Sales Data

Kennedy, Ecker and Isakson measure the accuracy of AVMs and their FSDs, demonstrating that most of the time FSDs underreport forecast error.

Using appraisals seems like a great way to get more benchmarks for testing AVMs, and since AVMs and appriasals are trying to do the same thing: predict sales prices, it seems intuitively appealing. However, there are some fundamental problems with using appriased values to test AVMs. Appraised values are subjective and inconsistent, and they introduce bias and compound errors. This article explains why arms' length transactions are the only appropriate way to test AVMs.

This op-ed offers an affirmation of the core tenets of what is becoming the industry-standard testing framework: a data-driven testing methodology grounded in sound and prudent validation principles. While Veros challenges this approach, the broader AVM ecosystem—including regulators, lenders, and nearly all major AVM providers—have embraced a process that prioritizes objective, real-world performance measurements over now antiquated methods allowing for data leakage into the Automated Valuation Model.

Many lenders use the "15% rule" for AVM-based quality control, assuming that when an AVM estimate aligns with an appraisal value, both are accurate. This widespread practice reveals a fundamental misunderstanding that could be undermining the entire valuation ecosystem. The problem? Comparing AVMs to appraisals measures precision, not accuracy. This distinction isn't just semantic—it's the difference between knowing whether your valuations actually predict market value or simply agree with other estimates. When institutions use AVMs to validate appraisals in QC processes, then turn around and use those same appraisals to test AVM accuracy, they create a dangerous circular logic that masks true performance. This article examines why the only meaningful benchmark for AVM accuracy is the actual sale price—the endpoint of market negotiation. We explore the critical importance of blind testing, MLS data suppression, and alignment with real-world use cases. Most importantly, we demonstrate how the Interagency Appraisal and Evaluation Guidance explicitly supports testing AVMs against actual sales data, not appraisal estimates. Understanding this distinction is essential for lenders, AMCs, and AVM vendors who need to move beyond the comfort of circular validation to achieve genuine predictive accuracy.