The Condition Record Deletes Itself

A sale closes. Within weeks, in most markets, the interior photographs come off the listing. What remains in the record is square footage, bedroom and bathroom count, lot size, year built, and a price. What has gone is the only evidence of what the property looked like on the day it traded.

That property is now a comparable sale. It will be used to price neighbouring homes for the next 6 to 12 months, and the single largest adjustment anyone makes against it, the one for condition, will be made against a record that no longer contains any information about condition.

The adjustment with no evidence behind it

Condition and update quality routinely move a valuation by more than any other single factor short of location. Two houses of the same size on the same street, one renovated in 2023 and one untouched since 1994, are not close to the same number. Every valuation method in use, automated or human, has to account for that difference.

The automated method cannot, because the fields it reads do not describe condition. The human method can, but only for as long as the evidence survives. An agent who saw the property, or who pulled the photos before they came down, holds something the record does not. An agent who arrives 8 months later is working from the same thin file the model is.

So the quality of a condition adjustment depends on when it was made relative to when the photographs were removed, which is a strange thing for a valuation input to depend on. It sits in the same family of absent fields as what a Zestimate cannot see, with the difference that this one was present and then taken away.

Nobody chose this and everybody lives with it

The deletion is not a conspiracy and not an oversight. Listing photographs are licensed work, and the licence usually runs to the listing brokerage for the duration of the listing. When the listing ends, the right to display the images often ends with it. Retention rules differ by market, some hold images for a period and some do not, and the ones that hold them frequently hold them somewhere an agent pricing a home will not think to look.

The result is that the most valuable pricing evidence in the system has the shortest retention of anything in it. Sale prices are kept for decades. Photographs of what was sold are kept for weeks. If the purpose of the record is to support future valuation, that is precisely backwards.

What this does to the tooling argument

A recurring claim in valuation technology is that better models will close the gap between an automated estimate and a considered one. On condition, they will not, because the gap is not a modelling problem. It is an evidence problem, and the evidence is being destroyed on a schedule.

If comp selection is the whole game, then the evidence supporting each selection is the board it is played on, and the board is being cleared on a schedule. A model trained on every closed sale in a market still has no image of the interiors of those sales, because the images are not there to train on. More compute against an absent field returns nothing. The only way to price condition is for someone to have seen the property, or to have preserved what was published about it while it was still published, and both of those are human acts performed inside a window that closes.

Why are MLS photos removed after a home sells?

Listing photographs are licensed to the listing brokerage, and that licence is commonly tied to the life of the listing. When the listing closes, the display right often lapses, so the images come down. Retention rules vary by market and some systems archive images for a period, but the archive is frequently not where an agent pricing a nearby home would look, and in many markets the interior photographs are simply gone.

Does the loss of photos affect a comparative market analysis?

It affects the condition adjustment, which is usually the largest adjustment made against any comparable. Without images, the person or system pricing the home has no evidence of finish level, renovation quality, or deferred maintenance in the comparable property, and has to fall back on assumption. Two valuations of the same home can differ substantially on this basis alone, depending on whether whoever built them saw the comparable while it was listed.

Can an automated valuation account for property condition?

Not from the public record, and not from the closed-sale record after photographs are removed. Automated models price against fields that persist, which are size, room count, lot, age, and price. Condition is not a field in that set. This is not a limitation that improves with a better model, because the input does not exist in the data the model reads.

The record is thinner than the number suggests

Every confident valuation of a home rests on a set of comparable sales, and every one of those comparables is a file that has already been stripped of the thing that most explains why it sold where it did. The number still arrives to the dollar. The evidence behind one of its largest components expired months ago.

This is Context Blindness™ with a timestamp on it. The model cannot see condition because the record no longer holds condition, and the agent who can see it can only do so by having been present, or by having captured what was published before it came down. A reasoned valuation, the approach CMAflow builds, treats condition as a documented judgment rather than an inferred one, which means the reasoning is written where the evidence is thinnest rather than left implied. The record will keep deleting itself. What can be preserved is the account of what was there.

---

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

The Independent Agent
Substack | Spotify | CMAflow FAQ | YouTube | Free CMA | Home valuation | Insights | Blog