How many comps do I need for a reliable value?

Three to five solid comps beats ten loose ones. If you can't find at least three genuinely similar recent sales, widen the timeframe before you widen the neighborhood — crossing into a different school zone or subdivision distorts value more than an older sale does.

How far back can a comparable sale be?

Six months is the standard window; 12 months is usable in slow markets if you mentally adjust for the market's direction since the sale. Anything older is history, not evidence.

Can I trust the Zestimate or other online estimates?

Treat automated estimates as a sanity check, not an answer. They model typical condition from public data — they can't see the 1978 kitchen or the roof tarp. They're most useful for flagging when your own comp-based number is way off.

What if there are no good comps at all?

Thin-comp properties (rural, unusual size, mixed-use streets) are where bad buys happen. Widen the time window, bracket with the closest superior and inferior sales, and price conservatively — or pass. A deal you can't comp is a deal you can't confidently sell to a buyer either.

How to Comp a Property Without MLS Access

Agents comp with MLS access. Most wholesalers don’t have it — and don’t need it. Sold prices are public record in most of the country, and the skill that actually separates good valuations from bad ones isn’t database access. It’s discipline about which sales count.

What makes a comp valid

A comparable sale is evidence about your subject property only if it’s genuinely similar. The working filters:

Distance: within about 0.5 miles — and never across a major road, school boundary, or into a different subdivision “just to get more data.” Micro-location is the strongest force in comping.
Recency: sold within the last 6 months (stretch to 12 in slow markets, adjusting for drift).
Size: within roughly 20% of the subject’s square footage. A 1,100 sqft house tells you little about a 1,900 sqft house.
Type and era: same property type, same general construction decade. Ranch to ranch, not ranch to new-build townhome.
Bed/bath count: within one bedroom and one bathroom.

Every filter you loosen adds noise. Loosen recency first, size second, distance last.

Where to get sold data without MLS

County records. The assessor or recorder site in most counties publishes every deed transfer with price, date, and property characteristics. It’s the ground truth everything else scrapes.
Free listing sites. Zillow and Redfin both expose sold filters with photos. The prices come from the same public records — the photos are the bonus, because they let you judge condition at sale, which county data never shows.
Your buyers. Active flippers know what renovated houses sell for on specific streets, because they just sold one. Asking “what did yours on Elm close at?” is comping too.

The workflow: build the candidate list from records, then use listing-site photos to sort each sale into renovated vs dated. That condition sort is everything — it’s the difference between an as-is value and an after-repair value.

The adjustment process

You’ll never find a perfect twin. Adjust the comp’s sale price toward what it would have sold for as your subject:

Start with price per square foot. Divide each comp’s price by its size. Similar houses in one neighborhood usually cluster within a tight $/sqft band — that band is your value corridor.
Adjust for condition. A renovated comp against a dated subject: subtract roughly the cost of the renovation gap. This is the biggest and most-fumbled adjustment.
Adjust for concrete differences. Garage vs none, pool, lot size, an extra bath. Use modest, defensible numbers — thousands, not tens of thousands, for most single features.
Bracket. Aim for at least one comp slightly better than your subject and one slightly worse. Your value lives between them, and a bracketed number is one you can defend to any buyer.

The traps

Non-arms-length sales. Family transfers, foreclosure deeds, intra-LLC moves — they’re in the public record at prices that mean nothing. If a price looks bizarre for the street, check the deed type and the parties’ names.
Outlier worship. One sale at $310k on a street of $250k sales isn’t your comp; it’s a question to investigate (addition? double lot? bad data?).
Active listings as comps. Asking prices are hopes. Only closed sales are evidence — actives only tell you your ceiling and your competition.
Condition blindness. Comping a house with a 40-year-old kitchen against fully renovated sales is the single most common way wholesalers produce fantasy ARVs that buyers laugh at.

A worked example

Subject: 1,400 sqft, 3/1 ranch, dated but solid. Three qualifying sales nearby: 1,350 sqft renovated at $195,000 ($144/sqft), 1,500 sqft renovated at $210,000 ($140/sqft), 1,300 sqft partially updated at $172,000 ($132/sqft). The renovated band is roughly $140–144/sqft, so renovated, your subject supports about $196,000–$201,000 — call the ARV $198,000. As it sits, with a $35,000 renovation gap against those comps, the as-is picture is materially lower. (Which of those two numbers you use depends on the exit — see as-is value vs ARV.)

Doing this by hand takes 30–45 minutes per property when you’re careful. This exact pipeline — pulling the sales, applying the similarity filters, sorting by condition, and producing both the as-is value and the ARV — is what PropTitan automates on every address you look up. See the comps and ARV feature for how it works, and pressure-test any valuation by hand with the free ARV calculator.

The standard to hold yourself to

A comp job is done when you can hand a skeptical cash buyer three addresses and your adjustments, and they nod. If your value needs a story to survive scrutiny, it isn’t a value — it’s a wish. Comp like your buyer will check, because the good ones always do.