Real estate comps are recently closed sales of similar nearby properties, and the free sources worth using are county recorder and assessor records, the public portals' sold filters, and a licensed agent's MLS search. A sale qualifies as a comp when it is closed, within about half a mile, sold in the last six months, within roughly 20% of the subject's square footage, and comparable in condition and style. Everything else is a nearby house.
| Source | Cost | What it gives you | What it misses |
|---|---|---|---|
| County recorder / assessor | Free | Recorded sale price and date, legal description, ownership history | No condition, no photos, no square footage you'd trust; blank in non-disclosure states |
| Consumer portals, "sold" filter | Free | Sale price, photos from the listing, beds/baths/ft², days on market | Photos are of the house as sold, not as bought; some sales never appear |
| MLS, via an agent | Relationship | The complete record — condition remarks, concessions, price history, withdrawn listings | You need someone to run it, so it doesn't scale to screening |
| Data providers / APIs | Paid | Sold sets and automated valuations at volume, across markets | Coverage and freshness vary by county more than the marketing suggests |
| Appraisal | $400–$700 | A defensible, adjusted set a lender will accept | Too slow and too expensive for anything but a deal you're committed to |
In practice most investors use two: a free source to screen widely, and an MLS or provider set to confirm the handful of properties that survive. Screening on the expensive source is how people spend $300 a month looking at houses they were never going to buy.
Only a recorded transaction is evidence. Everything else is an intention.
Tighten to a quarter mile in dense grid neighborhoods where value changes street to street; loosen toward a mile only in rural areas where there is nothing closer. Never cross a school-district, municipal or flood-zone boundary to get a comp — those lines move price without changing anything you can see in a photo.
Three in a fast-moving market. Twelve only when the alternative is having no set at all.
Price per square foot is not linear: small houses sell for more per foot than large ones in the same neighborhood, because a share of the value is in the lot and the location rather than the building. Comparing a 1,100 ft² cottage to a 2,400 ft² colonial on a per-foot basis systematically overvalues the larger house.
A 1920s bungalow and a 1998 colonial two streets apart are not comparable however similar the numbers look. Nor is a townhouse to a detached, or a condo with a $600 monthly fee to one with $180.
This is the filter that changes with your purpose. Valuing a house as it stands? Comp against houses in similar condition. Estimating what it will be worth renovated? Comp against renovated sales. Mixing the two sets is the single largest source of wrong ARVs — see how to calculate ARV for the version aimed at the finished house.
Enter an address and Faro assembles the comparable sales around it — recorded transactions and current listings, weighted separately — and shows the value they support along with the confidence band.
A set of qualifying sales is the input, not the answer. Turning it into a defensible value means adjusting for the differences that remain, weighting the closest matches more heavily, and stating the range the set actually supports. That process is its own guide: how to run comps on a property.
And if what you're after is the value after renovation rather than today, the comp set is chosen differently from the start — how to calculate ARV covers that, and ARV vs market value covers why the two numbers diverge.
County recorder and assessor sites publish recorded sale prices in most US states and cost nothing. The major consumer portals let you filter map results to "sold" and cover the last 12–24 months. Both are genuinely free; the trade-off is that neither tells you the condition the house was in when it sold, which is the one field ARV work depends on most.
Six months is the working standard, and three is better in a market that is moving. Past twelve months you are no longer measuring today's market, you are measuring last year's and applying a correction you cannot verify. In a thin rural market you may have no choice — say so out loud and widen your range rather than quietly accepting an old sale as current.
Yes, and it is worth knowing whether you are in one. In roughly a dozen states — Texas, Utah, Kansas, Idaho, Missouri, Louisiana, Mississippi, New Mexico, North Dakota, Wyoming, Alaska, and parts of Montana among them — sale prices are not recorded publicly. There, MLS access through an agent or a data provider is not a convenience, it is the only reliable route to a sold price.
As context, not as evidence. Actives tell you what you will be competing against and how much inventory is chasing the same buyer, which is genuinely useful. They do not tell you what anyone paid. A value built on asking prices inherits every seller's optimism at once.
Three qualifying sales is the minimum for a median that means anything; five or six is comfortable. If you have to reach past a mile or past a year to get to three, that is information — thin data means a wider range and a more conservative offer, not a more confident number.
The full process on one address — search box, adjustments, weighting, and the range you should quote.
The five-step method, with a worked example and the adjustment table most investors skip.
Two different numbers on the same house — where the gap comes from and who uses which.
The signals worth screening on, and how to run them across a market rather than a listing at a time.
Faro pulls both recorded sales and active listings around a property, weights them, and shows the value they support.
Estimates for analysis and educational use only — not financial, investment, tax, or legal advice. Verify every number independently before making a purchase decision.