Vector tiles are a method of serving map data in which features are cut into a tiled grid and delivered as compact packets of geometry plus attributes, rather than as pre-rendered images. The client (browser or app) draws and styles the features locally, enabling smooth zooming, interaction, and on-the-fly restyling.
Why it matters
Unlike raster tiles, which are fixed pictures baked at a chosen style and zoom, vector tiles keep the underlying geometry and properties. That means you can change colors, toggle layers, query features, label dynamically, and rotate or tilt the map without re-fetching anything from the server. For web GIS with large geological or terrain datasets, this delivers a responsive, queryable experience at a fraction of the bandwidth of full vector downloads.
A concrete example
The dominant format is the Mapbox Vector Tile (MVT) specification, a Protocol Buffers encoding usually served at URLs like /{z}/{x}/{y}.pbf and packaged in an MBTiles or PMTiles container. Tools such as tippecanoe generate tilesets from GeoJSON, and renderers like MapLibre GL or deck.gl draw them client-side. Coordinates within a tile are stored in local tile space (commonly a 4096-unit extent), not geographic units.
Common pitfall
Vector tiles are a visualization and delivery format, not an analysis format. Geometry is simplified and clipped per tile and zoom level, so a polygon may be split across tile boundaries and its detail reduced at low zoom. Never compute areas, lengths, or topology from vector tiles — go back to the source data in a proper analysis CRS for measurement.