Give me a break

[Mejappz]

If you think a MAPE of 7% is good, then how would you characterize a model with a MAPE that is less than half of that?

One thing I forgot to mention. In most jurisdictions in the United States, a MAPE of less than 5%(especially in the 1,000,00 to 1,800,00 range) would be a red-flag and would trigger an audit by the state. And a review by an outside **independent** third party would be warranted.

 

[DTB]

As a retired appraiser, I do not see a significant problem using a non appraiser but conscientious person to do the property visit and use one of the fancy systems to measure/floorplan the property and have a cascading AVM provide the number in my market.

 

[Mike Ault]

As a retired appraiser, I do not see a significant problem using a non appraiser but conscientious person to do the property visit and use one of the fancy systems to measure/floorplan the property and have a cascading AVM provide the number in my market.

What if the person pulling the levers on that AVM could manipulate their employers stock price at will simply by pulling the right one? See any performance-based compensation issues that might arise?

 

[DTB]

What if the person pulling the levers on that AVM could manipulate their employers stock price at will simply by pulling the right one? See any performance-based compensation issues that might arise?

I think using several independent AVMs, if there are such things, would overcome that. No?

 

[Terrel L. Shields]

If a “full appraisal” is used how would that lessen the burden on the borrower?

I am merely pointing out that the cost of an appraisal is a bargain. If it was important to have an appraisal in 1993, then why isn't it important in 2021. It certainly cannot be a issue of the comparative costs. The present mortgage holder is borrowing money so cheap it is unbelievable for those of us who borrowed money for 40 years or so at rates far higher than what people now complain about being "too high". 5% is too high? I never had a mortgage that cheap. It almost makes me wish I had a mortgage...err, no it doesn't. But I am saying the cost of the appraisal is a nothing burger yet everyone seems to want to cut the price down to the price of a credit check.

 

[Mike Ault]

I think using several independent AVMs, if there are such things, would overcome that. No?

Yes, as long as the output from those is controlled and reconciled by someone not employed by the end user. Prior to the crash financial innovation gurus were talking about "blended AVMs" and how wonderful they were, don't know how those loans actually performed. The fact that "risk management" considers itself sufficient authority to issue mortgage-backed securities without independent valuations is what should be bothering everybody about this whole scenario. Not whether appraisals and AVMs differ by 5 or 50%.

 

[djd09]

Coester Appraisal Group Launches Cloud-Based Appraisal Tool​

Coester Appraisal Group has launched Cloud Control, its revolutionary new appraisal management technology. Cloud Control is the only appraisal management software built on the award-wining platform of Salesforce.com, which Forbes magazine designated as the most innovative company in the world in July 2011. Cloud Control enables lenders to customize their appraisal processes far beyond the levels offered by other appraisal management technologies. Cloud Control offers virtually limitless customization—users can create business rules to automate virtually any function, from protecting standards through firewalls and safeguards, to business-unique sales and marketing activities that help companies generate new business in addition to enhancing compliance and efficiency.

Coester is providing Cloud Control completely free of charge to all lenders—Coester customers and non-customers alike—without obligation to use any additional Coester valuation service, whatsoever. Cloud Control is a cloud-based, end-to-end appraisal management technology that efficiently manages the entire appraisal cycle. In addition to automating functions and tasks ranging from relaying the initial request to the lender’s desired appraisal panel to filing and storing the completed report, the system automatically extracts and converts data into UCDP-compliant formats and submits appraisal data through the Uniform Collateral Data Portal (UCDP). Cloud Control provides the property value as estimated by an automated valuation model (AVM),

Coester Appraisal Group Launches Cloud-Based Appraisal Tool

Coester Appraisal Group has launched Cloud Control, its revolutionary new appraisal management technology. Cloud Control is the only appraisal management software built on the award-wining platform of Salesforce.com, which Forbes magazine designated as the most innovative company in the world in...

nationalmortgageprofessional.com

what could go wrong?

 

[Mike Ault]

Cloud Control provides the property value as estimated by an automated valuation model (AVM),​

what could go wrong?

They probably required all the appraisers on their panel to submit reports in "AI Ready".env format for data mining.

 

[J Grant]

Yes. And those are all elements of risk management for the lender/investor.

That 's all good, the lender/investor can call their use : "risk management"

But ( as you know ) it states on the appraisal : Purpose: an Opinion of market value ( as defined ) of the subject property ( as of x effective date )

It does not state on the appraisal : Purpose: "provide an opinion of risk management"

 

[J Grant]

I also believe that in the future, an improved appraisal product might include AI ( if developed by then relevant to RE ) and portions of an AVM or statistics.

However, what an appraisal really does is model the behavior of a typically motivated buyer and seller of a property, and they do not normally make decisions based on statistics nor on robot like "supported" adjustments. They make decisions about the pluses and minuses of a given property within a price per their budget /wants and needs. A borrower ends up with a smaller set of similar properties that might substitute ( the comps ) .

Well derived adjustments prove themselves right there on the grid, by narrowing down the prices to a more narrow value range ( via sensitivity analysis/extraction from pairing sales ) - when too much data shows a mathematical "support" for an adjustment, the data itself can be suspect because the larger data set for statistical results starts to include properties less similar to subject and thus can skew results . And if the wrong comps are chosen, what good are well supported adjustments ? And we also have the raw, not adjusted sales prices to check against -