A hard appraisal rule cost me two deals and saved one provable loss

14 out of 19 confirming the automated number is the part I'd sit with. That's 74 percent of your spend buying you nothing you didn't already have. If the townhouse hadn't been in the sample, this post reads as $5,400 spent to reprice two liens.

@tally's objection is the standard one and it has a specific answer, which is that the value of a screen isn't the average case, it's the tail.

Automated valuation models estimate from recorded data and statistical relationships. They're accurate on typical property in liquid markets, which is why 14 of 19 came back close, and they degrade predictably in a few situations: unusual property types, thin recent sale activity, condition or legal encumbrances that never touch a public record, and neighborhoods where a model's geographic boundary lumps dissimilar submarkets together. The townhouse in tender's post hits three of those at once. A pending association assessment is not in any dataset the model reads, and phase level pricing differences inside one complex are exactly what a radius based model averages out.

So the screen isn't buying accuracy on the 14. It's buying identification of the ones where the model's assumptions don't hold. Whether $5,400 across 19 files is a rational price for that depends entirely on the loss distribution in the portfolio, which is a question about the specific tape and not something anyone here can answer for you.

On the regulatory side, there is a federal quality control rule for automated valuation models used in certain credit decisions, and requirements differ by transaction type and institution. Confirm what applies to your own purchases in writing with counsel rather than assuming a rule aimed at lenders covers a note buyer.

One mechanical note on the turn time problem: desktop and exterior products vary a lot in what the appraiser is allowed to assume about interior condition, and that assumption is stated in the report. Read it, because on a lien behind a first it's often the assumption carrying your position.

The assessment on that townhouse is the kind of thing I'd expect to find in association minutes and estoppel documents rather than from an appraiser, and on second liens almost nobody orders those. Your appraiser caught it as a byproduct of standing in the parking lot, which is luck as much as method.

If you want the townhouse catch to be repeatable rather than fortunate, the trigger probably shouldn't be your 78 percent number alone. Any lien secured by a unit in a condo or townhouse association is a candidate regardless of the loan to value, because the thing that impairs you there isn't the value drifting, it's a lien or an obligation sitting in front of you that no model reads. I'd run the exterior on all of those and keep the desktop for detached houses.

Your turn time cost is also understated. Eleven days on an exterior means you're bidding against people who don't have that step, and the sellers who won't hold are going to be the ones with the cleanest tapes, because they can sell those to anyone. That's an adverse selection problem sitting inside a good rule.