Give me a break

[Mejappz]

I did a 1004 yesterday that had significant updates in the last 5 years. I was engaged only after a desktop came in way low for not accounting for said updates. I have numerous stories of this and the flip side as well where a house was in horrible shape on the inside.

Short of recent interior MLS or homeowner supplied pictures, I find it quite incredible that a model will predict condition better than a full inspection would. Quality I can buy somewhat but condition?

I have to think there may be more to this story. I am willing to consider some misunderstanding by appraisers of UAD rating definitions--for which I blame those who created those definitions--but not as much for observational deficiencies.

What would have been really nice all these years would have been for Fannie to communicate with appraisers and say look, you gave subject a C3, Q4 rating. We think it should be x, and y and here is why...

It's all BS don't believe it.

 

[KHS445]

Well, when a report says a property is in C3 condition, but it contains photos showing a bad roof, collapsing ceiling, carpets with bare spots, etc etc, I would submit I really do not need someone else to go look at the property. We are not playing a game of gotcha over C3 vs C4, we are looking for serious errors that affect our risk

Ok, so without benefit of interior pictures showing condition how does you model know what the condition is? I fully agree with djd09 that is sounds like someone either doesn't understand the rating system or how to apply it or they are out and out misrepresenting what is present. One of the last things I do before submitting a report is to review all of the photographs to make sure the good, the bad and the ugly shown in any photo has an appropriate comment in the narrative.

In the example you sighted it appears that the appraiser needed to classify the subject as C3 in order to use comps that would allow him/her to hit the sale price. If that is the case it is either neglect, stupidity or fraud and should not be tolerated. Turning someone into their State Board should not be taken lightly and should be reserved for only the most obvious cases of lying or fraud, but in the case you sight it might very well be worth it. The sooner we as a profession weed out the incompetent the sooner we will gain the level of respect that some many on this forum claim they want.

 

[Overimprovement]

Oh, there is a lot more to the story. I think most appraisers would be shocked at the type and extent of data that is available. I know I was

More than MLS pics, google aerials/street scenes, and prior appraisals/inspections? Most of us here strongly suspect that the Fannie/Corelogic database of information is immense indeed.

I guess I am just struggling to comprehend how a home not listed for sale in 10 years could possibly have its interior condition known by anyone. Most assessors/auditors never step foot into homes. Or is the basis premise that an exterior inspection is a good enough proxy for interior condition? And yes, I am aware of the irony for anyone who has ever done a 2055.

 

[NC Appraising]

The ivory towers have gotten so deep in your head your not able to think straight. How about this. Why don’t you let me test your model? Please send me your data in a spreadsheet format of a neighborhood or market area. I get to pick. I will test it out and provide you with the outputs, including ratio studies of how well the model did. Deal?

That is harsh. Try not to make it personal. I appreciate D coming on here and telling us the insides, nothing wrong with that. I can sound harsh at times too, but it is because I care and only want the good for appraisers.

 

[NC Appraising]

More than MLS pics, google aerials/street scenes, and prior appraisals/inspections? Most of us here strongly suspect that the Fannie/Corelogic database of information is immense indeed.

I guess I am just struggling to comprehend how a home not listed for sale in 10 years could possibly have its interior condition known by anyone. Most assessors/auditors never step foot into homes. Or is the basis premise that an exterior inspection is a good enough proxy for interior condition? And yes, I am aware of the irony for anyone who has ever done a 2055.

It has been a long time and maybe D can set me straight on this one. About, guessing here, 8 years ago I was at a live Fannie Mae training session with a small group of appraisers in my market. One said that Fannie kinda demanded access to MLS systems. My MLS board kinda hinted to the same thing. Not sure how much truth is to that, it was one of those you didn't hear that....

Another "black helicopter" but gut feeling that I have is that Fannie and other big data companies have access to building permits. I was wondering why so many data companies were selling access to permits.....never know, are the computers are Zillow, Fannie, etc. using permits to make adjustments?

 

[Mike Ault]

One said that Fannie kinda demanded access to MLS systems. My MLS board kinda hinted to the same thing. Not sure how much truth is to that, it was one of those you didn't hear that....

It's called a RETS feed https://realtyna.com/blog/rets-data-feed/ and I'd bet you be surprised how many data aggregation firms are signed up and pay a monthly fee to your local Association for access to your information.

 

[Overimprovement]

It's called a RETS feed https://realtyna.com/blog/rets-data-feed/ and I'd bet you be surprised how many data aggregation firms are signed up and pay a monthly fee to your local Association for access to your information.

That site has a great nationwide listing of all known MLS's. I had never seen an aggregated list like that. I would not be surprised to see a centralized database by 2030. We are certainly heading in that direction. It certainly is inefficient to have so much overlap. I have some areas I cover where the house could be listed in any or all of 3 different MLS's. Very stupid!

 

[Mike Ault]

That site has a great nationwide listing of all known MLS's. I had never seen an aggregated list like that. I would not be surprised to see a centralized database by 2030. We are certainly heading in that direction. It certainly is inefficient to have so much overlap. I have some areas I cover where the house could be listed in any or all of 3 different MLS's. Very stupid!

Yes, I currently pay MLS dues to 4 different associations within a 60 mile radius. Wouldn't mind having a single centralized one with standardized, consistently populated data fields however I hate for that to be imposed upon independent associations. The "Walmartization" of the real estate business will probably make those of us that are left more efficient, yet slightly diminished in spirit worker bees.

 

[DWiley]

I guess I am just struggling to comprehend how a home not listed for sale in 10 years could possibly have its interior condition known by anyone. Most assessors/auditors never step foot into homes. Or is the basis premise that an exterior inspection is a good enough proxy for interior condition? And yes, I am aware of the irony for anyone who has ever done a 2055.

Models do not “know” things. They provide predictions based on data. Give the model enough data, turn on the ML and AI, and give it a large test set, and those predictions can become quite good. And, unlike humans, models don’t substitute their own definitions of the condition ratings.

No one is claiming that the model is perfect. But the humans aren’t perfect either. I have seen gutted homes that were reported to be C4 or C3 (as is) by humans as well. It is impossible to eliminate all risk.

 

[J Grant]

Models do not “know” things. They provide predictions based on data. Give the model enough data, turn on the ML and AI, and give it a large test set, and those predictions can become quite good. And, unlike humans, models don’t substitute their own definitions of the condition ratings.

No one is claiming that the model is perfect. But the humans aren’t perfect either. I have seen gutted homes that were reported to be C4 or C3 (as is) by humans as well. It is impossible to eliminate all risk.

such egregious examples are either a one off mistake in a career, or not done by a competent appraiser- in a case where C condition is substantially different from what photos show, the appraiser should get a warning or not hired again. Though a gutted house on inside can still be C 3 outside when some components were not replaced so there can be honest reason why a C condition by appraiser is what it is.

A model prediction of a condition unseen by anyone can self fulfill as "accurate"? How is it proven unless each of those houses were to be inspected inside and out to check the data assigned condition from the model against what an inspection shows and appraiser concludes? If the user's agenda is outcome accepted of a model thus they find results acceptable, there is no breaking of a circular argument.

I do know that receiving Fannie alternate "comps " from CU that it is incredible that given their vast data, they get it wrong so often - the "comp" they send is within a mile, within X sf to subject , but a key feature makes it not similar and thus not a comp ( such as in a gated luxury community and subject is not a PUD or vice versa ). Some alternate comps they send are good and others terrible, so how can a model of condition be any better if their comps are half the time "off" , even with their vast data resources? That is the problem with modeling and machines, though the results don't have the weakness of a a personal error in judgment, there is an element of rote/ "dumbness" to the results .