Why is the Zestimate wrong on my house?

The honest answer, with the numbers Zillow publishes about itself.

The short version

If your home is not currently for sale, the Zestimate you are looking at is the least accurate version of that number Zillow will ever show you. Zillow's own published figures put the median error for off-market homes at roughly 7% to 7.5%. On a $600,000 home, that is a swing of about $45,000, and because it is a median, half of all homes are off by even more than that.

That is not a knock on Zillow. It is a limit built into what an automated estimate can and cannot see. This piece explains exactly where the number comes from, why the off-market version is the weak one, and what moves a real sale price away from the estimate.

What a Zestimate is

A Zestimate is an automated valuation. Zillow's algorithm takes public records, tax assessments, recent nearby sales, and any facts homeowners have submitted, and runs them through a machine-learning model to produce a single number. It has been doing this since 2006, and the model has improved over that time.

The important word is automated. No one walks through your home. The estimate is built entirely from data that can be collected without ever opening your front door, which is the root of everything that follows.

The two accuracy numbers, and why only one of them matters to you

Zillow publishes its own accuracy data, which is more transparency than most automated valuation tools offer. But the headline number and the number that applies to you are usually not the same one.

For homes currently listed for sale (on-market), Zillow reports a national median error around 1.9% to 2.4%. That sounds excellent, and it is the figure most often quoted.

For homes not currently for sale (off-market), the median error rises to roughly 7% to 7.5%. That is more than triple the on-market figure.

Here is the part that matters: if you are checking your Zestimate right now and your home is not listed, you are looking at the off-market number. The impressive 2% figure is not the one describing your house. The 7%-plus figure is.

Why the on-market number looks so much better

There is a quiet reason the on-market estimate is so much tighter, and it is worth understanding because it changes how much trust the number deserves.

When a home gets listed, the asking price becomes a data point the algorithm can lean on. The appraiser Ryan Lundquist has made this point clearly: Zillow measures its accuracy using the most recent estimate before the sale, by which time the estimate has often already drifted toward the list price. In other words, once your agent sets a price, the estimate adjusts toward it, and then Zillow grades itself against the eventual sale.

The estimate is, in effect, checking the homework after seeing part of the answer. The off-market estimate, the one you see while you are simply deciding whether to sell, has none of that to anchor to. It is the version with the fewest useful inputs and the widest margin for error, and it is the version most homeowners are looking at.

What "median error" really means

A median error of 7.5% does not mean every estimate is within 7.5%. It means half are closer and half are further off. Your home could easily be in the half that is further off, and nothing on the page tells you which half you are in.

It also varies enormously by location. National medians hide the spread. In a market like Pittsburgh, the off-market median error has been reported around 11%. Vermont has run near 12% to 13%, while a data-rich, uniform market like parts of Colorado sits closer to 5%. A tract home in a large subdivision, where dozens of near-identical houses sell every year, is far easier for an algorithm to price than a renovated or unusual home on a street where little has sold recently.

What the estimate cannot see

This is the core of it, and it is the same limitation across every automated tool, Zillow, Redfin, the bank estimators, all of them. They read square footage, bedroom and bathroom counts, lot size, and age. They cannot see:

Condition. A renovated kitchen with quartz and new cabinets and a kitchen with 1998 fixtures can carry the identical square footage and room count, and the algorithm will value them the same.
Upgrades it was never told about. A new roof, a replaced HVAC system, a foundation repair with a transferable warranty, none of it registers unless someone entered it into the record.
Lot position. Backing to a greenbelt with no rear neighbor versus backing to a busy road is a real difference in price and an invisible one to the model.
The quality of the comparables themselves. Two recent nearby sales might look identical on paper, but one was a tired rental and one was move-in ready. A person who saw both knows which should count. The algorithm only sees the numbers.

These are not edge cases. They are the ordinary facts of ordinary homes, and they are exactly where the gap between the estimate and the real sale price comes from.

The cautionary tales worth knowing

Two are worth remembering because they come from Zillow's own side of the table.

Zillow's former chief executive, Spencer Rascoff, sold his Seattle home for close to 40% below its Zestimate at the time. That is a documented transaction, not a rumor, and it illustrates the limits of the tool even for the people who understand it best.

And Zillow's own attempt to buy homes based on its estimates, the iBuying program, lost roughly $880 million and was shut down by 2021. The company built the algorithm and then could not safely trade on its numbers. That is the clearest possible signal about what an automated estimate is good for and what it is not.

So what is the number good for?

Treat the Zestimate as a compass, not a map. It tells you roughly which direction you are facing, the general neighborhood of value. It does not tell you the distance to the destination or what stands in the way.

For tracking the rough trend of your equity over time, it is fine. For a real decision, pricing a listing, making an offer, planning a sale, it is a starting point that needs a second, human step. You can improve it modestly by updating your home's facts on Zillow directly, square footage, room counts, renovations, but even a perfectly updated estimate still cannot account for condition, demand, or the micro-location factors that decide the final price.

What produces a reliable number

A comparative market analysis, a CMA, prepared by someone who knows your specific market and has seen comparable homes in person. The difference is not that a human is mystically better than a computer. It is that a human can see the things the data cannot: the condition, the upgrades, the lot, and the real quality of each comparable sale. A good CMA also shows its work, the comparable homes used, the price per square foot, the adjustments, and an honest confidence range rather than a single false-precision number.

That last part matters most. A single number hides its own uncertainty. An honest valuation shows you the range and tells you why the range is as wide as it is. The estimate gives you one number with no explanation. The honest version gives you the reasoning, which is the thing you can act on.

Frequently asked questions

How accurate is the Zestimate for an off-market home?

Zillow's own published figure is a national median error of roughly 7% to 7.5% for off-market homes. On a $500,000 home that is about $37,000, and on a $750,000 home roughly $56,000. Because it is a median, half of all homes are off by more than that.

Why is the on-market Zestimate so much more accurate?

Once a home is listed, the asking price gives the algorithm something to anchor to, and the estimate tends to drift toward the list price. Zillow then measures accuracy against the eventual sale. The off-market estimate, which you see when your home is not listed, has no list price to lean on and is the least accurate version.

Is Redfin's estimate more accurate than Zillow's?

They are very close. Reported figures put Redfin slightly better on-market and Zillow slightly better off-market, but the differences are small. Both rely on the same kind of data and share the same fundamental limit: neither can see inside your home.

Can I make my Zestimate more accurate?

Somewhat. Updating your home facts on Zillow, square footage, bedroom and bathroom counts, and renovations, can help, especially where public records are outdated. But it cannot capture condition, buyer demand, or lot-level factors, so it will still be an estimate, not an appraisal.

Is a Zestimate an appraisal?

No, and Zillow states this directly. It is an automated starting point. An appraisal is a licensed professional's formal opinion of value, and a CMA is a market-based pricing analysis from a local agent. The Zestimate is neither.

Sources: Zillow's published Zestimate accuracy page; Redfin's published Estimate accuracy page; appraiser Ryan Lundquist on how on-market accuracy is measured; widely reported figures on the former Zillow CEO's home sale and on Zillow's iBuying losses. Error rates are self-reported medians and vary by location and property type.

This article is general information, not an appraisal or financial advice. For a pricing decision, get a comparative market analysis on your specific home.