MISMO 2.6 XML vs. PDFs: Which Is Best For UMDP?

After UMDP takes effect, the GSEs will accept appraisal data in two formats; either MISMO 2.6 XML, or a first-generation PDF. A first-generation PDF is a PDF file printed from the appraiser’s desktop with no other modifications made to the PDF. If you only deliver a first-generation PDF, it will be converted to XML before it’s analyzed. This conversion process is error prone and can lead to delays or errors in the appraisal data. Because of this there will be fees associated with the PDF scraping choice. The bottom line is that you should choose to submit the MISMO 2.6 XML to avoid these fees and delays. 

Here are a few key resources:

• UMDP Readiness Checklist  This free .XLS tool will help you ascertain where your data is coming from and what you'll need to do to get ready for UMDP.
• Video:   UMDP:  What kind of appraisal data do the GSEs want?
• Article in HousingWire:  The Slow Death Of The Appraisal PDF
• Article in DSNews:  "LPS' Appraisal Unit Turns To Mercury Network for UMDP Compliance"  -- This story explains why industry giant LPS chose Mercury Network and Native XML for UMDP compliance. 
• Article in HousingWire:  "CoreLogic turns to Mercury Network" -- This story covers why CoreLogic chose Mercury Network and Native XML, streamlining compliance for their clients.

To get the appraisal data in Native MISMO 2.6 XML format, you or your appraisal management company needs to work with appraisers that have this capability built into their software. If you get the XML directly from the appraiser, you don’t have to worry about errors with the conversion of the PDF to XML because the PDF isn’t being converted. The XML is created from the source file, thus making it Native XML.

If your appraisal vendor is going to receive PDFs from appraisers, scrape the PDFs themselves, and then submit the non-native XML to the GSEs, you may be in for some serious issues. You will run into errors that can’t be corrected by the appraiser because it’s an error with the converter and not with the way the appraiser filled out the report. It’s not a question of “if” this method will fail. It’s how often and how much it will cost you. A typical URAR has over 1,000 fields. Even at 98% accuracy, 20 or more fields will be corrupted. Some of those will be critical, you probably won’t know exactly what went wrong (the converter or the appraiser), and your pipeline will stop due to bad data.