The agent tool landscape

Why every home-value tool shows a different number for the same house

6 min read

Type one address into three tools and you get three prices. The spread is structural, not a bug.

The short version

Every automated home-value tool starts from a different slice of data, picks comparable sales by a different rule, and updates on a different clock, so the same house returns a different number on each. Zillow, Redfin, realtor.com, and a county assessor are not measuring the same thing, and none of them can see inside the home. The number that matters is not any single estimate but the range across all of them. How wide that range is tells you how little the models really know about that specific property.

Four tools, four different numbers

Put one address into four places and four numbers come back, each defensible on its own terms because each is answering a slightly different question from a slightly different data set.

Tool	What it is built from	Why its number drifts
Zestimate (Zillow)	Public records, tax data, prior sales, comparable sales, and owner-submitted facts where available	Wide national coverage, so it reaches homes with thin data and fills gaps with the model
Redfin Estimate	Heavily weighted to current MLS listing data and recent sales	Tighter on actively listed homes, weaker where there is no recent MLS signal
realtor.com estimate	A third-party automated model on public records and sales, often shown as several vendor numbers	Different vendor, different comp rules, so it lands on a different figure again
County assessed value	Set by the tax assessor on a schedule for taxation, not for sale pricing	Built for tax, updated slowly, and often well below market by design
Agent CMA	Hand-selected recent comparable sales adjusted for condition, location, and features	A person weighs the specific home, so it can move away from every automated number

The three reasons the numbers diverge

Almost every gap between two estimates traces to one of three causes:

Data coverage. One tool has a recent MLS record for the home and the block; another is working from stale public records. Different inputs, different output.
Comparable selection. Each model chooses which recent sales count as comparable by its own rule, radius, recency, and similarity. Change the comp set and you change the number.
Update cadence. Estimates refresh on different schedules. In a moving market, a tool that updated last week and one that updated last month will disagree for that reason alone.

What none of them can see

Every automated tool shares the same blind spot. It cannot see the inside of the home. Condition, a recent renovation, an awkward or excellent layout, the view, and street-level facts such as a busy road or an adjacent commercial use all move a real sale price, and none of them appear in public records. A model prices the house on paper. The market prices the house that buyers walk through. The distance between those two is where the estimates scatter. For a sense of scale, the most used of these tools, the Zestimate, carries a published median error near 7.5% on off-market homes, which is roughly $45,000 on a $600,000 home, and half of all estimates miss by even more.

Why the spread itself is the useful signal

The instinct is to ask which number is right. The better question is how far apart the numbers are. A tight spread across tools usually means dense, recent, similar sales nearby, so the models are on firm ground. A wide spread means thin or mixed data, so every number is a guess wearing a dollar sign. The spread is a free confidence reading. Treat a narrow one as informative and a wide one as a warning that the specific home needs human eyes, not another algorithm.

What this means for an agent

When a seller opens with a number from one site, the losing move is to say it is wrong. The winning move is to show the spread. Pull two or three of the automated estimates, put them side by side, and let the client see that the tools do not agree with each other. Then explain why, different data, different comparable sales, none of them able to see the finished basement or the new roof, and bring the recent sales that set the price on this block. That is the difference between an argument and a reasoned range, which is the case for showing a range at all in what confidence should mean in a valuation tool. For why the automated numbers behave the way they do, see how AVMs work and where they break, and for how the estimate types line up against each other, AVM, CMA, or reasoned valuation. A reasoned valuation that shows an explicit range, like our home-value page, is built to make the spread the starting point of the conversation rather than a number to defend.

Frequently asked questions

Why does Zillow show a different price than Redfin?

Because the two tools use different data and different rules. Redfin leans heavily on current MLS listing data, so it is often tighter on actively listed homes, while the Zestimate draws on wider public records to cover homes with thin data. Different inputs and different comparable-sale rules produce different numbers for the same house.

Which home-value website is most accurate?

None is reliably most accurate for every home. Each tool is stronger on the homes that match its data, listed homes for MLS-weighted tools, common tract homes for records-based models, and weaker on unique or off-market homes. The more useful reading is the range across several tools, not a single winner.

Why is the county assessed value so different from the market estimate?

Because the assessed value is built for taxation, not for pricing a sale. Assessors update on a schedule and often set values below market by design, so the assessed figure can sit well under what a home would sell for and should not be read as a market estimate.

Should I trust the Zestimate or the Redfin Estimate?

Treat either as one input, not the answer. Both can be useful as a starting point and both can miss on a specific home, because neither can see condition, renovations, or layout. Compare them, note how far apart they are, and confirm with recent comparable sales.

What does it mean when the estimates are far apart?

A wide spread across tools is a sign of thin or mixed data, which means every automated number is less reliable for that home. It is the signal to bring human judgment and recent comparable sales rather than to pick one estimate and trust it.

Sources: Zillow and Redfin published descriptions of their estimate methods and accuracy, realtor.com and third-party AVM vendor descriptions, county assessor guidance on assessed versus market value, and general real estate valuation references. Methods and figures vary by tool and market; treat them as general references.

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