Blog

Shapefile vs feature class: what actually changes when you switch

Dry, scrub-covered hillside landscape under a clear sky.

If you've been maintaining the same .shp files since field school, you already know the shapefile's quirks: the 10-character field name cap, the dBASE attribute table that chokes on long text, the fact that one "layer" is actually five or six files sitting in a folder together and god help you if one goes missing. A feature class fixes most of that, but the real question for a working GIS lead isn't "which format is newer." It's whether the switch buys you anything on the project in front of you.

The structural difference

A shapefile is a standalone file set: .shp for geometry, .dbf for attributes, .shx for the index, plus whatever sidecar files your software wants that week. Each shapefile stands alone. There's no shared schema, no enforced relationship between one shapefile and another, and no way to store null geometry (a shapefile with no features in it is just an empty husk of a file).

A feature class lives inside a geodatabase, either a file geodatabase (.gdb) or an enterprise one, and that container is what changes the behavior. Feature classes in the same geodatabase can share a coordinate system automatically, get grouped into feature datasets, participate in topology rules against each other (no slivers between your wetland boundary and your buffer polygon), and carry field names longer than ten characters without truncating your "disturbance_type" column into "disturban." You also get domains and subtypes, so a field crew can't type "wetlnad" into an attribute and corrupt your classification scheme.

None of this is exotic. It's the same vector data, same points-lines-polygons, just with a schema wrapped around it instead of loose files in a folder.

When a shapefile is still fine

Plenty of environmental consultancy work doesn't need any of that overhead. If you're handing a single delineation layer to a regulator who's going to open it in whatever free viewer they have on hand, a shapefile is the safer bet for compatibility. If it's a one-off boundary for a permit submittal and nobody downstream needs domains or topology, building a geodatabase for it is just extra setup with nothing to show for it. Shapefiles also travel well: smaller teams, simpler email attachments, fewer "what version of ArcGIS do you have" conversations with a client's junior staff.

When to reach for a feature class geodatabase instead

The switch earns its keep once you've got more than one related layer that needs to stay consistent with itself: a stream network and its buffer zones, a series of habitat polygons that should never overlap, or a multi-year monitoring dataset where you're tracking the same features across site visits and need stable IDs that survive edits. Topology rules catch the sliver gaps and overlaps that a reviewer would otherwise find by eye at 1:500 scale. Domains stop a tired field tech from inventing a sixth land cover category that doesn't match your legend. And if you're building anything with relationship classes, like linking sample points to a separate table of lab results, you need a geodatabase; a shapefile has no mechanism for that at all.

There's also the versioning and editing side: a file geodatabase handles concurrent edits and larger attribute tables (shapefiles cap out at 2 GB and get unstable well before that) in a way loose shapefiles were never built for. If your project has a GIS team of more than one person editing the same dataset, that alone is reason enough.

The practical answer

Neither format is "better" in the abstract. A shapefile is the right tool for a single, simple, portable deliverable. A feature class inside a geodatabase is the right tool the moment you've got related layers, attribute rules you need enforced rather than hoped for, or a dataset that's going to get edited by more than one person over more than one site visit. Most consultancies end up keeping both around: shapefiles for client handoffs and quick exports, geodatabases for the working project data underneath.

Where this gets harder is when the feature you need isn't a standard layer at all, say, a one-off class like "illegally dumped tire piles" or "culvert outfalls with visible staining," something no off-the-shelf classifier has ever heard of and no existing shapefile template covers. That's a different problem than format choice, and it's the one Few-Shot Mapping was built around: you label a handful of examples, it maps the rest of your site from high-res imagery.

If you've got a custom feature nobody's built a layer for yet, that's worth a look before you start digitizing it by hand.

Request early access

← Back to the blog