Data & Analytics

Learn how SmartZip uses "big data" to make listing predictions in your area.

Oneeb Hassan avatar
Written by Oneeb Hassan
Updated over a week ago

In order to predict behavior like home buying and selling, you need a great deal of "big data".  SmartZip sources raw data from more than 25 trusted national sources, giving us deep insights on more than 95 million homes and homeowners nationwide. 

      

SmartZip Home Values:  How Are They Determined?

SmartZip Home Values are based on our patented AVM.  AVM  stands for "automated valuation model" - a service that provides real estate property values based on mathematical modeling.

Here are some of the variables we use, just to name a few:

  • Listing price trends

  • Time in home

  • Loan status

  • Number of residents

  • Owner savings

  • Home equity

  • Last selling date

  • Delinquency

  • LTV ratio

  • Price Appreciation

SmartZip's AVM

SmartZip estimates home values nationwide using our own automated valuation model (AVM). Our algorithms assess real estate data on various levels- from state, county and city to census tract and neighborhood areas. Each algorithm is slightly different depending on the circumstances, assessing variables related to publicly available property records: transaction history, comparable sales, listing history, and monthly rents; as well as area demographics: income, school ratings, crime, lifestyle, population growth and job growth and property attributes: beds, baths, building area, lot size, just to name a few. More importantly, it relies on our neighborhood-level house price index (NHPI).

SmartZip's NHPI

SmartZip’s neighborhood-level house price index (NHPI) measures how much the same homes have sold over time. The hyper-local NHPI currently assesses more than 70,000 neighborhoods across the U.S., including up to 100 million properties. By comparing how the same homes performed under different market conditions, we can more accurately predict a home’s current value in comparison to its last sales price.

Crude example of the NHPI: A home on a cul de sac in Richmond sold for $300,000 in 2008 but prices of similar homes on that cul de sac have (on average), risen 20% since then. The NHPI would show that the home may currently be worth $360,000 — 20% more than its 2008 sales price. (Note: The NHPI is only one factor taken into account when calculating an AVM value.)

Calculating AVMs in Non-Disclosure Areas

Most real estate transaction information in the U.S. is public record. However, some counties and states allow home buyers to keep the sale value of their home transaction confidential. In these non-disclosure counties and states, we may still receive mortgage data and transaction prices from buyers who choose to make it public record. We use this as the basis for each home sale price estimate.

Next — and this is something no one else does — when non-disclosure areas are being analyzed, our NHPI and AVM can adapt to “borrow” data from properties that match the geographical area and/or property features of the home being evaluated. We also compare against local listing prices and local sold prices, (that have been reported), in order to finalize our AVM values.

Example: The HPI is a repeat sales index. So, what happens if we only have one sale? Borrow from other properties with similar qualities, in similar areas.

SmartZip’s AVM Performance

We follow a very rigorous process to ensure that our AVMs are of the highest quality possible. One important thing to remember is that a home is only as valuable as what people are willing to pay for it. For that reason, we measure the accuracy of our AVMs by benchmarking them against actual sales prices, not listing prices. From there, we give each AVM a confidence score.

The majority of markets have a score above 80. The markets with low AVM confidence are generally those without much historical data, such as nondisclosure counties or states.

Data Updates

Some of our products, like our AVM estimates and Pre-mover (rankings) scores, are updated on a monthly basis while most property information is updated as soon as it becomes available.

How often our data is updated depends on many factors: what data, where it’s coming from, and what we need to do with it. In general, our raw data comes to us daily, but there are various delays between an actual event in time and when that data makes it to us, and then processed through our algorithms.

As an example, when a home is sold from one person to another, (or perhaps a mortgage is refinanced), a deed is recorded with the County Recorder's Office. A recorded deed, (normally a set of physical papers), could take up to 3 months to reach us digitally depending on the county in which it occurred. However, this is happening every day across the country, so we are getting hundreds of thousands of records daily!

Similarly, tax assessments are performed annually, (in some counties biannually); and so, although we get daily updates, this data doesn't change very often and relies on each county’s completion of their tax assessment.

On the other hand, listing events happen in real time so they are updated regularly. We work hard to push this data through our system and to our clients as fast as we can while being mindful of the data quality.
​       

Data & Analytics FAQ

Q:  What Makes Someone Have a High Likelihood to Sell?

A:  It comes down to the targeted homeowner's probability to list compared to everyone else's. We are analyzing "What could happen" in a farm. Smartzip combines current and historical property data, consumer behavior data, and market data to identify particular patterns in specific geographic areas.

We then apply our algorithms to capture significant relationships between these data sets. We are assessing the probability of homeowners needing particular professional services. It's important to remember that no algorithm can "predict" the future with 100% certainty, but we can test the efficacy over time.

As an example, a common application of predictive analytics that most are familiar with is to produce a "credit score". These scores are used by financial service providers to determine the probability of customers making future credit payments on time. We do a similar thing with SmartTargeting, except we're assessing the probability of homeowners needing particular professional services.

Q:  If a house goes on the market, what happens to the ranking?

A:  As soon as we receive a listing record for a property, their rank will drop significantly and will likely be updated to a 'Low' likelihood to sell. Normally this process is instantaneous; however, in some rare cases, there may be up to a 60 day delay time depending upon the data available in your territory, as we are updating information on over 92 million homes across the country.

Did this answer your question?