A comparative market analysis, or CMA, is a written opinion of what your home would sell for today, built from recent sales of similar homes that a person has actually examined. An instant online estimate is a formula applied to public records. The formula does not know your kitchen was redone in 2021, that the house two doors down had a flood, or that your floor plan is the one everyone in the neighborhood wants.
We get asked why our website doesn't show an instant number. This post is the long answer.
What goes into a CMA
The work is in choosing the right comparable sales and then adjusting for the ways they differ from your home. A good CMA for a Central Florida home looks at:
- Closed sales, not asking prices, of homes as similar as possible in size, age, condition and location, usually from the last several months. Asking prices tell you what sellers hoped for. Closed sales tell you what buyers paid.
- Adjustments for what is actually different: a renovated kitchen, a newer roof, a pool, a larger lot, a busier street. These are judgment calls, and they are where experience matters.
- Condition, which no database records. Two homes with identical square footage in the same subdivision can sell tens of thousands of dollars apart because one has been maintained and the other hasn't.
- What would change the number. A useful CMA tells you not just a range but what would move your home toward the top or bottom of it. That is the part you can act on.
Ours is a written document. You can read it in about ten minutes, and we go through it with you.
Where the instant estimate goes wrong
Instant estimates are built from tax records, prior sale prices and listing data, run through a model. For a home that is currently listed, with fresh photos and a stated condition, the model has a lot to work with. For a home that is not on the market, it is guessing from the outside. Zillow publishes its own accuracy figures, and the error for off-market homes is substantially larger than for listed ones. On a home you are about to sell, the off-market number is the only one you've got.
The failure cases are predictable. The model overvalues the dated house because it looks like its renovated neighbors on paper. It undervalues the one with the new roof, the repipe and the updated bath because none of that is in the record. It cannot tell a lakefront lot from one backing to a road. In the neighborhoods we work most, Oviedo, Winter Springs, Winter Park and east Orlando, the housing stock is largely from the 1970s through the 1990s, and how much of it has been updated varies house to house. That is exactly where a formula struggles.
What the Zestimate is actually for
Our read, and we think the record supports it, is that the Zestimate was never built to be a pricing tool. It was built to get you onto Zillow. Zillow makes its money selling your attention to agents and feeding its own mortgage business, and a number next to every house in America is a very good way to get people to show up and start clicking. For that job the number does not have to be right. It has to be interesting.
The one time Zillow bet its own money on the Zestimate being right was Zillow Offers, where it used the model to buy houses directly and resell them. It lost hundreds of millions of dollars, wrote down the homes it was holding, laid off about a quarter of its staff, and shut the program down in 2021. Zillow itself said that forecasting home prices had turned out to be far less predictable than it expected. That is the clearest evidence we know of: the Zestimate works well as a marketing instrument and poorly as a pricing instrument, and Zillow found out which business it is in.
When an instant estimate is fine
If you are curious what your home is roughly worth and have no plans to sell, an instant estimate is a perfectly good way to satisfy the curiosity. The problem is using it to set a list price, decide whether to sell, or judge an offer. For those decisions the gap between the formula and reality costs real money, in either direction.
What we've seen
Two Winter Springs homes make the point from opposite directions. On one, the Zestimate came in too low. We priced it from the actual comparable sales and sold it for close to $30,000 more than the Zestimate said it was worth. A seller who had trusted the number would have left that on the table.
On another, the Zestimate came in too high, and the reason was simple once we looked: Zillow had the house recorded as three bedrooms. It had two. The model priced a home that did not exist. Nobody at Zillow had been inside, and no formula can correct a record it does not know is wrong.
We have also watched a home's Zestimate move significantly, up and down, in the days after that home was listed for sale. Nothing about the house changed. What changed was that an asking price now existed, and Zillow has said that for a home on the market its estimate uses the listing itself. That is worth sitting with: a number that moves to meet the asking price is not an independent read on what the home is worth. It is a reflection of whatever price someone just typed in.
Same town, same tool. The miss went one way, then the other, and then the number itself moved once a listing went up. That is not a tool you price a sale with.
How we do it
Tell us the address and a little about the home, and we prepare the analysis ourselves, usually within 24 hours. There is no charge and no obligation. We do it because some of the people who ask go on to list with us, and most of the rest tell a friend.