Data Units are consumed in three different ways:

1. Every time you run an algorithm we need to run a specific machine or several machines. Data Units are then machines running in x period of time. For every hour a machine runs (standard) we charge $1 usd. It will also depend on the type of machine you select if you select them manually and not in default mode.
2. Every time you run a SQL query you consume data units. The more complex or the less optimized your query is the more data units it will consume. To help you out in monitoring your Data Units we included in the SQL editor the amount of Data Units a query you will consume based on your code.
3. Every time you run an ML Operator we track BigQuery usage (fetching data from BQ) which translates in Data Units.
4. Everytime you run API Exporters, APIs, run an integration, etc you will consume Data Units as well in less or more proportion.

In every plan we include an x amount of data units. For example in the free plan we include 2 usd so that clients can get a feel of our product.