Percentile band25 / 50 / 75Outlier-resistant — one mispriced sale can't widen the band.
Minimum comps4Below this floor we suppress the range and render a thin-data state.
Split outputsARV + RentSeparate bands for value and rent — two distributions, two answers.
The problem03 reasons existing AVMs frustrate investors
01 · Black boxYou can't see what comparables drove the estimate.
Most AVMs hand back a single number. No comp set, no math, no way to argue with the answer if your gut says it's wrong.
02 · Confident on thin airA narrow range hides a small sample.
When only two recent sales exist in a submarket, a confident-looking band is the worst possible output — it's a guess wearing a suit.
03 · Mixed sale + rentRent and value get conflated.
Sale comps and rent comps are different distributions. Lumping them together blurs the two questions an investor actually needs answered.
How Domora does itPublic method · 5 steps
- Step 1
Pull comparable sales in the same market.
We start from the same-market candidate pool — the addresses already indexed in the property feed, filtered to recent sales with the facts we need.
- Step 2
Take a percentile band, not min / max.
For each comp we compute $/sqft. The 25th / 50th / 75th percentile of that distribution becomes the band. One mispriced sale can't widen the headline.
- Step 3
Anchor the band to the subject's size.
We multiply the per-sqft percentiles by the subject property's own square footage, so the ARV reflects this house — not the average house in the comp set.
- Step 4
Suppress below four comps.
If fewer than four comps carry the fact we need, the range is hidden. We render a thin-data state instead of fabricating a number from a sample too small to trust.
- Step 5
Show the evidence underneath.
The comps that drove the band sit directly beneath it as a table — address, $/sqft, beds/baths/sqft, similarity. Every estimate links back to its evidence.
Why this mattersReal estate is local, sparse, and noisy. A percentile band built from a visible comp set degrades gracefully — when the data is thin you see thin data, not a fake answer. When the data is rich you get a defensible band you can take into a conversation.
What you’ll see in productSample · Houston, TX
Live-style example of the Comps + AVM panel.
Estimated value (ARV)Estimated
$364,000 – $437,000
Midpoint $396,000
Median comparable sale price: $218/sqft × 1,820 sqft
Estimated monthly rentEstimated
$2,380 – $2,720
Midpoint $2,550
Based on 4 comparable rent estimates
Comparable properties
4 comps · the evidence behind the estimateSample comparable properties used to derive the estimated value and rent ranges, with price per square foot, bedrooms, bathrooms, living area, and similarity to the subject property.| Address | $/sqft | Bed / Bath | Sqft | Similarity |
|---|
| 4019 Almeda Rd Houston, TX 77004 | $211 | 3 / 2 | 1,840 | 94% |
| 3742 Wichita St Houston, TX 77004 | $215 | 3 / 2.5 | 1,905 | 91% |
| 4310 Charleston St Houston, TX 77004 | $213 | 3 / 2 | 1,750 | 88% |
| 3915 Burlington St Houston, TX 77004 | $215 | 4 / 2 | 1,980 | 84% |
Estimated ranges are model outputs derived from the comparable set above (25th–75th percentile band). They are not appraisals. This sample is for illustration; live property pages render this section against real comparable sales in the subject market.
- ARV bandThe estimated after-repair / market-value range, derived from the 25th–75th percentile of comparable $/sqft applied to the subject's square footage.
- Rent bandA separate 25th–75th percentile of comparable monthly rent estimates. Rent is a whole-unit figure, so it's not re-normalised by size.
- Evidence tableThe exact comps that drove the band — each row links back to its own property page in the live product.
- Thin-data stateBelow four comps, the band disappears and a thin-data note takes its place. The page never fabricates a confident answer from a small sample.
The Comps panel ships on every property page. Open a sourced record from Search, or sign up and import the addresses you’re already watching.