How Comparable Listings Are Picked — and Why the Comp Radius Changes Your Number

What actually goes into a comp set

A short-term-rental revenue projection isn't pulled from a single number on a market report — it's built from a set of comparable listings, filtered and averaged. The core filters are the same ones an appraiser would recognize: bedroom and bathroom count, property type (single-family vs. condo vs. townhome), and a location radius around the subject property. Each comp then contributes its own trailing-12-month occupancy and average daily rate, and those get blended into a projected annual revenue figure.

The part that's easy to miss is that every one of those filters is a judgment call, not a fixed rule. Tighten the bed/bath match and you get fewer, more relevant comps. Loosen it and you get more data points but noisier ones. The radius does the same thing in geographic terms — a wider radius pulls in more listings, which sounds like a good thing, but it also starts including comps in adjacent neighborhoods with different demand drivers, different price points, or different regulatory status.

Why the radius is the lever that moves your number most

Picture a subject property a few blocks from a lake versus a walkable downtown core. A tight radius keeps the comp set inside that same lake-adjacent pocket, where demand and pricing behave similarly. Widen the radius by even a mile and the comp set can pull in properties closer to downtown, or on the opposite, less-desirable side of a highway — properties that rent for meaningfully different money for reasons that have nothing to do with bed/bath count.

This is also why two lenders — or two revenue tools — can hand you two different projections for the exact same address. They're not disagreeing about the property; they're using different radius defaults, different comp-count minimums, or different weighting between recent months and the full trailing year. Neither is necessarily wrong. They're answering slightly different questions.

There's a related lever that gets less attention than radius but works the same way: the minimum comp count a methodology requires before it'll return a projection at all. Some approaches will generate a number off a small handful of nearby listings; others hold out for a larger sample before treating the projection as reliable. A methodology with a low comp-count floor will happily produce a confident-looking number in a genuinely thin market — which is exactly where that confidence is least warranted.

Why trailing-12-month data is the standard, not just a convention

Comp data is almost always pulled on a trailing-12-month basis rather than a snapshot of last month or a forward guess. That window exists specifically to absorb seasonality — a beach market's August and a ski market's February look nothing alike, and a single month would badly misrepresent the annual picture. Trailing-12 smooths the peaks and troughs into one number that's meant to represent a realistic full year.

It also matters which 12 months are in the window. A comp set pulled today versus one pulled six months from now will overlap heavily but not completely — older months roll off, newer ones roll on. In a market with a real, structural trend (new competitive supply coming online, a steady rise or fall in demand), that rolling window means the projection itself drifts over time even if nothing about your specific property changes. That's a feature, not a bug — it's the methodology staying current — but it's worth knowing that the number you get today isn't frozen.

• Bed/bath and property-type match narrows the comp set to genuinely similar listings.
• Search radius trades precision (tight) for sample size (wide) — neither is free.
• Trailing-12-month occupancy and ADR per comp absorbs seasonal swings into an annual figure.
• Comp count matters: a handful of comps in a thin market carries more noise than a large comp set in a dense one.
• The trailing window itself rolls forward over time, so a projection pulled today can drift from one pulled later even with no change to the subject property.

What to actually do with this

You don't need to become a data scientist to use this. The practical takeaway is to treat any single revenue projection as a range, not a point estimate, and to ask what radius and comp count produced it before you anchor a purchase decision to it. If a market is thin — few short-term rentals nearby, or a lot of variance between them — expect the projection to carry a wider margin of error, and underwrite conservatively rather than to the headline number.

The feasibility check runs this comp logic against your actual subject property and shows you the resulting cap-adjusted, income-method figure that would actually be used to qualify a DSCR loan — which is the number that matters, not the market-report average.