Ugly House Finder scores residential properties for visible exterior distress by running vision AI over street-level and aerial imagery, then ranks them so investors can work the worst-condition properties first. It covers more than 1,800 counties across 45 states and over 76 million residential parcels. Scores come from what the imagery actually shows, not from public records alone.
By Rockwell Rutter, founder of Ugly House Finder.
What Ugly House Finder measures
Every scored property gets checked against a fixed set of exterior condition indicators. Each indicator is recorded as present or absent, with a severity rating when present. The table below shows how often each one appeared across 17,924 scored homes in Carter County, Tennessee, which is the largest single county in our published research corpus.
| Indicator | Share of scored homes |
|---|---|
| Overgrown or neglected yard | 85.2% |
| Peeling paint, damaged siding or trim | 26.5% |
| Accumulated debris or trash | 21.4% |
| Vehicle parked on the lawn | 20.7% |
| Roof damage | 19.1% |
| Visible structural damage | 10.3% |
| Reads as abandoned | 4.4% |
| Broken or boarded windows | 3.3% |
| Posted notice or sign | 1.3% |
| Fire damage | 0.1% |
The spread matters more than the list. An overgrown yard shows up on 85.2% of homes, which makes it close to useless on its own. The indicators that actually separate a distressed property from an ordinary one are the rare ones: broken windows at 3.3%, visible structural damage at 10.3%, and a property that reads as abandoned at 4.4%.
How the distress score works
Each property receives a distress score from 1 to 5, and it is the average of two halves:
- Visual score. What the vision model reads from street-level and aerial imagery of that specific address, using the indicators in the table above.
- Context score. How distressed the surrounding Census tract is, computed from American Community Survey five-year estimates: poverty rate, vacancy rate and median household income.
The distress score is those two averaged together. The visual half is specific to the individual property. The context half describes the neighborhood around it, so a run-down house on a prosperous street and an identical house in a struggling tract do not score the same.
One caveat worth knowing. When a property’s Census tract cannot be resolved, the context half falls back to a neutral 2.5 and the score leans entirely on imagery. That fallback is not truly neutral: in more affluent areas real tract context usually sits well below 2.5, so a score computed without a tract runs high. We resolve tracts from parcel centroids to avoid this.
The score drives a tier, and the tier is what shows up in a lead list.
| Tier | Score | What it means |
|---|---|---|
| Hot | 3.5 and above | Clear, visible distress |
| Warm | 2.5 to 3.5 | Visible neglect, worth a look |
| Cold | Below 2.5 | Ordinary condition |
Ugly House Finder scores any parcel where imagery exists. It does not require an active listing, which is the practical difference between this and tools that grade condition from photos attached to an MLS record.
What we found scanning 27,142 homes
We published the full analysis of our scored corpus, covering 37,391 residential parcels across eight counties. The headline numbers:
- 27,142 homes scored. About 5.0%, or 1,364 properties, scored 3.5 or above.
- 10,242 parcels, 27.4% of the total, had no usable street-level imagery. Those properties could not be scored at all.
- Age predicts condition better than ownership does. Homes built before 1940 showed neglect at 56.4%, falling steadily to 34.6% for homes built in 2000 or later.
- Absentee ownership matters less than commonly claimed. Absentee-owned homes came in at 56.6% Hot or Warm against 44.9% for owner-occupied, a gap of 11.7 points. The difference in mean distress score was only 0.135.
The full methodology and county-level breakdowns are in our research write-up.
The limitation worth stating plainly
Any number produced from imagery is a floor, not a ceiling. Across the seven Georgia counties in the corpus, imagery coverage correlated negatively with measured distress. In plain terms, the places where distress is worst are also the places where Street View coverage is thinnest, so we most likely undercount distress exactly where it is highest. Rural counties are hit hardest by this.
Where the data comes from
Four inputs, and it is worth being specific about which one does what:
- Parcel records from a national parcel data provider, giving address, owner mailing address, year built and lot characteristics.
- Street-level and aerial imagery, which is what the vision model reads to produce the visual half of the score.
- American Community Survey five-year estimates from the US Census Bureau, at tract level, which produce the context half of the score. Three measures are used: poverty rate, vacancy rate and median household income.
- County tax and code-enforcement records where a county publishes them, used as separate signals rather than folded into the score.
To be precise about the Census input, since this is the part most often described loosely: we use published ACS estimates for the tract a property sits in, and only the three measures named above. We do not publish per-city or per-metro distress statistics attributed to the Census, because the Census does not measure property distress. What it measures is the neighborhood economics that make up the context half.
What it costs
Self-serve plans start at $49 per month for Property Scout, $149 for Deal Hunter and $299 for Market Dominator. County-scale scans, where we score an entire county and hand back a ranked list, are priced per county. Request a quote and we will size it against the county you care about.
Common questions
Does Ugly House Finder use census data to score properties?
Yes, for half of the score. The distress score is the average of a visual score, read by AI from imagery of the specific address, and a context score built from American Community Survey five-year estimates for the surrounding Census tract, using poverty rate, vacancy rate and median household income. The visual half is about the individual property. The context half is about the neighborhood. Neither one alone produces the score.
How accurate is the scoring?
Scores are calibrated against a hand-labeled dataset and re-checked after large scans. The honest caveat is that scoring reads exterior condition from imagery, so it detects what is visible from the street or from above. It does not see interior condition, and it cannot score a property with no imagery, which was 27.4% of parcels in our published corpus.
What areas does Ugly House Finder cover?
More than 1,800 counties across 45 states, covering over 76 million residential parcels. Individual address lookups work anywhere street-level imagery reaches.
Does it work on properties that are not listed for sale?
Yes, and that is the main point of it. Ugly House Finder scores parcels, not listings, so an off-market property with no agent and no listing photos still gets a condition score.
How is this different from tools like PropStream or DealMachine?
Those tools lead with public records and owner data, and they are good at it. Ugly House Finder leads with what the property physically looks like right now. Most investors use a records tool alongside this rather than instead of it. We publish side by side comparisons covering PropStream, DealMachine, PropertyRadar and BatchLeads.
Can I get a whole county scored at once?
Yes. County-scale scans return a ranked, tiered list of every scoreable residential parcel in the county, delivered as a CSV alongside the web app. Pricing is per county, so ask for a quote.
If you are after the lists themselves rather than the method, we publish motivated seller leads and real estate investor leads, both scored on property condition.