When a homeowner asks me what their home is worth — or a buyer asks whether a listing is priced right — I know what they're really asking: "What does the online estimate say?" Those automated valuations are a tempting starting point, but they're just a starting point. A comparable sales analysis is the actual answer.

Online estimates run on algorithms fed by public data. They're genuinely useful at the extremes — sorting markets, ballparking a neighborhood, giving you a rough range before you've seen a single comparable. But they can't see inside the house. They don't know that the kitchen was renovated last year, that the roof is fifteen years old, that the floor plan is awkward, or that the house two doors down is a comp killer because it sits on a much busier corner.

Comparable sales are actual market decisions — what real buyers paid for real, similar homes, in the recent past. A good comp set answers the questions the algorithm can't: what's similar in size, condition, age, and features, sold in the last three to six months, in the immediate area? When I build one, I'm making judgment calls an algorithm can't — adjusting for the finished basement, throwing out the foreclosure that sold as-is, weighting the sale across the street over the one a mile away.

That doesn't mean comps are perfect. Thin markets with few sales, rapidly shifting markets, or unusual properties all make the picture harder to read. That's exactly when a professional's judgment matters most — and when the online estimate, which smooths over all that nuance, is least reliable.

Here's how I use both: the online estimate is the weather forecast. Comps are looking out the window.

When you're sizing up a price — buying, selling, or investing — do you start with the estimate, the comps, or something else entirely?