Traditional Appraisers Are Screwed ...

[RCA]

Look, please understand, this is still a ways off. Advanced appraisers are still in a good position long-term. They will be needed to manage the new systems as they come online. -- However, traditional appraisers are going to be a dying breed.

I can clearly see this now with Meta's free SAM 2:

GitHub - facebookresearch/segment-anything: The repository provides code for running inference with the SegmentAnything Model (SAM), links for downloading the trained model checkpoints, and example notebooks that show how to use the model.

The repository provides code for running inference with the SegmentAnything Model (SAM), links for downloading the trained model checkpoints, and example notebooks that show how to use the model. -...

github.com

What does this mean?

1. With RCA, 80% of the valuation is capable of being objectively automated to high precision.

2. The other 20%, more or less, at least in the San Francisco Bay area, with its high land prices, is, with respect to the subject property, subject to subjective opinion on the value contribution of unmeasured features.

3. However, about #2, we now have the ability of computer software to extract specific features from MLS photographs, and treat them as measured variables via their numerical RGB color value matrices of pixels. New software such as Meta's SAM are pretty good at this extraction, and the valuation of extracted photos can be accomplished with CNN. So, we can go through a stack of property MLS photographs, automatically categorize them by type (exterior vs interior, rooftop, front of house, back of house, kitchen, living room, master bathroom, ...), and then also automatically extract items from the photos and appropriately label them, e.g. toilet, sink, shower, bathtub, floor, refrigerator, gas range, oen kitchen sink, kitchen counter-top and so on. Then these photos, as grids of numerical RGB pixels, can be run through CNN (Convoluted Neural Network) programs to be valued on their impact on sale prices. The subject can also be ranked based on its photos.

Conclusion: The subject property valuation is currently capable of being nearly 100% objectively valued by current technology. The process needs to be managed and monitored by a knowledgeable appraiser, but the appraiser won't do the vast majority of the work. It is a question of time and resources. Processing photos, even with a good computer, can be time-consuming. However, efficiency will improve over time. No surprise.

I am currently working on such a system --- and it will be executable online by others - demo data only.

 

[Digger88]

All the more reason to set up a Google Business Profile like my mentor in this endeavor @Functional Obs showed me how to do. He is 100% private work only and frequently turns down work or refers them to me and another guy.

As far as tech goes I don't doubt it though it still can't and never will address issues that you uncover only on a onsite visit like smells etc. And for the relatively cheap cost of an appraisal you get someone to sue if something goes wrong as evidenced by the fact waivers haven't taken over as wildly as some thought.

 

[RCA]

All the more reason to set up a Google Business Profile like my mentor in this endeavor @Functional Obs showed me how to do. He is 100% private work only and frequently turns down work or refers them to me and another guy.

As far as tech goes I don't doubt it though it still can't and never will address issues that you uncover only on a onsite visit like smells etc. And for the relatively cheap cost of an appraisal you get someone to sue if something goes wrong as evidenced by the fact waivers haven't taken over as wildly as some thought.

MARS + RCA is 80% of the value, and SAM + CNN is the other 20% ( in California, with its high land values ). But MARS + RCA + SAM + CNN is actually a lot of work at this point, requiring a high skill set.

What I think appraisers are unaware of is how much work is actually being done on housing at the highest levels. It's a lot more than you think, not at all with respect to appraisers - but with respect to AVMs and companies like Zillow, Redfin, House Canary, and so on.

I mean, Claude, ChatGPT and Grok are completely fluent in using the most advanced statistics for analyzing house prices. You can start a conversation with them yourself.

Many of the highest-level mathematicians and statisticians in the world are quite familiar specifically with the valuation of homes. The issues of concern on this forum - are a concern at the highest levels in academia.

So, no, appraisers aren't the only ones living in this world of home prices. Not at all, not by a long shot.

I think the key obstacle here is this: Can SAM + CNN really replace the appraiser, even surpass the appraiser in valuing the impact of these almost intangible features such as condition, quality, functional utility, and aesthetics/design? Can SAM + CNN make the grade? The breakthrough this year is the introduction of MARS + RCA, which can do an excellent job of valuing the measured features of real estate that impact sales prices. That is about 80% of the value, at least in California. That leaves about 20%, which, as of today, I am sure can be covered by SAM + CNN -- given some more time to improve efficiency and adapt to the higher cost and training.

Next year, expect to see movement in this direction. I am sure a few on the forum, such as Chase Pursley, are already involved.

BTW, high speed processing of photos, will take a powerful AI system. A new Mac Studio M3 Ultra is mandatory for the appraiser. But he will likely have to spend money for extra AI processing on the cloud with AWS, THAT COSTS $4-$5/hour of processing. There are cheaper services that can get that down to maybe $1.50-$2.00/hour, but probably less powerful computers. There are services that do photo processing, but they are expensive. ..... There will be overhead.

 

[RCA]

BTW, if computers can completely handle residential valuation, commercial is certainly doable, but it is just more of a mess, it's simply more complicated because there are a lot of wierd special cases and anomalies, although the widespread us of the Income Approach makes things in some ways easier to detal with.

And yes, the rural and ag appraisers rightfully add that good quality digital data for their properties won't allow it. But it's just a question of time for them as well. Maybe, though, that is 20+ years off. I don't know.

 

[Fernando]

Technology has its limits.
Tesla can self drive most of the time and company recommends driver to stay alert.
Tesla still have problems at train tracks. Computer can't analyze many factors regarding tracks, train horns, etc and some end up tragically.
Lenders understand this with AI doing appraisers' work and is willing to take some losses.
Just like auto companies willing to pay out for some defective cars killing passengers. It's unfortunate but still profitable for the companies.