Platinum Data Solutions has completed an initial round of accuracy testing on the 18 automated valuation models that it resells to lenders and servicers.
The accuracy testing provides a third-party analysis of an AVM's strengths and weaknesses and is designed to complement OptiVal, Platinum's cascading tool that ranks AVMs according to user preferences and requirements. Using the results of the AVM testing, lenders and servicers can rank the AVMs they use based on which models fit their needs at a price point lower than the $50,000 to $100,000 fee that third-party AVM testers charge.
The initial testing included running the AVMs against a pool of more than 27,000 properties across the United States. Platinum Data says its testing program is a part of its ongoing AVM accuracy efforts, and advises that AVM accuracy testing should be conducted on a quarterly basis.
There are two parts to an AVM: its algorithms and datasets. Variables in those two components differentiate various AVMs and make them better suited for specific types of valuations—like residential versus commercial properties, new versus existing construction, and rural versus urban locations, among others. The accuracy of AVM results can also vary depending on the intended use, like home equity lending, quality control and portfolio analysis reviews.
Those differences can cause an AVM's valuation to vary by 20% to 30%, the company estimates, adding that valuations that are too high or too low can lead to missed sales opportunities, lost liquidity and a heightened risk of compliance violations.
"Companies need to be aware that just because AVMs have a low price point, it doesn't mean they don't have an impact on the balance sheet. They do, and it's significant," says Phil Huff, CEO of Platinum Data, in a press release.