Most American parcel data is published city by city, in whatever schema the county assessor happens to use. Massachusetts is the exception. MassGIS Level 3 Standardized Assessors' Parcels puts all 2,558,412 parcels in the Commonwealth into one schema — every municipality, the same field names.
For anyone drawing site analysis in Boston, that is a gift. One query pattern gets you Back Bay, Cambridge, Somerville, Worcester and Springfield. You do not write an adapter per city, which is exactly what New York, Philadelphia, Los Angeles and Chicago each require.
We use this dataset in production for the Building age and Building use panels on sitedia. What follows is what we learned wiring it up.
| Field | What it is |
|---|---|
YEAR_BUILT |
Year of construction, as recorded by the municipal assessor. 2,198,910 parcels carry a plausible one — about 86% of the state. |
USE_CODE |
Massachusetts Department of Revenue property classification code. The hundreds digit is the category; the rest is detail. |
Massachusetts is old, and the data shows it: 15,566 parcels record a construction year before 1800, and 199,321 before 1900. In a half-mile circle on downtown Boston we see years running from the 1760s to the 2020s in a single drawing. That range is the reason a building-age map of Boston reads so differently from one of Los Angeles, where the histogram spikes hard in the 1900s–1920s.
The code is a three-digit number and the first digit tells you almost everything:
| Leading digit | Category |
|---|---|
| 0xx | Multiple use (mixed residential / commercial / industrial) |
| 1xx | Residential — 101 single family, 102 condominium, 104 two-family, 105 three-family, 111+ apartments |
| 2xx | Open space |
| 3xx | Commercial — 300s lodging, 320s–330s retail and auto, 340s office |
| 4xx | Industrial |
| 6xx / 7xx / 8xx | Chapter 61 forest, 61A agricultural, 61B recreational |
| 9xx | Exempt — government, education, religious, charitable, cemeteries |
The digit rule takes you a long way, but it is wrong often enough to matter. Some examples we had to special-case:
316 Storage and warehouse sits in the commercial 3xx block, but
reads as industrial in a drawing.322, 323, 324 — discount stores,
shopping centres, supermarkets — are a different animal from the small
storefronts that dominate 3xx.336 and 337 are parking garages and lots. Painting
them as commercial buildings misreads a block.380 golf courses and 384 marinas are commercial by
tax classification and open space by every other measure.130–132 are vacant residential land, not housing.942 colleges,
955 hospitals and 960 churches all end up in the
same hundred but belong in different colours.We built our mapping by pulling the actual DESCRIPT text for
every code that appears in the Boston metro — 297 distinct values — and reading
them, rather than assuming the hundreds rule held. It did not.
MassGIS publishes an ArcGIS feature service already in WGS84, so no coordinate transform is needed:
https://arcgisserver.digital.mass.gov/arcgisserver/rest/services/AGOL/L3_Parcels_FeatureService_4326/FeatureServer/1/queryAsk for a bounding box and just the fields you need:
?geometry=-71.075,42.348,-71.045,42.368
&geometryType=esriGeometryEnvelope
&inSR=4326
&spatialRel=esriSpatialRelIntersects
&outFields=YEAR_BUILT,USE_CODE
&returnGeometry=true
&outSR=4326
&orderByFields=OBJECTID
&f=geojsonThat comes back in about two seconds for a couple of thousand parcels. Note
layer 1 — layer 0 is Devens, a separate jurisdiction,
and will look empty for anywhere else.
Ask the service for the minimum and maximum YEAR_BUILT across the
state and you get 22 and 23023. Those are typing errors that have
been carried faithfully from municipal records into the standard schema. If you
map colour to year without a sanity range, one mistyped parcel stretches your
entire scale and every real building collapses into one shade. Accept only years
in a plausible window — we use 1700–2100 — and treat the rest as unknown.
You will get 102, but also 1020, 1021,
330V, 985I, 010J. Municipalities append
their own sub-codes and letter suffixes. Parse the first three characters
and ignore the rest, and match on the leading digit rather than on exact code
equality — otherwise a large share of parcels quietly fails to classify.
The record is the taxable lot. Attaching the year to a single building and stopping leaves garages, rear additions and everything else on a multi-building lot blank. Assign the lot's attributes to every building footprint whose centre falls inside the parcel polygon.
Related: if you simplify the parcel geometry to shrink the payload, keep the tolerance well under the width of a narrow urban lot. We measured a comparable grid where a 9 m tolerance dropped matching from 77% to 61%; at 2 m the result was identical to the unsimplified polygons at half the vertices.
Measured on our own database inside a half-mile circle on downtown Boston, 89.3% of buildings carry a year built and 99.9% carry a use. That is the second-best result of any US city we have wired up, behind New York — and it arrives with Cambridge, Somerville and the rest of the Commonwealth attached, which no other state-level source we have found does.
If you want the drawing rather than the dataset, sitedia runs this pipeline and returns building age, land use, footprints, heights, green space, transit, terrain and aerial as SVG, PNG, DXF or 3D. Coverage figures for every city are on the data sources page, and the same walkthrough for New York is at NYC PLUTO.