房地产估价师和评估师

anon (极有可能) 6 months ago

every single appraisal done for a major bank is given to machine learning researchers

0 0 Reply

Josh T. (低) 2 years ago

The data for automation is only as good as the input data. The primary source for this data comes from realtors and brokers who often don't input data correctly into their respective MLS systems. Realtors and brokers are required to input very detailed listing information into the system.

They also need to measure homes to ANSI standards and provide the year a property was updated, or if it was not updated at all. When I say "detailed information", I mean the following:

- 3” vs 6” trim or crown molding
- Custom crown molding
- Slate or marble tile vs builder grade
- Custom cabinets vs builder grade
- Level 1-5 granite/quartz/marble countertops
- Frigidaire appliances vs Wolf/Viking/Subzero
- Handscraped, engineered wood or Luxury Vinyl Plank floors
- 2x4 construction or 2x6
- Insulation types

In the end, the biggest factors for appraisals come down to the quality and condition of the subject property relative to its comparables. Even with all the detailed information laid out for a computer to analyze, it still has to choose comparables, which it can't do based on data points alone.

It would need to be able to identify quality and condition similarities, which is where Zillow and all AVMs have failed so far. This is something that, so far, only a trained appraiser can do.

0 0 Reply