Is SmartExchange Not USPAP-Compliant?
Is Smartexchange Resulting in USPAP Non-Compliance?
This blog is likely going to stir up controversy. It is likely going to have a few folks calling me a loyalist to a la mode, or other similar things. Am I a loyalist? I do not see myself as such, but a la mode has been the only software that I have used in my career for residential form reporting. I do use ACI in the corporate setting as a reviewer and manager in my firm. I like and consider many of the staff at a la mode to be my friends, and I have had long relationships with them because of my time as a customer. I will always have an affinity with those folks, but that does not make me a supporter of CoreLogic one way or the other. I am agnostic when it comes to CoreLogic. They exist and as such I have no choice but to deal with them on some level. They are too big of a company in the real estate and valuation space to not deal with them on some level. The whole point of this piece is to discuss the acceptability to standards when dealing with sharing comparable property data. It is not a barometer for what CoreLogic means to valuation.
With all the social media chaos over the CoreLogic acquisition of a la mode, there has been some very loud commentary coming from appraisers. CoreLogic is not looked at in a positive light by many in the valuation profession, but that is not what this article is about. I wanted to deal with the issue of USPAP compliance and their comparable data sharing product, SmartExchange. I have seen and read many posts from others that claim it is not USPAP-compliant to share comparable property data among one another. I must disclose up front that yes, I did at one point receive compensation as part of the a la mode Labs group between 2007 and 2009. I am no longer a paid employee or contractor with the company. I am offering my opinion on the product to discuss and look at the issue of standards compliance.
What is SmartExchange? This is what a alamode posts on their site:
“SmartExchange is a nationwide appraisal network that puts property data back in your control…SmartExchange is a nationwide appraisal network that puts property data back in your control by giving you immediate access to pure, UAD formatted appraisal data. This level of data is unprecedented and is unlike anything you’ve been able to gather from MLS systems, public records, and other sources. It’s going to improve the quality and consistency of your appraisal reports.” [1]
Image Courtesy of CoreLogic – smartexchange
When UAD was pushed out by Fannie Mae several years ago, they let us all know that they would be monitoring and tracking what we used for condition and quality ratings to compare against other “appraisers in” the same market. In other words, they wanted to see if appraisers were materially misrepresenting data to support bias. a la mode has created a tool that will allow all participating appraisers using their software to share the ratings used so that one may look at what the peer group is saying in their reports. It is possible to opt in to the program or opt out.
So, is this something that appraisers can and should use? I have reached out to several USPAP experts and walked through a series of questions with them and with other practicing appraisers. I have also done my own research to come to my own opinion. Much of social media is asking an important question about this type of technology: Is comp data shareable or does it fall under assignment results which would be deemed confidential? Let’s look at some definitions:
ASSIGNMENT RESULTS: An appraiser’s opinions or conclusions developed specific to an assignment.
Comment: Assignment results include an appraiser’s:
- opinions or conclusions developed in an appraisal assignment, not limited to value;
- opinions or conclusions, developed in an appraisal review assignment, not limited to an opinion about the quality of another appraiser’s work; or
- opinions or conclusions developed when performing a valuation service other than an appraisal or appraisal review assignment.
Physical characteristics are not assignment results. [2]
CONFIDENTIAL INFORMATION: Information that is either:
- identified by the client as confidential when providing it to an appraiser and that is not available from any other source; or
- classified as confidential or private by applicable law or regulation. [2]
What we have above are two definitions central to this discussion. The first, Assignment Results, essentially draws one to understand that assignment results are opinions or conclusions that the appraiser supports through the course of the assignment. The obvious question is then, “Aren’t condition ratings and quality ratings assignment results?” On the surface I felt that they may be but have further talked with others and looking at all the language in the definitions. The reason I initially felt that the ratings might be assignment results is that assigning a property a rating requires judgment, and as such judgement is an opinion or some type of conclusion as we use the words.
After speaking with a few experts on this very issue, I changed my opinion. Most importantly, what I was drawn to is the last line in the definition of assignment results:
“Physical characteristics are not assignment results”.
When it comes to comparable data, we are all using third-party sources to assign the condition and quality ratings for each property. A couple of the experts, many being residential practitioners, go back to the definition for condition and quality that Fannie Mae developed. Many felt that there is little room for opinions of even conclusions, that the ratings are -defined, and all the appraiser is doing is assigning to the most obvious tier. So, the sharing of comparable data that is supported through MLS databases and other third-party sources, readily available to our peer group, does not violate USPAP.
Where we did get into some interesting discussions revolved around the subject property. There is certainly confidentiality involved when dealing with the subject property. Let’s say that we are doing a refinance transaction. There would be no way to support assigning ratings other than an onsite inspection where the information gathered was only available to the appraiser as part of the assignment. What made the conversation interesting was that something came up that I never thought of before. What if you find out something about the subject that could create a misleading situation to the client?
Let’s say that you do a divorce use report. 3 months later the home goes up for sale and sells. The MLS data indicates that the home is in much better condition than you saw when you did the inspection. Upon further research to use it as a comp, you learn that the MLS listing incorrectly indicates it is a C3 versus the C4 that you felt it should be from the previous information that you have. Or, the square footage is wrong. What should you do now that you are faced with what you know to be reality versus what a normal peer would be left to conclude form the data available? What I gleaned from these conversations was that we should not be using information that only we have privy to through a previous assignment.
USPAP requires that we do not mislead in our reporting, but it also says that we must stay beholden to assignment confidentiality. This a bit of a conundrum. If the difference in information is enough to affect assignment conditions, can we take on that assignment? If we do, how do we report what we know is contrary to what the overall market accepts as fact? I think the correct answer is that one must recuse themselves from this situation.
Getting back to smartexchange, this led me to contact a la mode and ask the question of whether subject data was being shared or not. What I was told is that it is only comparable property data. That the shared database contains only the comparable properties, and that our local database (meaning available only to you and your office) contains your subject data but that is not shared with the users of smartexchange.
Is this really any different than calling up your appraiser colleague across town and asking if they have used a comp and what they considered it? I think the answer to that is both yes and no. It is not different in that colleagues are sharing info, it is different in that the users have no idea where the data comes from. When I call a colleague, it is normally someone that I respect and have a relationship with. With respect comes trust, but without being able to see who supplies the data, it does make me uneasy. I can see multiple uses of the data though, so I am able to see if there has been consistent codification of the comp over several users. At least I can qualify my own rating and write something proactively if I am using a different code level.
In conclusion, is smartexchange a problem from a standards perspective? I do not think that it is if it is simply dealing with comparable property data. It certainly can be helpful if the appraiser using it wants to compare the condition and quality ratings over a market area. Beyond that, it could save some time with data input, but that is something that the user needs to vet carefully and fully every time, and honestly, by the time I do that I am better off just entering that myself.
Footnote:
By Woody Fincham, SRA, AI-RRS, is based in Charlottesville, Virginia, where he is vice president and chief residential appraiser for the national valuation firm Valucentric. His work involves complex valuation assignments, including historically significant real estate and equestrian properties. He advocates on behalf of residential appraisers and is co-chair of the Appraisal Institute’s Residential Appraiser Project Team. His passions include listening to and creating music and riding his Harley-Davidson Road King.
“Assignment results: opinions or conclusions developed in an appraisal assignment, not limited to value;”
Lets literally take a stroll down the URAR:
Sale Financing/Concessions/Date/Rights Appraised – in non disclosure states, these would be confidential
Location – perhaps one of the most subjective opinions on the form
Site – Whether to include surplus/excess here or elsewhere is an opinion
View – highly subjective opinion
Design – based on my polling of appraisers on this issue, there are no hard/fast rules on this, and it is largely opinion
Quality and condition are obviously assignment results, I’ve never seen anyone attempt to say they are not “opinions developed during the assignment.”
GLA determination is based on opinions and conclusions of functional utility, market acceptance, adequate heating, public record accuracy, etc
Room count – is that small room with no closet a bedroom or office… opinion
Below grade – same as GLA and room count
Functional utility – clearly an opinion
Heating and cooling – are wall AC units central or no?
Garage – is an oversized stall a 1 or 2 car?
There is not a single line on the URAR that is not based on the opinions and conclusions of the appraiser. Many may blast past the reasoning (because they trust the assessor, et al) but that is an opinion and conclusion they have made. The counter is offered: “Does that mean you can’t keep your own database of comp for use between reports?” In so much as it is MY opinion from one report to another, of course I can’t “unknow” a thing. However, if more accurate data becomes available then I must discard the old and start new for the current report.
USPAP is painfully clear in its language, and no court would defend this Commode Logic.
AMEN, good on you. And also Protecting (our) data is a waste, it is implying that the info is ours, not generated by others.
Most of us are in states where disclosure is the LAW. The sales prices are published as a matter of record, MLS prides itself on accuracy and disclosure.
WHO is the prideful appraiser that tries to capture, only interpretation is allowed.
“Physical characteristics are not assignment results”
Focusing on this as the emphasis.
What is the purpose of sharing this data from the get go, it is to HARMONIZE it make it all similar.
The Tech realm and those reading it UAD etc DO NOT UNDERSTAND why there is a difference or HOW
THERE CAN BE A DIFFERENCE from one Appraiser to another on the EXACT SAME PROPERTY. So as we move into A.I. the future code trying to become HUMAN and thinking like a human it requires that data to be HARMONIZED i.e. similar to understand LOGIC.
Here is the problem with that theory and lets take Mathew, Mark, Luke and John. 4 folks focused on the same person yet their stories differ and are not in 100% harmony. How is that possible, because as HUMANS we all think and see things and convey them differently. So my point is this fast forward to today and the age of the MACHINE and TECH and we believe the interpretation should be HARMONIZED.
That is utter poppy cock nonsense. Taking this further if appraisers believe this is how the real world works with emphasis upon THE MARKET VALUE DEFINITION we are orphaning it and need to rewrite that definition with any such data harmony.
Why cause as a Realtor and an appraiser I can tell you this, you can get 4 buyers and they all see that exact same property differently, very differently on the very items we are now being told to HARMONIZE. So if the market see it different why are we being asked to nail down datapoint to such a degree that we are no longer mirroring buyer and seller activity because WE ARE DOING NOTHING HERE BUT BUILDING A MACHINE that desires to think like a HUMAN in order to take over our industry. That machine with UAD logic as the baseline just can’t get how or why there can be differences between appraisers when in FACT THERE IS and should be. I’ve provided historical evidence of that from even the Bible and to the current market today via buyers and sellers.
It’s data, nothing more than data and since when did a data field in a form become so important beyond the art of appraising and stating here is the market figures and market value. We are not just providing a value opinion these past 10 years we are utterly providing food to the very machine to become more human and for that A.I. system to learn how to logically think more and more like a human to take over the appraisal realm entirely.
You can thank UAD and the creators for this along with all these other firms like core logic for stealing data from the heartbeat of each of your markets.
Appraisers need to file a massive class action lawsuit as that data that built this entire system is their intellect and they should get paid for it in the form of pension and severance if the end game it WE ARE OUT it is in.
Brokers and Realtors are next and that shoe will drop even faster as Corelogic runs almost all MLS systems and all data is being ported as well. NAR has power to aid but once a new system is built there as well and the public is not longer having to pay a % vs. using an application on the phones most are hypnotized by GAME OVER.
The end result is THE DATA IS LEARNING, PAY ATTENTION IT’S LEARNING NOTHING MORE and the more it can compound this learning (massive data sharing and harmony) the faster the A.I. and new data systems can tap it.
Fantastic response. BTW – all readers should understand why you emphasized harmony repeatedly (& thank you for highlighting it). It is the new ‘buzzword du jour’ for Feds and the corporate hucksters they are supporting. Apparently LEVERAGING Big Data isn’t harmonious enough.
Automated mortgage processing that (insert tense here) harmonizes or is in harmony with other processes/millennials demands/Amazon-like efficiency/etc. has replaced leveraging.
Everything else you write is spot on except the term “AI”. AI does not exist. Period. It’s a myth. ALL that we have are computers that can answer “yes” or “no” extremely fast with further yes/no questions dependent on how the first query is asked and answered. The thousands or millions of permutations of how yes and no are used may make it seem like our computers have intelligence…but they do not.
They just have an extremely big database that has been ‘leveraged’ to provide answers that are in harmony with what is being specifically asked, very, very, very fast.
Don’t over estimate Alexa’s or Siri’s abilities. They are illusory…just like the concept of AI is.
For those in rural areas or less populated areas, SmartExchange may be meaningless, however big city appraisers encounter used comps all of the time. I’m working in a report now where a comp has been used 17 times (1143 S Nevada St, Oceanside, CA 92054), and yet its information is in no way consistent. Public records data indicate 804 sf today, but yet past reports vary from 804, 943, to 1,001 sf. How about the Garage/Carport (2 garage per public records)? It either has 1ga1dw, 1gd1dw, 2gd2dw, 2gd2cp2dw, 2ga1gd1cp2dw, 1gd2dw, or 1gd1cp2dw. How about the Heating/Cooling? I love how the property had central AC, then didn’t, then did, then apparently had the forced air unit (FAU) removed and a wall unit put in, then had that removed and the FAU was put back in (what ?).
You guys can argue all you want about the legalities of the system, but don’t gloss over the fact that much of the data is still garbage.
Seek the truth.
If you want to see how smart appraisers are, or how smart SmartExchange is, then enter the following address (1143 S Nevada St, Oceanside, CA 92054)? You will find 17 uses, and inconsistency galore. I like how the GLA bounced around, how the property had central AC, before it did’t, then did, then had a wall unit before again having central AC. Don’t forget to check out the Garage/Carport variances (were there 10?).
I will let you guys determine the legality of the system, while I sort through the garbage (our garbage) it contains.
Seek the truth.
It’s a great example! It’s also a multiple flipped property since 2015 I’ll assume the first $520k sale may be a fixer, selling for $620K 4 months. Then only a couple years or so later it flips again for $760k on 7/30/2018.
Nothing remarkable about that for SoCal RE EXCEPT CoreLogic’s own market trend data shows the median price in its area DECLINING steadily over the past 2-3+ years. PS-SF is now 932 in CoreLogic’s Realist, and it looks like a pool in the aerial Bing photo.
RealValue (Corelogic) is $799,023 as of 7/18/2018 in a range of from $695,000 to over $902,000 I rounded. CL is even more accurate than what I wrote. “Tight!” From a low end of range 8 1/2% LESS than it sold only two weeks later, to a point value 5+% ABOVE what it sold for, to a high end of range more than 18% above its actual sale price.
Make sense of THOSE numbers, lenders and regulators! Oh, and definitely qualify it for a PIW – after all it sold FHA for $620k back in 12/2015 about 4 months after it sold for $520k in 08/2015.
To each and every single regulator that thinks PIWS and AVMs are “ok” I have only one thing to say:
“Asshat”.
The other issue is the fact that properties do change. Remodeled, updated, garages added. additional parcels purchased. C’s and Q’s. The tech world thinking that properties do not change yet this is the AI world where everything is constantly changing is not thinking this through. A properties data may be good for several months who knows. Just depends on if they installed central air conditioning yesterday or tomorrow. The garbage in garbage out rule could be validated mostly by where the data is coming from. County records, realtors or appraisers.
Appraisers using data is limited to there resource plain and simple. Every market is different as one mentioned pertaining to non disclosure states. Texas is a non disclosure state however a vast amount of information can be found about properties. New Mexico is a non disclosure state where data is very limited however it is improving. The example is the home that was bought in foreclosure then almost torn down and rebuilt. I guarantee you the reports that have the least amount of similar data than others is probably the most accurate data. This means they independently keep records and have a large data base AS WELL AS THIS IS THE MOST IMPORTANT OF ALL THEY PROBABLY APPRAISE IN THEIR OWN MARKETS. Remember when the lending industry decided that appraisers shouldn’t work in their own markets. Boy you want to talk about the point when turn times went from a week to three weeks and everyone is scratching their head. Give me a break.
I do like smart exchange just for my own personal data and I rarely rely on other data in smart exchange but it does make data entry faster which is one of the main issues we drag time in appraising. Data entry, constant calls from AMC’s are the two biggest issues we have today. Smart exchange does make data entry faster
Well… I type 100wpm or faster which was a quite easy skill to develop with fun old Typershark online gaming, and never get any calls from amc’s.
Now, on to the actual problems which are not self imposed. The problem with big data systems is that it’s easy to set them up but increasingly difficult to maintain cohesive current data as time passes. Was at the hospital today and although we’re under totally different insurance, their complex data systems still notes our old employer, even though we’re covered by the new one. Sure, update that data, again, and again, and again, maybe one day it will actually happen. Big data is not just a central repository of data, it’s mirrored and segregated and mirrored again on a vast network of servers that are unlikely to be in the same location or under cohesive management in the first place… Until they have D-Wave computers managing all of that, such persistent inconsistency will continue because administrators will never have an adequate hierarchy or agreement regarding who’s data is most current, who’s servers are most secure, and where updated and introductory data flow must through because they seek to avoid bottlenecking and bus overload. There are after all, as many entry points for data input as their are data points themselves.
Big data. Big joke. Put a call in to test the alamode systems at the next whitehat gathering and watch how easy they make it look, it’s a rule of thumb these days. Computers are not some grand mystery, they’re just tools and like any tool, in the hands of an inexperienced mechanic they can do more harm than good.
I’d bet not a single licensed appraiser or lender or amc manager whom promotes these items knows how to code themselves. Tech people are not as smart as generally given credit for, they stand on the shoulders of giants and we’re approaching ‘idiocracy’ at a breathtaking accelerating pace.
Where is that time machine again? Learn to code I guess, what more is there to say?
#!/bin/bashcat SmartXchangeDatabase.txt
if [ $? -eq 0 ]
then
echo "The data has been successfully updated"
exit 0
else
echo "The appraiser is a moron and data rejected" >&2
exit 1
All good observations Woody. Great article and thank you for taking the time to write it.
“When it comes to comparable data, we are all using third-party sources to assign the condition and quality ratings for each property.”.
No, I am using my own observations and opinions which in turn may be derived from third party guidelines, but the interpretation of those guidelines even for “absolute” ratings is my professional opinion. The same holds true if I am guided by M&S (never) or Craftsman or other reputable sources including area builders.
My understanding is that SE never shows adjustments made. That’s fine, but then how does it communicate that two properties – both rated C3 may still have had $20k; $50k or even hundreds of thousands of dollars difference in condition within the same rating?
I suspect that like most tools there is a way and use in which SE could be used and still result in a compliant report. I also suspect that human nature being as it is; compounded by demands for ever faster processing, that many appraisers won’t take those extra steps.
It was an awesome article, Woody is the man. I just shortcut the argument by entering unique data for each and every comp, every time, even if I have used it before. What makes the entry consistently unique is the entry is relative to the subject, and therefore is irrelevant when compared contrasted or applied elsewhere. Typing fast is so easy, the manual comparison and subsequent analysis can be the most defensible point for an appraisers analysis, I fail to understand how appraisers benefit in terms of quality work product by cutting out the most essential part of analysis, relevant and direct comparisons of comps vs subject. This knowledge gained from only a half hour of manual type up, photo and mls review, and cross referenced data of gridded comps often provides the best framework for all other analysis points. Or just outsource the duties of typing and data entry so you can get back to ‘appraising more’, whatever that means, nonsensical jargon from the tech crew…
I suppose if one does not understand the underlying process and function, the infinitely repackaged tech may certainly still hold some sort of wondrous amazing power to the laymen. I prefer my tech to be painfully static and managed by people whom can not actually write code, with cleverly applied permissions structures so I can feel good about subscribing indefinitely and never receive a finished product. Do you offer that in analog?
On a side note, for those that don’t understand, giving permission for A la mode to enable SmartExchange on your computer, is giving them approval to not just reflect from that date moving forward (your data), but rather ALL of your data from YEARS PRIOR. If you work at a medium or large appraisal company where the system is networked, what is the market value for the potential tens of thousands of reports on file (again dates back years)?
Seek the truth.
Does anyone know if Mercury and all other portal softwares will implement a shelf life limitation for stored data? It occurred to me that I’m still able to look into full detail for reports completed something like ten years ago… Suddenly it’s not fun to use these third party systems anymore. Eventually an appraiser is going to argue the right to be forgotten, like that’s where this is headed in the bigger longer term picture.
Tech is so amazing, I’m glad we outsourced all these jobs overseas, crushed domestic production facilities, pushed a literal continent of ewaste forward, compromised privacy on an epic unheard of scale, and bred an entire generation of people with no manual labor skills whom prioritize convenience over liberty. What could go wrong? I guess now we’re supposed to debate the merits of it, how specifically to best use it, or what exactly is happening here? I’ll take an appraiser who knows how a furnace and electrical box works anyday over someone who understands coefficient application and subscribes to big data systems. We need a new stereotype, or something, whatever. Learn to code.
What Is Artificial Intelligence?
Artificial intelligence (AI), sometimes known as machine intelligence, refers to the ability of computers to perform human-like feats of cognition including learning, problem-solving, perception, decision-making, and speech and language. Early AI systems had the ability to defeat a world chess champion, map streets, and compose music. Thanks to more advanced algorithms, data volumes, and computer power and storage, AI evolved and expanded to include more sophisticated applications, such as self-driving cars, improved fraud detection, and “personal assistants” like Siri and Alexa. Today, medical researchers are using AI to develop technology that will detect a range of diseases, improve radiology imaging, fine-tune radiation treatments, simplify DNA sequencing, and advance precision medicine for more individualized health care.
Source: https://www.psychologytoday.com/us/basics/artificial-intelligence
Will Artificial Intelligence Soon Tell Us How To Live?
Source: https://www.forbes.com/sites/cognitiveworld/2019/01/27/will-artificial-intelligence-soon-tell-us-how-to-live/#4090339d2938
At the same time, the AI is evolving now in an unprecedented way, and the interim results of this transformation are expected to be truly impressive: according to the report, 79% of the banking sector representatives anticipate that in the next two years, technologies will advance to such an extent that artificial intelligence could work in financial organizations equally with living people as an employee and reliable advisor.
Source: https://www.fintechnews.org/artificial-intelligence-in-banking-industry-conversion-to-genuine-benefits/
Over the past couple of years, Artificial Intelligence (AI), including Machine Learning (ML), has gained traction by businesses that deal with large amounts of data to reduce human error and improve operational efficiency. With AI, computer systems are able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, language translation, and decision-making. And with ML capabilities, they also have the ability to process large amounts of data (structured and unstructured) from various sources and recognize patterns in the data to identify opportunities or risks.
Major areas of AI/ML application to the mortgage industry may include identifying anomalies, assessing risk, exploring non-credit bureau data to enhance prediction of loan performance, and answering customer questions (e.g., search tools, improved guides, and chatbots).
Source: http://www.fanniemae.com/resources/file/research/mlss/pdf/mlss-artificial-intelligence-100418.pdf
Artificial Intelligence (AI) prediction models. This study attempts to analyze and summarize AI techniques, which gives insight to application of various techniques for prediction related to property valuation. Comparison of various techniques shows that Artificial Neural Network (ANN) and fuzzy logic are better suited if attributes and model parameters are appropriately selected.
Keywords: Real property, value forecasting, artificial intelligence, neural network, valuation.
Source: http://www.enggjournals.com/ijet/docs/IJET13-05-03-087.pdf
Pay attention to the fuzzy nature of logic cause like I said IT’S LEARNING, a bit or 1 or 0 reverse engineered into a database needs to know how to apply variance to the emotions and variances including all this inconsistent UAD stuff we are to aid by HARMONIZING this dam database even more going forward.
What you want to pay attention to is the movements of core logic for one as they will buy up key companies or startups with good A.I. and with access to data, historically this has proven to be true. All pay attention to NN and ANN these are neural networks to be combined with all this later. Remember the property is static and lets say 150 million homes in the US. Once they all get loaded up and they can link up the datapoint the variances are next, cycles, condition, market trend, cost of money, etc etc. From there just like the A.I. that lost the to worlds best Chess player originally that A.I. came back later and kicked his ass. So like I said we are feeding this machine that right now humans are empowering and just like the horse and carriage the car took over, just like the factor worker the robot took over and just like the appraiser the algorithm when it reaches it’s A.I. potential (ability to do what a human does) it too will take over.
P.S. for those not believing in this stuff, google how they are now using/considering DNA for storage of data. Yep that is how far advanced we are into this rabbit hole that the dam machine now wants our DNA to store data and the dumbass humans allowing this are insane.
https://www.technologynetworks.com/genomics/articles/dna-the-future-of-digital-data-storage-299234
Thanks
Alan
Great post Alan.
I do pay attention to CoreLogic. Just as I paid attention to FNMA’s fatally flawed Collateral Underwriter Process since its patent application was submitted, and before it was adopted as policy in January 2015.
CoreLogic is INFAMOUS for providing both incorrect and conflicting data concerning physical attributes of property. Even when the conflicting data is from their own alternate products!
One of the earliest computer axioms still applies today.
GIGO “Garbage In-Garbage Out.” What they are really doing today in many so called “AI” scenarios is lowering the bar of acceptable accuracy, and then pretending that subjectively scrubbing the data to produce the desired results still maintains scientific reliability.
There is no such thing as cognitive ability in computers-yet. All they do is say yes or no very, very, very fast to many sequential questions.
OH Yeah! Detailed tech topics!? That’s a treat for this forum. Please allow me to bridge this gap with best of ability efforts.
AI is a popular buzzword of the day, the race to automated superiority. AI is the necessary ingredient right? Presumably because the people who code can’t get there on their own with programmed parameters and we need a higher intellect with more computational power than the most advanced computer on the planet, the human brain, to reconcile all this data? Perhaps the tech guys are taking weekends off, have a 30 hour workweek, just can’t find the time. Or perhaps the dream of complete automation in real estate is a wingnut theory in the first place. Big tech and big data is suffering the publish or perish symptoms right now. They claimed a massive market stake and now are seeking to justify their presence in a more legally compliant and more widely applicable setting after the fact.
Appraisers are suffering from a similar identity crisis where individual appraisers have confused themselves with wall street investors and tech companies, promoting automation, absent of the fundamental understanding of how the tech works in the first place. Reaching beyond the appraisers appropriately limited scope of work participation as a check to the balance and independent analyst point, operating apart and independent from the larger integrated mechanics of the interested companies. The push to use integrated systems is synonymous with abdicating our role as independent checks to the balance.
Big data systems like Zillow rely on the theory of computation. This is reliant on 3 parts, language theory, computability theory, and computational complexity theory. The problem with real estate integration and big data is the human factor. It remains impossible to accurately move to complexity theory application (3rd stage), when during the assimilation of data through computation (2nd stage), after hopefully interpreting and mapping the language correctly (1st stage), the human and real property behaviors remains largely unpredictable without updated data on that behavior and states of condition. Don’t believe the hype, without human presence the data modeling systems will never work in real estate. A human must be there to input current data otherwise the complex calculations will be based on inadequate or fictitious data, aka in appraisal; stale data. The rush to AI and big data is a consequence of a myopic understanding of how tech works, as the majority of decision makers hefting tech around left and right do not actually understand coding themselves. Constantly dressing the same old tech up in a new package, perhaps with some clever new coding but really never properly managing the question of what is reducible data and what is irreducible data.
Right now, somewhere, a tech programmer is hunched over his pc smoking the keyboard with code entries, while some corporate ceo is whipping him in the back telling him to get it done if he wants a job tomorrow. Not surprisingly, he is able to dress something up, integrate language, slap a graphic on it, and even publish that work, all the while never actually having to withstand adequate measures of peer review prior to or after publishing. (Who has time for peer review of tech with this much demand?) It appears Zillow and their counterpart companies spends more time and money on pr and image control than R&D, actually tacking the question of how to have adequate computability without actually having a current measurement.
The hubris of tech companies and those whom promote such, is the notion that the available public record and sales data is good enough, no human appraiser is necessary, believing firmly that success is achievable through reactive principals of available data interpolation tools. They believe their systems can be programmed to be reactive to apparently reducible data, therefore make more accurate predictions after that data is present in the computation stream. The problem with that in real estate is that because the data is stale and early results will be calculated prior to reducible data cleaning or correcting, someone would have to lose before the next party could theoretically benefit from cleaned up interpolated data. And it’s a recurring systemic concern which cumulates with time as all data will eventually become stale, often with exponential unpredictable differences regarding how long that data takes to become stale, aka the X factor of human behavior. It is truly remarkable that FNMA is not grasping these concepts in their race to advocate on behalf of institutional investors, and have apparently abdicated their duty to demand irreducible data, aka; full service human appraisal inspection. It will be painfully entertaining to see what gimmicks they come up with on the new forms to try and solve this unsolvable problem which is the fact that human behavior and real property conditions constantly evolve and change as swiftly as new technology and building and lending tools are made available.
Welcome to the brave new world of the digital age. It ain’t your grandpa’s vcr anymore, even if that seemed new to us at the time and we still remember that. It’s ancient history as computational power steams forward. I’m like John Henry over here, scribbling notes with pencils and form filling manually while everyone is telling me to automate. Not going to happen. Knowledge is power. We’re not there yet, not even close. If anyone bothered to listen to Elon Musk on the matter, it would be a wise decision to slow this all down and let tech systems develop for a very long time in less consequential fields before we attempt to implement them into such crucial social infrastructures.
whenever we fit the data, we should expect 2 types of errors 1) reducible errors 2) irreducible errors
reducible errors: this is encountered due to either “squared bias” or “squared variance”. (refer to bias variance trade-offs for supervised learning for more details)
irreducible errors: it is an error that we cannot reduce by choosing a better model. it is due to randomness or natural variability in a system.
https://en.wikipedia.org/wiki/Theory_of_computation
I might have confused reducible and irreducible in the post language. Oh well, I’m not a tech guy, but I’m sure you’ll understand. Cheers.
Baggins,
Your more tech. than most as you looked it up on WIKI.
Too bad wiki didn’t have an appraisal dictionary which says the same thing (Good Comp v Bad comp).
The AVA’s will suffer their reputation and demise naturally. When Appraisers use both anecdotal comparison and statistics our credibility will improve, as well as our necessity
I still can’t find a cell phone I like. I’m holding on to this 7 pro like my life depends on it. This analog call blocker has a 10,000 number capacity. When I’m away from the desk I am not consumed by tech. Life is good.
Now copy and paste that educational post into your letter to the Feds on deminimis Baggs!
You just explained why snake oil is snake oil. All most of the rest of us know is that snake oil isn’t good.
Thank you Mike.
That might make a good article, thinking about it. Everyone is tossing around big tech language lately, but often I wonder if they really actually understand what they’re talking about. They’re good at arguing about it but do they actually understand the details of the subject matter. Individual casino dealers are subject to more oversight than these tech nerds, and they’re only half as dangerous.
Letter submitted today. Something about consumer protection, something something.
Thank you.
It really would make a good article. Interesting aside. USPAP requires that we support our adjustments and conclusions. Some purveyors of magical ePixie dust & Big Data solutions suggest that we can / should now be able to prove our adjustments and conclusions. All via the God of Science.
Few remember that the human brain is and always has been the best computer ever created. It has a life long accumulation of data that it reconciles on a daily basis, with or without our direction. It’s a bad & arbitrary mistake for our regulators to dismiss it (life & professional experience) out of hand for “science” that has also been arbitrarily tweaked and massaged.
And just think it all started with an answering machine
Appraisers may be needed more than ever once big data replaces the appraiser in the lending process. Buyers, sellers and borrowers will now want to dispute their denial for a loan due to a bad report provided by big data. And that is where an appraiser can capitalize. I hope anyway
They’re probably going to need a brand new special insurance and regulatory structure for that. Ha! The results of new and influential business structures in our modern government is predictable, inevitable regulation. It would just be easier to skip all that and continue to use traditional full service non integrated appraisal services. They’ve invested ten hundred million man hours and one hundred quadrillion trillion computational hours to trying to accomplish what we do in our sleep. lol!
Exactly. There is no way they can better replace market experience with big data. Unfortunately it’s a race to the top who will be the one stop shop for lenders. Yes they better have a great insurance plan.
Baggs: “The human brain consists of about one billion neurons. Each neuron forms about 1,000 connections to other neurons, amounting to more than a trillion connections. If each neuron could only help store a single memory, running out of space would be a problem. You might have only a few gigabytes of storage space, similar to the space in an iPod or a USB flash drive. Yet neurons combine so that each one helps with many memories at a time, exponentially increasing the brain’s memory storage capacity to something closer to around 2.5 petabytes (or a million gigabytes). For comparison, if your brain worked like a digital video recorder in a television, 2.5 petabytes would be enough to hold three million hours of TV shows. You would have to leave the TV running continuously for more than 300 years to use up all that storage.”
https://www.scientificamerican.com/article/what-is-the-memory-capacity/
Big Data? FNMA; FFIEC, TAF, AARO and State Regulators y’all don’t have a clue what really “Big Data” is. Every single appraiser in the country is walking around with far more Big Data storage than any software designer ever imagined. It’s application is no less scientific than any arbitrarily ‘scrubbed’ data.