using AURA QC to complement CU

Appraisal QC with AURA can complement CU, and reduce buybacks

Nov 2, 2023

This series of articles is designed to help sellers & correspondents understand how Fannie Mae Collateral Underwriter (CU) can be paired with AURA appraisal QC reports to address buyback risks before delivery, and how to be prepared if they receive appraisal repurchase requests.

Increases in repurchase activity throughout 2023 have highlighted for lenders the need to fully address appraisal-related CU findings and flags before delivery – the delays and additional costs associated with revisions, secondary products and even re-ordering an appraisal are far less than the financial consequences of a repurchase request.

As we’ll address in the series, if a lender is dissatisfied with an appraisal report and the collateral risk associated with the loan they can seek a series of escalated remedies. CU helps users determine which risks exist and the appropriate remedies. Here are the  first  and second articles in the series.

AURA appraisal QC summary

The prior article provided best practices for using CU to mitigate repurchase risk. This article addresses how to utilize the new Appraisal Underwriting Risk Analysis (AURA) collateral risk and appraisal QC report to complement the use of CU in situations where the lender determines there is an elevated collateral risk. AURA is at the forefront of AI-driven appraisal QC, and a big step forward in collateral risk management. AURA automatically reviews an appraisal and returns a report in a matter of seconds; but it does a lot in those few seconds, including:

  • Submits to UCPD (CU/LCA), if requested, to understand CU and LCA scores, and associated findings; these are not presented in the AURA report
  • Checks to ensure the appraiser has a valid license (and no disciplinary actions pending) for the assignment
  • Checks public records, MLS, and other local market insights for sales and listing comparables, and selects the most accurate recent data
  • Runs a Home Price Index (HPI) model against the last sale date to see if the current valuation is in line with what is expected based on the neighborhood price appreciation
  • Uses image recognition to check all the photos in the appraisal, to determine room type, condition, and the appropriate labels
  • Uses natural language processing (NLP) to check for problematic language and risky terms in the appraisal that are unacceptable and could indicate bias
  • Runs rules to check and highlight inconsistencies between the photos, adjustment grid, and sketch or floor plan
  • Runs an AI condition model to compute condition from photos and property data and compares this to the appraisers condition (C-rating) in the appraisal

Additionally, AURA provides three scores for a credit or collateral underwriter to rely on:

  • Valuation accuracy score – a numeric score that rates the accuracy of the appraisal when compared to the AVM. This is based on a comparison of the appraised value with ClearAVM and the low/high range of the AVM, giving a further indication of the overvaluation or undervaluation risk. Importantly, AURA does not include the actual ClearAVM value so lenders are not required to disclose a new valuation to the borrower. ClearAVM is consistently the highest-ranked lending-grade AVM by Fitch Ratings when assessing the use of valuations in private-label residential mortgage-backed securities (RBMS). Note that the true test of AVM accuracy is on refinance transactions where the purchase contract value is not supplied.
  • Valuation confidence score – this is the overall confidence of the ClearAVM value based on the available data to compute the automated value, most notably the subject property comparable data from MLS, public records, and other sources.
  • Report quality score – this is an assessment of the quality of the overall appraisal report – based on the number of rules that fail – giving the underwriter an indication of the soundness of the report and the attentiveness of the appraiser.

Finally, AURA presents the top 10 ranked sales and listings comparables to the subject property, showing key property characteristics and the source of data. This includes showing where the appraisal-supplied comparables rank within, or outside, the top 10 comparables. Again, a quick way for credit underwriters to validate if there may be issues with the selected (and adjusted) comparables.

Recap on repurchase requests

To recap, the main reasons for appraisal repurchase requests and look at how AURA can assist:

  • Failure to use comparable property sales that are locationally and physically similar to the subject, including:
    • Inappropriate selection due to GLA, location, site character & age, and/or
    • Inadequate adjustments to comparable sales, and ultimately, and/or
    • Failure of the adjusted comparable to support the appraised value
  • Misrepresentation of the physical characteristics of the subject property, improvements & comparables.
  • Failure to comment on negative factors in the neighborhood, subject property, or proximity of adverse influences.

You could summarize this as… make sure the appropriate comparable properties are selected, adjusted correctly, and support the value conclusion. Then check the property characteristics and condition for accuracy, and look for any external obsolescence.

How AURA appraisal QC can assist repurchase requests

After a human review utilizing CU, we recommend that the lender provides selected details, or the full AURA report, to the appraiser as supporting information to the request for clarification or revision. In the next article we’ll address using the CU score to help determine the appropriate level of risk-based review, and the next steps based on the response from the appraiser, but for now let’s focus on how you can use AURA appraisal QC to assist with the QC process.

  • Overview – with risk score & flags, data discrepancies and reconciliations
    • When ‘Overvaluation or Undervaluation Messages’ appear, lenders should check AURA Collateral Scores:
    • Appraised Value – compare the appraised value with the ClearAVM low/high-value range to see where the appraisal valuation, including checking scores:
    • Valuation Accuracy – indicates the accuracy of the appraisal when compared to the lending-grade ClearAVM; a score of 8 and above is more accurate
    • Valuation Confidence – This indicates the confidence of the ClearAVM computed value, based on supporting information; again, 8 and above is a higher confidence
    • Report quality – this is an overall check on the number of issues found in the report – the more rules that fail, the lower the report quality score – giving the underwriter an indication of the overall quality of the appraisal. As above, over 8 is considered higher quality.
    • Valuation Findings-Failed – check to see what AURA report rules have failed that indicates overvaluation or undervaluation
  • Comparables & Adjustments
    • Valuation Findings – Failed – check to see what rules have failed that indicate concerns with the comparable properties
    • Most notably, these will highlight discrepancies between comparables for condition and quality, and the count of beds, baths, and GLA
    • Also, check if the Comparable Price Range rule has failed
  • Nearby Similar Properties – Sales – review the rank and see where the appraisal (report_supplied) comparables are positioned
    • If they are lower in the top 10 rankings or outside the top 10, then this should be reviewed more carefully
    • This page also includes an Average Sales, Median Sales, and Low/High Sales to provide guidance on bracketing the comparables
    • The comparables list notes the property price, sale date, distance from the subject, GLA, beds, baths, lot size, and data source
  • Nearby Similar Properties – Listings – when sales comparable are sparse or not available, this provides background information
    • This also provides information on markets with increasing or decreasing market value adjustments
  • Sales History
    • Valuation Findings – Failed – check to see what rules have failed that indicate concerns with the sales and transfer history
  • Market Trend
    • Valuation Findings-Failed – check to see what rules have failed that indicate overvaluation or undervaluation, including the message that the forward-trended HPI from the last sale date has a material variance with the current appraised value


To wrap up this article:

  • Fannie Mae prohibits lenders from sharing CU findings and the CU web tool with appraisers and AMCs – this ensures an entirely independent appraiser opinion of value and is a pillar of effective risk management
  • Lenders can use CU to evaluate the appraisal, complete a human review, and determine the next steps
  • If underwriters decide that due to heightened collateral risk they need to contact the appraiser for clarification and/or revisions, they can utilize the AURA QC report to check their findings and share some or all of the AURA appraisal QC report with the appraiser and/or AMC who completed the assignment
  • This allows lenders to be in compliance with GSE directives, and provide supporting information to the appraiser to resolve the concerns identified

Coming next… the remaining two articles will address:

  • What to do next if the appraiser is unable or unwilling to address a lender’s concerns
  • How to effectively document the ‘well-informed human risk analysis’ and action taken before loan delivery

If you need to find out more sooner please…. Contact us