AFR's Evaluation Tools
Exterior Inspections, AVMs and Desk Top Appraisals
In keeping with the AFR theme "Your Satisfaction is Our Future" lenders can now order an AVM, an Exterior Inspection or desk top appraisal (ValueNet) from our website. Our program streamlines the collateral valuation process and saves our lenders and consumers hundreds of dollars on every transaction.
Automated Valuation Models (AVMs) and Exterior Inspections
Since its inception over 15 years ago, AVM technology has advanced the world of automated valuations from novelties to necessities. Today, AVMs are efficient, effective, and essential tools for staying competitive. AVMs expedite processes, lower costs, and minimize risk. AVMs are also a great tool for supporting a traditional purchase appraisal and measuring risk in your portfolio.
Exterior inspections are also available from AFR. With national coverage and turnaround time available from 1 to 6 days (service fee based on requested turn time) these reports include information as to the property and neighborhood condition along with photos of the exterior of the property and street view. Using these property reports, in addition to a market value evaluation, is a conservative yet cost effective approach to meeting the new compliance requirements under the Interagency Appraisal and Evaluation Guidelines.
Key Benefits, Features and uses of AFR's AVMs include:
√ Reduce closing costs
√ Shorten the time needed to close loans
√ AFR's AVMs provide three value points - A Low, Medium and High Value
√ AFR's AVM's provide a confidence score to indicate accuracy
√ Assist in Quality Control by:
● Using AVMs to calculate current LTV in declining real estate markets to measure the need to lock or reduce a HELOC (batch ordering available at discounted pricing, minimum order of 1,000 loans).
● Using AVMs to support a standard appraisal on purchase money mortgages (unlike appraisals, AVMs give you unbiased valuations based strictly on data).
The AVM providers AFR has incorporated into its AVM ordering program are the leading models in the AVM industry. Each has its own particular strengths that, combined in our program, give lenders the quality, coverage and service they have come to expect from AFR.
Freddie Mac's Home Value Explorer (HVE)
HVE has the best national coverage of any AVM and is the most precise AVM available. When you use HVE you benefit from Freddie Mac's unique proprietary algorithm that blends multiple model estimates returned by the Repeat Sales model and the Hedonic model. Freddie Mac has effectively employed AVMs internally for its own risk and portfolio management for over 15 years.
When choosing an AVM, it's important to consider coverage, vendor reputation, current technology capability, and price. But the most crucial factor in choosing the best AVM is precision. Relatively small differences in precision will quickly translate into lost customers, lost revenue, dramatically higher default costs and less liquid assets.
Greater precision = lower default cost on the loans you originate.
Greater precision = more loans processed without unnecessary upgrades or lost deals.
Recent Neighborhood Sales Defined
HVE Reports provide a list of recent neighborhood sales within a 1-mile radius of subject property sold within the last year. Up to 10 properties can be listed; however, the number will vary based on sales activity.
Confidence Scores Defined
In addition to the estimated value of a property, HVE reports provide a Confidence Score based on a FSD (Forecast Standard Deviation). Confidence Scores indicate the accuracy of the estimated value when the target price is not known in advance. These Confidence Scores show the accuracy of the model result, relative to the accuracy of an appraisal's forecast.
- A high (H) Confidence Score corresponds to a FSD in use of <= 13%.
- A medium (M) Confidence Score corresponds to a FSD in use of > 13% to <= 20%.
- A low (L) Confidence Score corresponds to a FSD in use of > 20%.
Over 70% of HVE estimates are high confidence estimates. Approximately 25% of HVE estimates are medium confidence estimates, and less than 5% of HVE estimates are low.
A FSD of 10% indicates that approximately 68% (i.e., one standard deviation) of the model results will be within 10% of the true market value. Using FSD is the most accurate way of measuring an AVM model result because it is statistically sound and empirically derived. It precisely quantifies the accuracy of a value estimate. HVE users will be able to fine-tune their risk management calculations and loan processing based on HVE's validated forecast standard deviations.
If a standard deviation is exceedingly high, indicating the data appears to be inconsistent or the neighborhood property values vary drastically, HVE will not return a value.
Of all the AVM scoring systems, the HVE Confidence Score is the most reliable and easiest to use because it is tied to the accuracy of the actual model in use, and to the accuracy of the traditional appraisal process.
Casa Values
CASA, based on the home price research concepts pioneered by world-renowned economists Karl Case and Robert Shiller, leverages a very large, meticulously filtered property records database and multiple market-specific analytic approaches to estimate current market values for individual residential properties. Eight of the nation's top ten banks rely on CASA to deliver fast and statistically accurate home valuations.
CASA measures home specific price dynamics by combining several valuation methods, whenever possible, including characteristics analysis (often referred to as "hedonics"), Case-Shiller (repeat sales) Indexing, non-linear regression techniques, home price forecasting and unpublished innovations. To compensate for varying degrees of data quality by geographic area, the optimal combination of data elements and estimation methodologies is determined and weighted for each property analyzed.
Unlike most AVMs, the CASA system calculates property value estimates in real time. This processing technique, which still renders CASA value estimates in seconds, serves to further enhance the accuracy, hit rate and safety of the service for our lender customers. While the CASA algorithms mobilize data inputs and techniques used by an appraiser (e.g., comparable sales, home characteristics and time adjustments), CASA does not emulate the traditional appraisal process. Rather, CASA is based upon dynamic analyses and multiple property valuation approaches, and is calibrated over time on a market-specific basis. CASA is engineered to optimize hit rate without compromising accuracy and collateral score veracity so that customers can benefit from CASA value estimates that are not merely fast, but reliable.
Characteristics Analysis is a regression technique that assigns values to attributes of individual houses. The CASA system calculates market specific regression results on statistically valid samples throughout the U.S. These statistics are re-calibrated on an ongoing basis as market conditions change and CASA's database grows.
Comparable Property Approach determines the property's value from estimated current values of neighboring homes. Unlike the traditional appraisal process, CASA is able to draw upon a very robust set of historical sales indexed to estimate their current values, rather than rely on merely a few that may have sold within the past few months. Indexing is used to time-adjust historical home values to their present market levels. For each historical value found for the subject home and surrounding homes, CASA applies the most appropriate metrics from the REdex LibraryTM for the given geographic area, time interval, and price level. Wherever possible, a segmented ZIP Code level index is employed for this portion of the CASA analysis.
Non-linear regression techniques comprise a set of advanced statistical processes. It is through these techniques, for example, that CASA can make assessments of local market and home characteristic quirks which may impact the determination of market value for the subject property.
Home Price Forecasting enables CASA to mitigate the lag inherent in data collection and filtration processes. CASA is the only AVM that addresses the inefficiencies of public record data compilation with credible forecasting technology.
Each of these methods has certain limitations, given market conditions and available data. For example, in markets where robust characteristics data sets are not available, a simple characteristics analysis model would not work well. The comparable sales approach tends to be less accurate in diverse areas with fewer similar comparable sales. Indexing is not optimal for homes that deviate from the typical price paths of their market segments. CASA uses and combines all of these valuation methods whenever possible to create a cross-weave that compensates for the limitations of any one method for a particular home. The exact means to combine the approaches used for any given property is determined on a market-specific basis and can change to reflect results of ongoing CASA testing and calibration. Examples of the kind of factors that can affect the valuation techniques deployed and how they are weighted by CASA include:
- Homogeneity of the housing stock where the property is located
- Uniformity of the property with respect to its neighbors
- Amount and nature of raw data available for the property and other homes within its market
CASA Collateral Scores are measures of the expected accuracy of each CASA value. Unlike "confidence scores" of many other automated property valuation systems, CASA Collateral Scores are statistically reliable for automated loan scoring and underwriting processes. CASA Collateral Score choices include Safety Scores and Confidence Grades.
CASA Safety Scores are for those customers who prefer an ordinal collateral scoring metric linked to a fixed 10% valuation accuracy threshold. This score is a statistically accurate measure of the probability that CASA's estimated value for a subject property does not exceed +10% of the "true" value of the home. For example, a Safety Score of 85 means that there is an 85% probability that the CASA value estimate delivered does not exceed the actual value by more than 10%.
In addition to augmenting CASA's database with multiple and sometimes redundant property data sources, CASA has developed proprietary raw data filtering techniques to apply to incoming data sets irrespective of the source. These techniques are most important (particularly in light of uneven quality within certain public record data sets) in ensuring that incoming data records are checked for reasonableness and accuracy.
For example, in addition to identifying data entry errors wherever possible, CASA will focus on identification of non-arms-length sale transactions and anomalous, incongruent data through a series of automated filtering mechanisms. Within markets where CASA utilizes more than a single data supplier, we have the luxury of checking a set of records as reported to us. Given this capability, CASA is able to review the overall quality of data provided by individual suppliers through time.
Accuracy
Accuracy is determined by comparing CASA's value estimates for a set of properties with actual recent sale prices for those properties. Tests are routinely performed on sample properties by county. Test sample size varies according to the size of the test market, but samples are generally comprised of thousands of recently sold properties to ensure statistical validity.
After many years of collecting home price and property specific characteristics data for the purpose of producing the Case-Shiller Indexes, and later, the CASA system, the result is perhaps the cleanest and most comprehensive residential property records database in the country.
VeroVALUE
Veros is shaping the future of mortgage technology through its continuous predictive technology leadership by providing innovative and highly accurate solutions in collateral assessment. Standard AVM reports include value estimate, value range, confidence score, available subject property information, market data information and historical price trends for the subject's neighborhood.
VeroVALUE is built with the latest in predictive technology that evolved from years of research and development in the aerospace, technology and manufacturing industries. This technology utilized advanced science to more accurately predict results. The same technology used in VeroVALUE has been used on mission-critical prediction problems such as orbital spacecraft reentry and aircraft engine design and manufacturing for organizations such as Boeing, the U.S. Military, Pratt & Whitney, GE and NASA.
Reliability is a key component of VeroVALUE through the use of Veros' Confidence Score (VCS). The VCS is highly correlated to our predictive accuracy and, as such, is extremely useful in decision making within a lending organization.
Innovation is the driving force behind the technology. Combining highly technical analytics with real estate fundamentals requires unique and creative thinking. Veros' design team has developed numerous proprietary methodologies which result in a more accurate and reliable service for lenders.
Combining multiple valuation methods, each using different predictive technologies (e.g. Bayesian, neural network, advanced statistical/mathematics, probabilistic, data mining) in one AVM increases our ability to render values and increase the accuracy of our valuation. Veros runs approximately 10 different valuation methods for any given subject property compared to other vendors that may run only 2-4 basic methods.
Veros currently delivers property valuations in 40 states and over 600 counties nationwide. In addition to VeroVALUE's meaningful confidence scores and the high usable hit rates, VeroVALUE also generates property fitness analytics, fraud detection, price trend information, secured valuations, and VeroFORECAST, an analytic tool that estimates future values and market stability.
ValueNet Desk Top Appraisals with Exterior Inspection
ValueNet appraisals are completed at the local level, by local appraisers, utilizing local data. ValueNet appraisers have deep familiarity and experience in your market. As a result, ValueNet offers a more cost effective appraisal solution which is typically completed in 24 hours.
Using a sophisticated process that parallels traditional appraisals, ValueNet appraisals include extensive research of the subject property, comparable selection from local data sources including MLS and the actual valuation of the property. The ValueNet appraisal process and the report form is USPAP (Uniform Standards of Professional Appraisal Practice) compliant so lenders can trust that the proper process was followed and the resulting valuation is accurate.
Our flagship product is a USPAP compliant desktop valuation process that utilizes quantitative and qualitative appraisal methodologies. ValueNet can be used when AVMs cannot produce a value or the risk factors involved in the loan require an appraiser be involved in the process.
ValueNet includes:
- Market Value of the subject property
- Neighborhood characteristics and property comments
- Satellite view of the subject property
- Housing trends
- Property characteristics (lot size, rooms, baths, age, sq.ft., etc)
- At least 3 sales comparables with map
- Appraiser Comments
- Appraiser's electronic signature
- Next business day turnaround time in most areas
ValueNet Plus
ValueNet Plus is our ValueNet product plus:
- an exterior inspection of the home and neighborhood
- Multiple photos of the home and neighborhood
- Turnaround in 4 - 6 business days in most areas
For additional information on AFR's services, please contact us at 1-800-995-8667 or visit us at www.afrservices.com
