AI and Scoring Models in Real Estate: What They Can and Can't Do

Scoring tools make research faster and more consistent. Here is what they are good at, where they fall short, and how ours is measured.

Scoring tools can make real estate research faster and more consistent, but it is worth being clear about what they can and cannot do.

What automation does well: it applies the same measurements to every area, so a neighborhood in Cleveland can be compared with one in Phoenix on equal terms, and it can check thousands of places at once for the handful worth a closer look.

What it cannot do: see things that are not in the data. Planning documents, council minutes, a walk around the block at night and a conversation with a local property manager will often tell you more about one specific area than any dashboard.

How nort.estate works: each area's score is built from Zillow's ZIP-level home value and rent indexes (the 12-month price trend, the last three months' momentum, and gross rental yield) using a method published on our methodology page. The 12-month projection is a damped trend, backtested on 198,895 past predictions across 25,751 ZIP codes. Its median error was 2.93%, against 3.86% for simply assuming prices stay where they are, and its 80% range contained the actual outcome 79.5% of the time.

For investors: use scores to shortlist, not to decide. The most useful thing a model can do is point you to the five places worth researching by hand.