RCA
Elite Member
- Joined
- Jun 27, 2017
- Professional Status
- Certified General Appraiser
- State
- California
The future of GSE work is banging on the door, and it will look like the link below. The only GSE work that will trickle down to true independent appraisers are the properties that AMC and staff appraisers can't figure out while trying to deliver a report in the field.
\
Nighthawk - Appraising Reimagined | Bradford Technologies, Inc.
www.bradfordsoftware.com
The following is with respect to residential appraisal software, that is to say, somewhat more broadly, software for estimating Market Value:
It is, as you should know, easy to advertise software that can do many things. Sometimes we do run into residential appraisal software, such as third party plugins for Alamode, that do a good enough job on simple tasks to be worth using, as long as the price is low enough. It's a question of whether you need it or not. If it is not really needed, the chances of long-term success maintaining and selling the software service is likely not good. [ Of course, as already state, non-residential DCF software such as Argus is a much different story. ]
Now as far as major appraisal software companies go, their major forms product currently has a real use in filtering and formatting input, ensuring compliance to GSE and other organizational regulations.
But, when these companies branch out to provide additional software that supports the actual task of appraisal analysis, they fail. They have always failed as far as I am concerned.
Anything based on Linear Regression is going to fail in an area like the SF Bay Area. That is to say an area with a mixture of homes of all ages that have been updated to varying degrees over the past 100+ years. Especially in areas with a great mixture of low to high income jobs.
Can any of the existing companies provide a new software product that goes beyond linear regression, such as using MARS? Well, I know of companies that have pushed a bit in this direction and discovered through trial and error that appraisers are generally too stupid to even begin to use the products they came out with. REALLY. So, they back off. And in the process, assuming you have read a number of posts in this forum over the past 10-15 years, there have been hard feelings created by this companies. Appraisers don't like to be called "dumb" or insulted by vendors.
Now that I have come out in the past several years and discussed on this forum and in some other channels the RCA method, and my various ideas such as:
1. Having advanced analysts provide price models for appraisers to use.
2. Modifying the appraisal forms to have a property column for Value Contributions.
3. Adding additional steps to the appraisal process.
certain individuals believe they can develop a common product that will magically allow our proverbial dumb appraisers to turn out sales grids that are perfect in all respects. They probably plan to
1. Sell access to a price model service that built by an Ph.D. statistical anaylst who understands MARS.
2. Sell access to market area analyses, appropriate for appraisers, including neighborhood boundaries create by cluster analysis and other methods.
3. ...
Just like they have sold dumb Linear Regression programs in the past.
These individuals at this stage of the game assuredly don't really know what they are doing. They are guessing. They are being preemptive sending out adds in advance.
They are also getting old and desperately trying to find a buyer for their business before it is too late. Yea. You see, we are going to go through a major transformation. This is no longer 1990-2020. Things change. It is NOT likely these old appraisal software companies will last much longer. It is not certain they will disappear, but nothing is certain, and nothing looks that good for the future.
Bottlenecks:
1. The more serious bottleneck is not simple, it is very complicated. Far more complicated than most realize. This first bottleneck collecting and understading the necessary real estate data. Digital data.
1a: The MLS data is sometimes of poor quality.
1b: The MLS data needs to be understood, cleaned, transformed and analyzed at the local level by professionals who have a sense of the local market, and they have this developed sense after years of straining to build meaningful models that work. They have a refined understanding of the various factors that impact price that can only be extracted using advanced non-parametric statistical tools such as MARS, K-Means, Principle Component Analysis and Deep Learning.
1c: So, to put this in perspective, the appraisers have to have a sense of what is in the market, so that they can answer questions about how a given subject property relates to the market in general: Is it in the middle of the market in a certain regard, at the low end or the high end. What are the differences, what features important or not. They will have an overall perception of where any given property fits in the market.
2. Ph.D analysts who don't know real estate, smart guys and gals, will invariably get all kinds of things wrong in doing their analyses using MARS for example. I have seen this (I am not going mention names). They ususally live in a far different world - than real estate. Most likely it would be experience in the Pharmaceutical Industry, Genetics, BioChemistry, Quality Control, Demographics or one of many other fields where non-parametric statistical analysis isused. They invaraible make assumptions about the data and models that are not correct. This is the way they see the world. They probably look down on Real Estate as retarded, although it is actually quite complex. They are not motivated to learn more about real estate, construction, GSEs, appraisal regulations or any of that. Notwithstanding that, they certainly could learn all about real estate. But they likely won't unless they are really interested in real estate and want to dedicate several years of their life working with it.
Ph.D analysts want high-paying jobs that have a future. No doubt you can find a few (relatively few) such people to work with real estate valuation analysis. But they will probably be few and far between - because there is more demand for such individuals in other industries.
It would take 5-6 yers minimum to build a limited population of such individuals. And they would not be able to handle the workload of building Price models for every City, County and Neighborhood in the US.
3. In the meantime, the poor appraiser these companies want to sell their product to has to be trained. In fact, they are probably going to have to be certified, I bet. We currenlty have a USPAP requirement that appraiers be competent at using whatever statistical methods they are using.
4. Price Models to be useful need to be updated at intervals. I wont' go into the ways this can be done.
5. Price Models have to be available for the past 10-20 years, for IRS appraisals.
6. Price Models have to be available for different property types: SFR, Duplexes, Triplexes, Quadplexes, 5+ unit ...
7. You may find that the existing Price Models have major defects and have to be updated. In fact, a whole series of Price Models may need to be updated.
....
OK, in my conclusion, a useable product such as the virtual Night Hawk (currently classified as "vaporware") is mostly likely out of reach, way out of reach.
====
The only solution I see is for appraisers to learn R and the relevant R packages, so they can tailor them to whatever need they have. It would take many years to develop a decent size population of such appraisers, or valuation engineers.
Last edited:
