New construction (ever added a unit / new or reclassified commercial parcel, FY18–24)
Likely renovated (structural signal, a flagged large swing, or cumulative growth well above the town's typical no-change rate)
Everything else — height/color = $ growth per sqft of land, FY17→24
How "likely renovated" is determined: most of these are the
identified_renovation / unclassified_large
parcels from the classifier, plus a newer check — parcels whose
total FY17→24 growth exceeds the town's own empirical no-structural-
change baseline (47.6% compounded, not the commonly-assumed Prop 2.5
2.5%/year figure, which caps the tax levy, not assessed values)
by more than 30 points. That second check is a bottom-up estimate,
not a certainty — it catches gradual value-adding activity the
per-year structural rules miss, but "grew faster than typical" isn't
proof of renovation on its own. See
METHODOLOGY.md ยง6 for full detail and
known limits.