Automate with ETL

Transform Stages

Shape your rows, join two inputs together, and run geometry operations between reading and writing.

Updated 2026-08-315 minutes read
Before you begin
  • Familiarity with the flow canvas

Between reading your data and writing it out sits the work — dropping columns, filtering rows, joining two sources, buffering geometries. The stage list on the left of the canvas groups all of it into four categories. Search it by name if you know what you want; browse it if you don't.

ETL Stages

Row and column work on a single input: "Select Fields", "Filter Rows", "Sort", "Field Manager", "Deduplicate", "Formula", "Group By" and more. These are the stages you reach for most.

"Filter Rows" is worth knowing well. Its editor shows an "Expression" footer that "Auto-updates as you edit", so you can see the condition you're building as you build it.

Joins

Everything in "Joins" takes two inputs and produces one — connect both before the stage will run. Join on matching values with "Join", or on location with "Spatial Join". "Union All" stacks two inputs into one set of rows.

Spatial / Grid

Geometry operations: "Buffer", "Centroid", "Convex Hull", "Dissolve", "Reproject", and spatial indexing for aggregation. Raster equivalents sit alongside them, with the imagery-specific stages on their own page — GeoAI and Raster Stages.

Custom Logic

When no built-in stage fits, write SQL. "Custom SQL Select" runs a SELECT over the rows coming in.

Next steps