The Human/Machine Valuation Team

The question of which is better, human or machine, has been around for a while. Which plays chess better, the chess master or the computer? Which carries on a better conversation, any human at all or Siri? Which drives more safely, the human or the automated car?

In most cases, experts say, the best outcomes are achieved not by either the human or the machine but by a combination of the two. Working together, human and machine produce a better result than either can working alone.

Automation has already found its way into appraisal in the professional recertification courses that appraisers take every two years. Appraisers have their choice: the instructor-led course that meets in a hotel conference room at a fixed time or the automated online course that the student can work through at any hour of the day. The advantage of the instructor-led course is the opportunity to meet with colleagues. The advantage of the automated online course is the freedom to take on the course material at the most convenient time. The online course won't let the student advance until the student has demonstrated mastery of some part of the material. In terms of learning, that is probably the better thing. The best of both worlds might be an online course that also allowed the students to meet in a Zoom meeting to discuss their real world experiences that relate to the cases presented in the course.

Automation has come to cars. It edges in little by little. It used to be that driving involved shifting gears. The demand for that kind of car is gone. Today, new cars make it difficult to so much as change lanes unless the driver has signaled to move left or right. The car itself will apply the brakes if it is about to crash. We are made safer, even if the car is less fun to drive.

Automation reaches medicine as well. As much as anyone, doctors take professional pride in their skill at diagnosis and treatment. The thought that automation might do a better job is offensive to doctors on a personal level. Yet the Symptom Checker at medical Web sites embodies a level of knowledge of uncommon conditions that no human could memorize. If ten patients worldwide suffer from Santilli Syndrome, what is the chance that a general practitioner would recognize the symptoms when the eleventh checks into the waiting room? For common ailments, certainly the general practitioner's knowledge is adequate. For uncommon ailments, not so much. The best outcomes in many cases are the result of the doctor working with the online service, so that a patient's serious condition is diagnosed.

The list goes on - automated auto repair diagnostics, automated tax preparation, automated chat bots, Siri, Alexa, and more.

The methods used by appraisers to do their work fifty years ago are almost laughable by today's standards. Inspecting the building hasn't changed much. But there the similarities end. In the 1970s, researching comparable sales for a commercial property appraisal involved leafing through fifty back issues of a real estate publication that gave a property's address, its selling price, and not much more. Finding comparable sales involved driving around town for days to find out whether a particular sale was of an office, an apartment, or what. Comparable rents came from an appraiser's files or from some other appraiser, using the barter method. Capitalization rates were gotten by algebra, not from investor surveys or brokers' reports.

Today we access data from services that allow us to find the best comparable sales, fully researched, in minutes, as well as comparable rents by the dozens and capitalization rates from specific sales.

The raw material for a valuation has gone from low quality, time-consuming, and laborious to high quality, efficient, and instantaneous.

Raw material is still only raw material. A high quality work product requires analysis. For that, what is needed is a human/machine team.

Zaxia is a semi-automated valuation model for commercial real estate. It requires a human user, who has the choice of letting the automated system control all the variables involved in the calculation of property value. Zaxia will then make its best estimate of market rent and of all the other variables - the vacancy rate, the expenses, and the capitalization rate. It will choose the comparable sales. But Zaxia has found that the most accurate valuations are made when the user fine tunes the inputs, especially in the assignment of the market rent. The user can go back and refine the inputs again and again. The result is a work product that is part human and part machine, achieved in a far shorter time frame than has otherwise been the case. Zaxia allows its user to "cut to the chase."

The future of valuation is all these things. It is improved research. It is a willingness on the part of the analyst to use better methods. And it is improved analysis not simply by a better trained human or a more refined machine but by both. It is improved because it is the product of a better human/machine team.