The Evolution of AVMs


What's the point?Automated Valuation Models (AVMs) are a handy tool for quickly and simply providing an estimated value for a property. Nobody has quite nailed the science, as all AVMs are inherently wrong. A new trend is to put several different AVMs side by side. Thus producing varying estimates of the same property. The value for consumers is in providing a starting point, a somewhat novel basis of what their home—or a home they're interested in buying—is worth. The value for agents and brokers is in those consumers reaching out with questions; giving the broker an immediate opportunity to demonstrate their expertise by providing a more accurate and locally informed estimate. While providing multiple AVMs gives consumers more information, it also can cause confusion. This provides additional likelihood that a consumer will reach out to a local real estate professional.

AVMs have come a long way in the past few years. For contrast, take a look at the AVM landscape two years ago when I first wrote about it here.

How could AVMs be improved? Currently most AVMs work by running comps – comparing the subject property address to nearby recently sold properties and current on-market listings. This process could be improved by incorporating more specific local data. A property's estimated value could be increased by proximity to well regarded school zones, and highly desirable local perks like express subway stops. Likewise the estimated value could be decreased for being too close to airports/flyover zones and busy intersections/street noise. Currently there aren't any AVMs that utilize this kind of rich local content.

Another issue with AVM predictions is that they don't have any way of automatically factoring in the quality of a property into the pricing model. A renovated kitchen will increase the value of a property, while a dilapidated roof is likely to decrease it. AVMs have no way of recognizing those alterations. Therefore, models assume the property is in average condition. Some companies are working towards a solution for this. I have heard ideas of allowing a homeowner or an agent to change certain aspects based on intimate knowledge of a property. Zillow is already allowing consumers to personalize their own Zestimates, which they can then run by an agent to see how accurate it looks.

How are AVMs already improving? Weiss Analytics offers an AVM that provides not only the current estimated value of a property, but also forecasts the estimated value of a property a year into the future. At Onboard we’re working on changing the scope of our valuations too. We are creating a view that shows the value of a property over a period of time, going back a series of months.

Another use for AVMs has come into play as a result of a changing market trend. A younger generation has shown a propensity towards renting instead of owning property. Addressing this, Zillow has created a product—a Zestimate—that estimates a property's monthly rental price. Lookout for this to become more popular in 2016.


For an example of Onboard's AVM check out Address Report.


If you have any questions or would like to know more, give me a shout.