How home valuation tools work

Comp selection is the whole game

6 min read

Change the comp set and the value moves. Everything else is arithmetic.

The short version

Strip the branding off any valuation method, the AVM, the CMA, the appraisal, and the same engine is underneath: find recent sales of similar homes nearby, adjust for differences, land on a number. The difference between methods is not the math, which is simple. It is the selection, which is everything. An algorithm picks comps by rule, radius, recency, similarity scores. An appraiser picks a handful and defends each one. An agent picks by a test no rule can run: would the buyer for this home have bought that one. The estimate is only as good as its comp set, which is why the first question to ask of any number is not how it was calculated but which sales it stands on.

Every method is a comps method

An automated model scores thousands of candidate sales and weights the nearest matches. An appraiser selects a small set, usually 3 to 6, and adjusts each line by line. An agent assembles the set a serious buyer would recognize. Different rigor, different transparency, same skeleton, which is the comparison drawn method by method in AVM, CMA, or reasoned valuation. Whatever the wrapper, the number inherits the comp set, including its mistakes.

The three dials

Every selection rule, human or algorithmic, is a setting of three dials, and every setting is a trade.

Dial	Set tight	Set loose	The trade
Radius	Same micro-market, few sales	More sales, wrong micro-market	Precision against sample size
Recency	Current market, thin set	Deeper set, stale prices	Freshness against depth
Similarity	True peers only	Near-peers admitted	Fit against coverage

The test a rule cannot run

A comparable sale is not a home that matches on paper. It is a sale the same buyer pool would have considered. A rule can match beds, baths, and square footage; it cannot know that the buyer for a renovated craftsman on a quiet street was never going to buy the flip backing the arterial two blocks over, even though the records rhyme. Condition parity and micro-location are exactly the variables that never reach a record, the inventory in what only the block knows, and they are the variables that make or break a comp set.

Reading any estimate by its comps

The practical use of all this is a habit: when a number arrives, ask for its comps before you ask anything else. If the comps are right, the number is usually close. If the comps reach across a boundary, the number is an average of two markets. And when several tools disagree about the same house, they are almost always disagreeing about the comp set, not the arithmetic, which is the spread examined in why every tool shows a different number.

What this means for an agent

Your comps page is the whole argument. Defend the selection, not the number: why these sales, why this radius, why that one was excluded, and the price defends itself. Sellers do not need to believe your figure, they need to believe your set. The appointment-ready version of that case, built against whatever automated number the client walked in holding, is our home-value accuracy page.

Frequently asked questions

What makes a good comparable sale?

A recent closed sale the same buyer pool would have considered: same micro-market, similar condition and size, sold under current market conditions. Paper similarity is the start; the buyer-pool test is the real filter.

How many comps should a CMA use?

Enough to show the pattern without diluting it, commonly 3 to 6 closed sales, sometimes with active and pending listings for current context. Past a point, adding sales widens the net into the wrong market.

What radius should comps come from?

As tight as the data allows. The right boundary is the micro-market, the school line, the block character, the road, not a fixed distance. A half-mile circle that crosses a market boundary is worse than a two-block set that does not.

Why do two valuations of the same home differ?

Because they stand on different comp sets. Different radius, recency, and similarity rules select different sales, and the number inherits the set. The arithmetic rarely differs; the selection almost always does.

Can I challenge the comps in an appraisal?

Yes. Lenders have a reconsideration-of-value process, and the effective challenge is better comps, closed sales that fit the subject more closely, with the reasons stated, rather than an objection to the conclusion.

Sources: Appraisal comparable-selection guidance including lender reconsideration-of-value processes, published automated valuation methodology descriptions, and standard CMA practice references. Selection norms vary by market density; the principle, that the comp set drives the number, does not.

This article is general information and analysis, not financial, lending, or appraisal advice. Verify any home value with a licensed professional before acting.