We tested six reasons a house sells. Five of them were noise.

The short version

There is a rate at which a house changes hands whether anything happens to it or not. In St. Charles County, over six months, it runs around 2.5% to 2.8% depending on which homes you are comparing against. That number is the whole game. A signal is only worth something if it beats the rate its own comparison group already moves at, and almost nothing does.

We ran six candidate signals against that baseline. Five of them came back flat. Here is the entire scoreboard, including the parts that make us look bad.

The one that fired

A comparison group is where most published numbers quietly fail, so here is ours. Probate homes are not ordinary homes — they are cheaper and they have been held far longer. Long-held homes sell less often. Compare them against “the county” and you are comparing two different kinds of house.

So every probate home is matched only against homes in the same price band and the same length-of-ownership band, over the same calendar months. That correction pushes the number up, not down, because long ownership drags the honest baseline lower.

The test almost nobody runs

A skeptic’s objection to any signal is that it is not the event doing the work — it is just the kind of house that turns over a lot. There is a way to check, and all it requires is the willingness to be wrong.

We ran the identical measurement on the identical homes, with the clock started one, two and three years before the probate was ever filed.

All three land at or below 1.0. These homes never sold faster than an ordinary house. They ran at or below the normal rate for three straight years, and then a case was filed and the rate jumped. That is not a type of house. That is an event, on a date.

The five that did not

Each of these was tested and did not survive; the method section states what each test was, because they were not all the same test. They are published for the same reason a negative result belongs in any honest report — a research program that only ever discovers good news is not measuring anything.

Falling behind on the tax bill

This is the assumption we most expected to hold, and it did not. On a 1,400-parcel study with the tax status frozen before the outcome window ever opened, the odds ratio came back 1.28, with a range from 0.68 to 2.40. That range excludes anything at or above 2.4×.

The detail that settles it is the direction. If financial distress were driving sales, deeper delinquency should predict more selling. It runs the other way: 1.79 at 31 days past due, 1.28 at 60 days, and 0.72 at six months. A ladder that descends as the trouble deepens is running the wrong way for a distress signal. We can say the effect is absent and that it weakens as the delinquency ages. We cannot tell you why, and we are not going to guess.

What the house is worth, and how old it is

Value is flat across all five price bands. Age is flat and does not even move in one direction. Within probate leads specifically, size, number of bedrooms and length of ownership all wash out too. Every physical characteristic of a house — the things a person reaches for first — tells you close to nothing about whether it is about to sell.

How many names are on the deed

One name, two names, or a deed reading “et al”: 1.07×, 0.92× and 1.20× against the rate those same homes move at anyway. All three ranges touch or cross 1.0. Whatever a title structure moves, it is small and its direction is not even consistent.

One honesty note we would rather state than have found: this one is measured over six and a half weeks, because that is the entire length of the ownership record it depends on. It is enough to say these are not large effects. It is not enough to call them zero at six months, and we will not.

Both co-owners having died

This one was not ruled out for being weak. It was ruled out for being redundant. Every property it identifies is already a property a probate filing identifies, so it delivers nothing the first signal has not already delivered. A signal that only ever agrees with one you already have is not a second signal.

Even the one that works is not uniform

A signal that fires is not a signal that fires evenly. Within probate filings, one of the sharpest things we can grade on is the calendar — how long after the death the case was opened. What the family actually filed separates them at least as sharply, and is covered below.

The bottom row is the one worth sitting with. 105 homes, none of them sold. Everything filed more than a year after the death converges on the rate of a house picked at random off the street, and the last row falls under it. Those are not weak leads. They are leads that make a list worse than no list, and the correct thing to do with them is not to send them.

What the family filed matters at least as much as when

Which probate document was filed tells you more than anything about the house. Independent administration — the filing you make when there is property to deal with — converts at 20.2%. A refusal of letters, where a spouse declines administration and the property simply passes, converts at 4.3%. The gap holds even after controlling for how fast the case was filed, and it replicates in both counties.

That the two should differ is a legal reading of what each instrument is for, not a statistical finding, and we state it as reasoning. The numbers themselves are measured.

What the five times does not mean

This is the part it would be easier to leave out. We will keep going, because the gap between these two sentences is where a customer gets disappointed.

14.1% is the chance the house sells to anyone. It is not the chance you get to buy it.

Of the probate homes that sold, 98.4% were ordinary arm’s-length transactions — slightly more arm’s-length than the county’s own baseline. The family lists with an agent and sells at retail. Narrow the outcome to a company owning the home afterward, the closest proxy we have for an investor purchase, and the rate is 3.3%.

So the honest arithmetic is this. Out of 100 probate leads, about 14 sell within six months. About 3 go to an investor. The other 11 sold the ordinary way — and reaching the family before that happens is the entire proposition. Anyone quoting you 14% as an acquisition rate is quoting the wrong number.

One caution on the 3.3%: it is a floor. Our sale records carry the parcel’s current owner on older rows, so a home an investor bought and later resold to a family reads as a family purchase. The real figure is higher. We cannot say how much higher, so we publish the floor.

What we still cannot tell you

• The five-times figure is one county. St. Charles, 184 homes, filings across about a year. The six-month sale rate replicates in St. Louis County at 14.4% on a much larger sample — but St. Louis has no comparison group we trust yet, so there is no St. Louis multiple, and we will not print one.
• The edge decays. Followed further out, the same homes convert more often while the ordinary rate rises faster: the advantage falls from about 5.4× at six months to 3.4× at eighteen. Half of what eighteen months delivers has already landed by month six.
• Most probate cases never become a lead. Only about a third of the decedent-estate cases in this window could be tied to a specific parcel. That costs us inventory; it does not make a delivered lead less accurate.
• Nothing here measures whether anyone closed a deal. That takes outcome data we do not have. When we have it, we will publish it, including if it is unflattering.

Why publish the misses

A list tells you something happened. Five of the six things in this report happened, and turned out to mean nothing about whether the house would sell. The difference between those two statements is the only thing worth paying for, and it is invisible until somebody measures both sides and shows you the losing half.

We would rather hand you five signals we killed than one we never checked.