How we calculate your home estimate




  • Risk-adjusted ROI (Return on Investment) projection on an investment property
  • Yield and Total Return estimate calculated using Price Estimate, Rent Estimate, expenses/ fees and risk-adjusted expenses and Fees

INVestimate® can be preliminary reference in determining the investment potential of a property; and is not an actual investment performance of the property. INVLestimate is auto computed for about 110 million homes nationwide based on millions of data points focusing on features such as price, rent, location, market growth, employment and rentability.

INVestimate is derived from two key HomeUnion's
Data points:

HomeUnion Price REALestimate

HomeUnion Price REALestimate (a.k.a Price REALestimate) is a "site unseen" fair market value for a property in the United States regardless of its recent sale. Price REALestimate comprises a predicted property value and a confidence interval or range, with lower and upper values, that accounts for the spotty knowledge about the condition of the property.

HomeUnion uses machine learning techniques to predict Price at a specific point in time by regressing values of comparable properties onto a set of features (predictors) that include:

  • Physical facts about that home and land
  • Geographical features like properties? geocodes and distance to various points of interest
  • Market conditions such as unemployment, income, crime, and school ranks
  • The tax assessor's indication of value and other tax assessment information
  • Previous valuations and historical home price movements
  • Neighborhood quality
We have more than 15,000 geographically varying price estimate models predicting prices for 111 million properties nationally.

Measure of Accuracy

Price RealEstimate

One measure of accuracy is to compare the predicted Price REALestimate with actual sale price of a property. To assess how well the models perform, statistically, we compare all recent sales (within the last year) and determine the number of estimate predictions that are off by 5%, 10% from the actual sale.

For an overall performance of the model at State, County, City and ZIP levels, we measure Median Absolute Percentage Error [MAPE aka MdAPE] and to understand over prediction or under prediction in a location, we measure Median Percentage Error [MPE].