"Fair" is doing a lot of work in that phrase. A fair offer price isn't a vibe, and it isn't whatever number makes you feel like you didn't get ripped off — it's a specific dollar figure built from recent, comparable, sold properties, adjusted for how the house you're bidding on actually differs from them. That number is your starting point. What you do with it once you have it (offer above, at, or below) is a separate decision covered in a different article — this one is just about how to build the number itself, the way an appraiser or a good agent would, so nobody can hand-wave you into a price with a shrug and a Zestimate screenshot.

Step one: what actually counts as a "comp"

A comparable sale ("comp") is a recently sold home similar enough to the one you're bidding on that its sale price tells you something real about market value. Not every "similar" house qualifies. The criteria that matter, roughly in order of importance:

  • Sold, not listed. A comp has to be a closed sale, not an active listing or a pending one. List prices are what sellers hope to get; sold prices are what buyers actually paid. This is the single most common mistake people make when they eyeball a neighborhood — they average what's currently for sale instead of what actually closed (Redfin).
  • Nearby. Same neighborhood or, at minimum, the same "market area" — the zone buyers are actually cross-shopping. Redfin's rule of thumb for buyers is roughly a 1-mile radius in denser areas, stretched to as much as 5 miles in less dense or rural ones (Redfin). Appraisers work the same way but describe it more formally — Fannie Mae's guidelines say comps should ideally come from the subject property's own market area, and if an appraiser has to pull from a competing area, they're required to explain why and note the differences (Fannie Mae Selling Guide B4-1.3-08).
  • Recent. The tighter the window, the more reliable the comp. Buyers should lean toward sales in the last 3–6 months; appraisers are allowed to go back up to 12 months, and only reach further than that when comparable sales are genuinely scarce (rural areas, unusual property types) (Fannie Mae). In a market that's moving fast in either direction, a 10-month-old comp needs a time adjustment before you trust it (more on that below).
  • Similar size. A common appraisal-industry guideline keeps comps within about 20% of the subject's square footage — a 1,800 sq ft house shouldn't lean on a 3,000 sq ft comp down the street just because it's the only recent sale (JVM Lending).
  • Similar age and style. Built within roughly five years of the subject, same basic property type (don't compare a single-family house to a townhome or condo, even next door — different ownership structures and buyer pools price differently).
  • Similar condition. A gut-renovated kitchen and a house that hasn't been touched since 1994 are not the same asset even if the square footage matches exactly.

Fannie Mae's own standard requires a minimum of three closed comps to support an appraised value, and appraisers are expected to "bracket" — include at least one comp that's better than the subject and one that's worse — rather than cherry-picking three houses that all happen to support one number (Fannie Mae). That's a good discipline to borrow even if nobody's making you follow it: pull five or six candidate comps, not two.

Step two: where to actually pull the data

Your agent's MLS access. This is the real answer, and it's not close. The Multiple Listing Service has the most complete and current sold data — actual closing prices, days on market, seller concessions, sometimes even the terms of the deal — and most of it isn't visible to the public at all. A good buyer's agent should be building you a Comparative Market Analysis (CMA) as a matter of course before you write an offer. If they're not offering to, ask.

Redfin and Zillow's sold filters. The free, DIY option, and a genuinely useful one for a sanity check between agent conversations. Search the neighborhood, filter to "Sold," and set your own timeframe and radius. Two caveats: automated home-value estimates (Zillow's Zestimate, Redfin's Estimate) are not the same thing as a comps-based valuation — they're algorithmic models with real error margins. Zillow states a median error rate of roughly 1.9% for homes currently on the market and about 7% for off-market homes, meaning the further a home is from an active listing, the shakier the automated number gets. Treat the Zestimate as a starting search filter, not a comp itself.

County property/tax records. Recorder or assessor offices maintain the official record of what actually closed on a property, and it's usually free to search once you know the parcel. The catch: this only works in "disclosure states." Roughly a dozen states — including Texas, Kansas, Idaho, Mississippi, Montana, Louisiana, and Wyoming — don't require sale prices to be publicly recorded at all, which means county records (and even some consumer sites pulling from them) can be missing or wrong in those markets, and an agent's MLS access becomes the only reliable source (Redfin on non-disclosure states).

Your own eyes. Drive or walk the comps if you can, not just the subject property. A photo can hide a busy street, a sloped lot, or a neighbor's derelict shed. Comps look identical on paper more often than they actually are in person.

Step three: doing the adjustment math

Here's the part most buyers skip, and it's the part that actually turns a list of sold prices into a defensible number. Raw comp prices aren't your answer — you have to adjust each one for how it differs from the house you're bidding on.

The logic runs like this: for every meaningful way a comp is better than your subject property, subtract an estimated dollar value from that comp's sale price (to bring it down to what it would have sold for without that advantage). For every way it's worse, add. Appraisers formalize this with a technique called paired sales analysis — finding two nearly identical homes that differ in exactly one feature (say, a finished basement) and using the price difference between them to estimate what that one feature is worth in that specific market (McKissock). You won't have the data to run a full paired-sales study as a buyer, but you can approximate it with a good agent's local knowledge or by looking for near-identical pairs yourself in the sold data.

Categories that typically get adjusted, in the order appraisers tend to apply them — sale conditions first, then market-timing (appreciation or depreciation since the comp closed), then physical features (McKissock):

  • Lot size
  • Total finished square footage (gross living area)
  • Bedroom and bathroom count
  • Garage / covered parking
  • Basement (finished vs. unfinished, or none)
  • Age and overall condition
  • Notable updates (kitchen, roof, HVAC, windows)
  • Time between the comp's sale date and today, if the market has moved since

A simplified walkthrough. Say you're bidding on a 3-bed, 2-bath, 1,800 sq ft house with an unfinished basement, and you've pulled three comps within a half-mile that sold in the last four months:

Comp AComp BComp C
Sold price$410,000$395,000$430,000
Difference vs. subjectHas a finished basement (~$20k value, subtract)Nearly identical1 extra bathroom (~$8k value, subtract)
Adjusted price$390,000$395,000$422,000

That gives you an adjusted range of roughly $390,000–$422,000, clustering around $395,000–$400,000 — that's your comp-based number, before you decide anything about strategy. (The dollar figures above are illustrative only — what a bathroom or a finished basement is actually "worth" swings enormously by market and has to come from local paired-sales data or agent judgment, not a national rule of thumb.)

A blunter, faster version of this same idea is price-per-square-foot: divide each comp's adjusted sale price by its square footage, average the resulting $/sq ft across your comps, then multiply by your subject's square footage. It's useful as a gut check, but treat it as a floor for further analysis, not the final answer — $/sq ft flattens out real differences in lot size, age, and condition that a straight adjustment approach captures better.

Where this goes wrong

Using list prices as comps. Worth repeating because it's the most common error: an active or pending listing tells you what a seller is asking, not what a buyer actually paid. It's not a comp until it closes.

Letting one outlier drag the whole number. If one of your comps sold for way more or less than the others, don't just average it in — investigate it. Common culprits: a cash sale between family members (priced below market as an informal gift), an estate or foreclosure sale (priced for a fast close, not top dollar), a bidding war that got emotional, or a sale that included unusual terms (seller financing, personal property thrown in, a rent-back deal) that inflated or deflated the recorded price. When a comp doesn't make sense, it usually isn't representative — set it aside rather than let it skew your range.

Treating a stale comp as current. A comp that closed 10 months ago in a market where prices have moved 5–8% since then needs a time adjustment before you use it at face value — otherwise you're pricing off of last year's market, not this one. As of the most recent national data (through early May 2026), price cuts remain elevated at roughly 36% of listings and median list prices were down about 2.2% year-over-year, though conditions vary sharply by metro — some markets are absorbing inventory quickly while others are still resetting expectations (HousingWire, as of May 2026). That kind of local, current context is exactly why "what did comps look like six months ago" isn't automatically "what's true today."

Confusing your comp-based number with the appraisal. These are related but not the same exercise, and mixing them up causes real confusion at the worst possible time in a transaction. Your comp analysis is you (or your agent) estimating market value to inform an offer. The appraisal is a licensed, independent third party doing a more formal version of the same sales-comparison approach — for the lender, to make sure the collateral backing the loan is worth what's being borrowed against it — using stricter rules about which comps qualify and how old they can be. They usually land close together, but they don't have to: roughly 1 in 12 appraisals nationally come in below the agreed contract price, a mismatch known as an "appraisal gap," and it happens more often in markets where prices have been moving fast or the subject property is unusual enough that good comps are scarce (Redfin on appraisal gaps). Your comp math is the thing that should make an appraisal gap a rare surprise instead of a routine one — if your own adjusted comps already cluster meaningfully below your offer price, that's useful information before you're contractually stuck, not after.

Skipping the adjustment step entirely. Averaging three unadjusted sold prices and calling it a day is better than nothing, but it's not the same analysis a lender's appraiser is going to run on the same house a few weeks later. If your number and the eventual appraisal are wildly different, this is usually why.

One more note on where this data doesn't come from: if you've read our other pieces on this site, you know we lean hard on HMDA (Home Mortgage Disclosure Act) data for lender and market-pattern research. It's not useful here — HMDA tracks loan originations and applicant/lending patterns, not property sale prices, so it can't tell you what a house down the street actually sold for. Comps come from MLS, sold-listing sites, and county records, not federal lending data.

The bottom line

A fair offer price starts with three to six recent, sold, genuinely comparable properties — not listings, not a Zestimate, not "well, my coworker's house was about this big." Adjust each one for the real differences between it and the house you're bidding on, throw out anything that doesn't make sense, and you'll land on a defensible range instead of a guess. What you actually offer relative to that range — at it, above it, below it — is a market-strategy question for a different day. This is just the math you need before that decision is even worth having.

Sources