Is there something missing — in the appraiser software designed to fill in the UAD 3.6 FannieMae/FreddieMac residential database form?
An appraiser is charged with analytically predicting the “most probable” selling price. The science is straightforward: 1) Identify the problem (scope); 2) Select the data; 3) Apply a model/algorithm; and 4) Pick a certain single-point number.
The UAD 3.6 is a receptor for data, information, results, and opinion. An appraiser cannot directly input into the official “not a form.” The appraiser must create, or buy a third-party input form as a go between. Here, we discuss what the several third-party products do. They:
- Provide a place to gather data, whether online, from a data collector, or personal observation.
- Select a path of fields to be filled, depending on property type, time, location and features.
- Optionally: offer algorithm(s) for the appraiser to select and accept analytic results.
- Replicate the required certifications attesting to truth and accepting liability.
More simply, the appraisal software enables data entry and analytic help. The data entry part is not a problem. The missing fix lies in the analysis part.
The analysis part in appraisal standards (USPAP) is called “Appraisal development.” The appraiser (the modeler), does two things: 1) chooses the algorithm to use, and 2) inputs the inputs.
An appraiser selects the algorithm for a computer to run, based on the modeling decision. This can be the “approach” to use, the logic to apply, or the function to use. This is the job of the appraiser – the analyst. The analysis part. Appraisal development. Standard 1.
So what’s the needed fix?
The problem is that software sellers have to sell software. They have only two product elements to sell: 1) the input fields part, and 2) algorithms to apply.
An appraiser-analyst makes numerous modeling decisions. And these decisions involve several major parts of the development process. And each of these larger development parts has specific sub-parts. Needing judgment. Expert judgment. Expert judgment is essential to the science of data analysis.
As an example, we consider a central element of appraisal: market analysis. And market analysis itself involves two elements: data selection and trend estimation. Trend estimation in turn, has ‘elements.’
In our Valuemetrics curriculum, starting with the introductory “Stats, Graphs, and Data Science 1” class, we note that there are six specific decision points for price indexing (“time adjustment”). These each involve analyst judgment. These six points also constitute an underwriter/reviewer checklist.
We provide the code, the AI prompts, and the needed understanding. Programming not necessary.
The problem that needs fixing is that software vendors want to provide analytics that are button-easy and sellable. The catch is — it is not reasonably possible to put every needed analyst judgment call into a checkbox, a pick-list, table, or correct graph. Not possible.
The analysis (appraisal development) “decision tree” is longer and wider than the UAD 3.6 decision tree. Much longer and wider. The software would be far more convoluted and complex than the UAD 3.6.
Analysis software without analyst competence does not improve things. But it can complicate things.
Appraiser competence can only come from education on these specific decision points — the judgment calls. Software cannot anticipate every possible property/market nuance!
The missing fix must be filled by the analyst-appraiser. That’s what we get paid for. Thinking.