Unfortunately, it is not possible to (accurately) calculate the AVM market value for all homes in the Netherlands. With every AVM prediction we provide a confidence percentage to determine accuracy.


First and foremost, machine learning will be applied to make almost all of the predictions. When a value cannot be predicted, indexing is being applied. In the extraordinary case the required amount of transaction data is insufficient for indexing, a reference model will be applied to create sufficient transaction data. This three-stage rocket, looks like this:
  1. 1.
    Machine learning prediction
  2. 2.
    Indexing: using the Kadaster index when there aren’t enough transactions within the same neighbourhood
  3. 3.
    Reference model in case no transaction data is available.

The end of the AVM API chapter.

Go back on how to implement the AVM API: