Will appraisers be replaced next year with big data provided by companies like Corelogic, Zillow, Cubicasa, and Open Door?
Well, no. And yes. I entered the profession in 2001, and one week after starting my new career I was stunned to find a front page article that laid out the end of the appraisal industry within a year. The culprit? Big data and appraisers who unknowingly
allowed companies to data mine their appraisals, (The content once belonged solely to the appraiser and use of the content for anything but the purpose stated by the appraiser was prohibited) but that's a story for another day. Today I want to talk about why
does big data and algorithms continue to fail to be accurate?
First a simplified explanation of what big data is and where it comes from. Big data in the appraisal industry is a compilation of everything real estate. It is
data mined appraisals uploaded through software companies, the entirety of Fannie Mae and Freddie Mac's appraisal portfolio, current and past mls data, county records, national real estate trends, local trends, and more recently, third party property data
such as building sketches, all input into an AI system that analyzes it in seconds and spits out a current market value. The key to accuracy is the algorithm(s) it uses. And in a subdivision of similar quality and condition homes, it can
be accurate. But it isn't always. Why?
I have explained for 21 years why online valuations aren't accurate on properties with non standard amenities, the easiest explanation being that if an appraiser failed to consider non standard amenities in developing an an opinion
of value for a property they wouldn't be accurate either. But the losses incurred by Open Door this year and my direct knowledge of homes they overpaid for last year when the market was increasing had me wondering how their model is so far off on a simple
tract home. The answer is they have a faulty algorithm. It may be as simple as they built their algorithm during a historic market increase and the AI determined the increase would continue, or the human component influenced the AI by buying properties above
the suggested value. In any case, like the platforms before, it failed to be accurate on a large scale. Imagine big data, algorithms, and private enterprise as the only means of valuation in a declining market and how that would affect lending abilities and
home prices? Yikes!
Markets change, neighborhoods change, buyers preferences change. What increased the sales price by 5% in 2010 may not affect the sales price today. Algorithms and AI can only consider data after enough of it has been compiled to affect
the algorithm. Any change to that and you end up with home values being based on the last house that sold. As appraisers we analyze the last hold sold, contact the selling agent, look through photo's, drive by the property, etc. and incorporate the result
in developing their opinion of value of your home.
The appraisal industry isn't the only one experiencing big data pains, and certainly isn't the 1st. In fact, big data has been around since the early 1980's. I embrace big data fully, and even wrote a big data program in the mid 1980's
(DB3 anyone?). But big data has it's faults, and those faults haven't changed much in 40 years. You ever wonder why your local hardware or big box store is always out of something? It may be because they use big data to order for them. Big data orders based
on what is selling, and if at the time of inception there wasn't much inventory on an item, sales would therefor be low, and big data can only order off that data. Example. An electrical supply house installed a big data ordering system. At the time they had
just sold all of the most popular romex wire they had in stock, 14-2 with ground. The computer didn't order any because it had no record of selling it and it took a couple months before a human was able to override the system. It was 1988 and they lost over
3 million dollars in sales, as well as several customers. Today we have algorithms all over the world basing what is needed on data acquired during the pandemic. Inventories are undersupplied and staffing is shorted today because a computer is telling humans
what is needed and humans aren't being allowed to override the systems.
There is a place for big data. The more data we have the more accurate an appraiser can be when comparing properties and market influences. When I started in this industry all the data was obtained in a printed book, on a black computer
screen with green writing, driving by a property, and calling the listing agent. Today we have multiple sources of photo's, permit records, assessor records, etc. at our fingertips which help, but big
data cannot (yet) replace a physical inspection by a knowledgeable appraiser who can consider the quality of materials, condition of the improvements, location influences (does it really have a view?), and functionality of your home.
So no, big data is not replacing your trusted appraiser next year. But yes, if appraisers stop providing more accurate opinions of value than big data does, then they will be replaced.