Why an address is not enough
Every valuation begins with the same raw material: what has recently sold nearby. The difference between a number that holds and a number that collapses in the first three weeks is what happens after that data is gathered.
This page describes the method. It does not produce a figure, because a figure that has not seen the property is not worth publishing.
Choosing comparables
A comparable earns its place by being similar in the ways that actually drive price for that particular property, not by being geographically close or recently sold. Proximity is a weak proxy. Two houses on the same street can belong to different markets if one is a renovated original and the other is new construction on a subdivided lot.
Fewer good comparables beat more weak ones. A valuation built on distant or dissimilar sales looks precise and is not accurate, and the appearance of rigour is exactly what makes that failure hard to catch.
Adjustment: what the data flattens
Recorded sale prices carry no information about condition, finish level, light, layout, or how well a renovation was executed. Two houses with identical square footage, bed and bath counts, and lot size can be separated by a wide margin on those variables alone, and at the upper end of the market they frequently are.
Adjustment is where judgment enters, and it is the part worth interrogating. Anyone presenting a valuation should be able to say which comparables were adjusted, in which direction, and why.
The number is the output. The reasoning is the product.
One method does not cross the county
An acre in Alpine and a high-floor condominium in Fort Lee are both Bergen County housing, and almost nothing that determines the value of one determines the value of the other. Acreage, setback, frontage and topography drive one; floor level, exposure, view line, monthly charges and building reserves drive the other.
A single valuation model applied across both produces a number for each and a reliable number for neither. The Northern Valley reads as one region and prices as many, which is why the method has to change with the housing stock rather than the county line.
What a valuation deliberately excludes
What the seller paid. What the seller needs. What the property is assessed at for tax purposes. What a neighbour is asking. Each of these is a real number and none of them is evidence of market value, though all four routinely find their way into pricing conversations.
How this differs from an automated estimate
An automated estimate is a statistical model built from recorded sales and public records. It is genuinely useful for direction and scale, and it is structurally blind to the variables that move the number most at the top of the market, because those variables are not in its inputs. It cannot see a bad floor plan, a compromised approach, or a renovation that was finished properly.
How this differs from an appraisal
An appraisal is a licensed opinion of value prepared to a defined standard, most often for a lender, and bound by the terms of that assignment. A pricing valuation asks a different question: not what a lender should rely on, but how to bring this property to market so that it meets its strongest audience at the right moment. The two can differ without either being wrong.
What to ask about any number you are given
Which sales were used, and why those. What was adjusted, in which direction, and on what reasoning. What the number assumes about condition and about timing. And what would have to be true for the number to be wrong. A valuation that cannot answer the last question has not been tested.
Related: what a sale actually costs a New Jersey seller at closing is set out in the closing-cost reference.