of the 2024 population, clamped to 0-100.

How young the typical resident is.

Source
Census ACS table B01002, median age (2020-2024 5-year)
How
100 minus the median age placed on a 28 to 52 year range.
Level
City
Newer places
Measured.

How old the typical resident is.

Source
Census ACS table B01002, median age (2020-2024 5-year)
How
100 minus the younger-community score, exactly.
Level
City
Newer places
Measured.

How evenly residents are spread across seven racial and ethnic groups.

Source
Census ACS table B03002 (2020-2024 5-year)
How
Shannon entropy of the seven group shares divided by its maximum (ln 7), times 100.
Level
City
Newer places
Measured.

How many adults hold a bachelor’s degree or more.

Source
Census ACS table B15003 (2020-2024 5-year)
How
Share of adults 25+ with a bachelor’s or higher, placed on a 12% to 65% range.
Level
City
Newer places
Measured.

How Republican the county voted in 2024.

Source
County-level presidential results (tonmcg/US_County_Level_Election_Results_08-24) (2024)
How
Republican share of all votes (third parties stay in the total) placed on a 20% to 80% range.
Level
County
Note
Alaska reports by legislative district, so its 11 original places keep a hand estimate.
Newer places
Measured. Except 31 Alaska places, which are unknown.

How Democratic-leaning the county voted in 2024 (the mirror of conservative).

Source
County-level presidential results (tonmcg/US_County_Level_Election_Results_08-24) (2024)
How
100 minus the conservative-community score, exactly.
Level
County
Note
Alaska reports by legislative district, so its 11 original places keep a hand estimate.
Newer places
Measured. Except 31 Alaska places, which are unknown.

How many religious congregations there are per resident of the county, of any faith (not the share of any one religion).

Source
U.S. Religion Census: Religious Congregations and Membership Study, county file (Association of Statisticians of American Religious Bodies, via the Association of Religion Data Archives), with the county population in the same file (2020)
How
Congregations per 100,000 county residents, on a log scale from 45 or fewer (0) to 300 or more (100), about the 1st and 99th percentile of places: 80 per 100,000 scores 30, 150 scores 63.
Level
County
Note
Counts the congregations of every faith that the 372 participating religious bodies reported, so independent congregations and groups that did not report are missing and every rate is a lower bound. It is a density of congregations, not of people: rural counties with many small congregations score higher than big metros (places under 50,000 average 41, places over 250,000 average 26). Connecticut is not in the file, which uses its eight historical counties rather than the nine planning regions the place data uses: 19 original places there keep their older hand estimate and 60 new ones are unknown.
Newer places
Measured. Except 60 Connecticut places, which are unknown.

Families & Schools

How good the schools look on inputs, and whether universities are nearby.

Source
NCES Common Core of Data (pupils and teachers) (2023-24); Census Annual Survey of School System Finances (F-33) (FY2024); NCES IPEDS (campus enrollment) (2023-24)
How
35% fewer pupils per teacher (percentile), 35% per-pupil spending (percentile), 30% students at 4-year campuses within 15 miles per resident on a log scale.
Level
County schools; campuses within 15 miles
Note
Inputs, not outcomes: test scores are not in it.
Newer places
Measured. Except 1 new place, which is unknown.

How many private schools there are per resident.

Source
NCES Private School Universe Survey (2023-24)
How
Private schools per 100,000 county residents placed on a 0 to 15 range.
Level
County
Newer places
Measured.

Affordable childcare

Partly measured

How low the price of infant child care is.

Source
U.S. DOL National Database of Childcare Prices, median weekly infant center price (2022)
How
100 minus the state median weekly price placed on a $120 to $400 range.
Level
State
Note
6 states have no 2022 figure: the 85 original places there keep a hand cost.
Newer places
Partly measured. Unknown for 570 new places in those states.

How few child care deserts there are.

Source
Our estimate of the share of the area that is a child care desert (original places only) (hand-assembled, 2024)
How
A hand-assigned score from our estimate of the share of the area that is a child care desert.
Level
County or metro
Note
No measured value reaches any place: no published dataset gives the share by place or county, so the original places carry our hand-assembled 2024 estimate.
Newer places
Unknown.

Trade-offs

Fit labels and the tension view

Next to each category you rated, a place gets one of three labels. A Strong fit is a category score of 70 or higher. A Likely compromise is a score below 50 on a category you rated Important or higher. Everything else is Mixed, including a low score on a category you rated below Important, since you told us it matters less. Where we have no data for a place (the score shows as unknown) it gets no label at all.

Show me the tension looks at your current top 30 and keeps the places that have at least one Strong fit and at least one Likely compromise among the categories you rated Important or higher: the places that nail something you care about while asking you to give up something else you care about. They are ordered by the biggest weighted shortfall (how far the compromise falls from 100, counted the way the scorer counts your slider), and ties go to the place with the stronger best fit. This view only reorders what you see: every place keeps its fit rank and its match score.

Trust

Data confidence

Every place carries a confidence from 0 to 1 that says how much of its data is measured for that place rather than borrowed or estimated. It starts from how precisely the Census data matched the place, how complete its fields are and how large it is.

The rule is relative

When a source reaches most places but misses one, that place loses 0.05 for the topic. When a source reaches fewer than half of all places, the gap is a limit of the whole dataset, so it costs nobody anything: it says nothing about one place compared with another, and charging for it would only lower every score equally.

The “Limited data” chip and the strong-data switch

A place below 0.8 gets a Limited data chip on its card and page. In the survey’s “Narrow your search”, Only places with strong data keeps only places at 0.8 or above.

Tempering toward the middle

Before weighting, each score of a place with less than full confidence is pulled toward 50:

adjusted = 50 + (score − 50) × (0.4 + 0.6 × confidence)

How confidence changes a score of 90 and a score of 10
ConfidenceA score of 90 counts asA score of 10 counts as
1.0090.010.0
0.9087.612.4
0.8085.214.8
0.6080.419.6

Stored scores never change; the card’s “Why this score” panel says when tempering was applied.

What “unknown” means

The 5,649 places added in the 25,000+ expansion, the 10,000 to 24,999 second tier and the small-town tier have no hand-assembled values, and we do not invent any. Where no public source reaches a place, the topic is marked unknown. Its score is a neutral 50, a placeholder that is shown as “unknown” instead of a bar, kept out of “why it matched” lists and out of “best in row”, and never presented as a measurement.

A score built from several parts replaces an unknown part with the median of that part among the original places, so a missing piece neither flatters nor punishes. Because gaps that affect the whole dataset cost no confidence, these places compete on the topics that are measured.

Coverage

Places covered

There are 6,398 places: the 749 in the original hand-assembled set plus 5,649 added from the Census Gazetteer (3,636 incorporated places and 2,013 CDPs). 3,644 of them are every incorporated place and census-designated place (CDP) with 10,000 or more residents (1,518 places of 25,000 or more and 2,126 places of 10,000 to 24,999). The other 2,005 are distinctive small towns of 2,500 to 9,999 residents, not every town in that range: a town is in when a measured rule holds, namely a four-year college of 1,000 or more students within 5 miles (1,235), a county in the top 5% for lodging per resident (81), the ocean or a Great Lakes shore within 3 miles (601), 3,000 feet of relief within 20 miles (264), or a county in the top 5% for performing-arts companies and museums per resident (193); a town can meet several. They are ranked only when you choose to include small towns, and their data is thinner. A CDP is a statistical area the Census Bureau draws for an unincorporated community. Population is the Census estimate for July 2024 (the ACS 5-year figure for CDPs, which have no estimate).

Photos on city pages come from Wikimedia contributors (Wikivoyage, Wikipedia and Wikimedia Commons). Only files under CC BY, CC BY-SA, CC0 or public-domain licenses are used. An AI judge scores each candidate against a visitors-bureau-brochure rubric and only high scorers are kept, at most eight per place. Each photo shows its author and license and links to its file page; a place with no qualifying photo shows none.

Nearby lists use straight-line distance. A result card lists places within about 40 miles that are also in your results, with your rank for each; a city page lists up to six places within 60 miles, nearest first.

Claude

AI features

  • What Claude writes

    Result cards carry no AI text: their fit and trade-off lines and their “Best for” lists (your priorities the place serves strongly, with the figures behind them) are computed from your answers and the data. Claude writes the personalized city summary, the ratings behind any criteria you type in your own words, and nothing else you read on a card. Photos for city pages are chosen by an AI judge from Wikimedia candidates.

  • Numbers come from the data

    Claude is given each place’s scores and raw figures and told to use only those. It never changes a score, and the ranking is computed before it writes a word.

  • Your own criteria are estimates

    A criterion you type (up to three) is rated by Claude for every place, from the measured facts we hold about it (for example business counts or registered nonprofits nearby) and from what Claude specifically knows about the place. A rating needs one or the other: where Claude has neither, the place shows “unknown” and counts as a neutral 50. Ratings are labeled “AI-estimated” everywhere they appear, name the datasets they rest on or show Claude’s reasoning, count for less when unsure, and are never presented as measurements.

  • Heritage criteria use Census data

    When a criterion names a nationality, ancestry or language (“strong German-speaking community”, “Korean community”, “Spanish-speaking neighborhood”, “immigrants”), the rating is also given what the Census measured for that place: the share of residents who report that ancestry or origin, who speak that language at home, and who were born abroad (American Community Survey 2020–2024, 5-year), next to the typical place. Margins are wide for small towns, and the rating is still labeled AI-estimated.

  • What happens to what you type

    The words are sent to Claude to rate places, and the ratings are cached by the wording so the next person who asks the same thing is not charged another model call. The text is not used for ad selection, for the other AI notes, or for anything else, and it is in a share link only if you share your results.

Sponsors

Advertising policy

Communities, agents and local institutions can pay for placement. These rules are enforced in code, not left to judgment:

  1. Never changes a score or an order

    Ads do not touch scores, rankings or search results. The code that scores places never receives anything about an ad.

  2. Native and labeled

    Every ad is a plain card marked “Sponsored” with the advertiser’s name, in a reserved slot, so nothing on the page jumps when it loads.

  3. Only where it is safe

    Never in the survey, the comparison view or the map, and never as a popup, interstitial, autoplay video or sticky overlay. One ad per type per page at most.

  4. No tracking

    Which ad you see depends only on the place on screen and your own priorities, which stay in your browser. No cookies, no third-party trackers; impressions and clicks go to our own log with no address, device or answers attached.

  5. “Why this ad?”

    Every ad has a button that states the real reason it was chosen, in plain words.

  6. Dismissible

    Every ad has a hide button. Your choice is remembered in your browser only, and an empty slot collapses rather than showing filler.

Want to advertise? See the advertiser page.

Privacy

What we keep

There are no accounts. Your answers stay in your browser unless you share them. A results share link carries your answers in the link itself, and a short share code keeps the same answers on our server until twelve months after it was last opened. A shared plan is stored on our server with its members’ chosen labels, priorities, reactions and suggestions, visible to anyone with its code, and deleted after twelve months without being opened. Our own first-party cookies hold random ids (a visitor id for 13 months, a session id for 30 minutes) so that we can count visits and apply the daily limits per browser; they are not linked to a name, an email, and our logs never hold your IP address, your answers or anything you type. What we keep, for how long, and who processes it is in the Privacy Policy, which is the version that counts.

Honesty

Known limitations

  • Crime counts offenses that police agencies reported to the FBI in 2025, per resident of the agency’s own population: not what residents feel, and not comparable between agencies that define or record offenses differently. A place has no figure (940 places) when its agency did not report all 12 months, no agency matched, or the agency has not been fetched yet, and for 1,596 places the figure is county-level (the sheriff’s, usually for the unincorporated area, or a metropolitan department covering far more than the place) and describes that agency’s jurisdiction, not the town.
  • Walkability and transit are missing for 41 small places (4 original, 37 newer) whose block groups lie mostly outside their limits: all of them read “Unknown” and score a neutral 50.
  • Sunshine reaches places with a weather station within 50 miles, as a measurement where the station records sunshine and as an estimate from cloudiness otherwise (about 3 points typical error); the 260 original places beyond both keep a hand estimate and 713 newer ones read “Unknown”.
  • Several flags are hand-assigned for the 749 original places and unknown for the rest: religious denominations (the survey’s denomination filter) and LGBTQ+ climate (pro teams come from a curated list). Power-4 college sports is measured for every place: a school of the ACC, Big Ten, Big 12 or SEC (2025 membership) with its campus within 30 miles, which drops one hand flag (Ithaca) and adds 57.
  • Religious community counts congregations of every faith that the participating religious bodies reported to the 2020 U.S. Religion Census in the place’s county: not people, and not any one faith. Independent congregations and groups that did not report are missing, so the rate is a floor that leans rural, and Connecticut has no figure (the Census uses its eight historical counties).
  • Distance to the coast is a straight line from the place’s coordinates to a generalized coastline, not a drive time, and not every shore is a beach: it counts bays and the Great Lakes, and leaves out tidal rivers, so a port up an estuary (Philadelphia, Portland, Sacramento) reads as farther than the nearest water.
  • Golf and cultural venues count employer establishments in the county (Census County Business Patterns): not seats, quality or programming. The Census leaves some detail out for small counties, so resort and small-town counties can read too low on culture, and golf is Unknown (or an older hand estimate) for the 854 places whose county count is withheld, and so are the venue counts for 528 places.
  • Many figures are county-level, so a suburb inherits its county, and a county rate cannot tell one town from the next.
  • Cost of living is measured for the whole metro area (or the state’s non-metro portion), not the city, so a pricey suburb and a cheap city in one metro score the same; home prices, which are measured per city, carry the difference.
  • School quality uses inputs (pupils per teacher, spending, nearby campuses), not test outcomes.

What is next

  • Fill in crime for the places whose police agency did not report the whole year (a second year of FBI data, or the state’s own reports).
  • Add state test proficiency to the schools score.
  • Replace the hand flags for the original places with measured ones, and fill them in for the newer places.

Help us fix it

Report a data problem

Found a figure that looks wrong, a place that is missing, or a source we should use? Open an issue on the project’s GitHub page and name the place and the dimension: github.com/watkynco/find-your-spot/issues (opens in a new tab).

How we score · Let's Live There