Your first import,
minus the mystery.
A tiny dataset. A visible transformation. One useful error. Run the fixed customer example locally and follow the source all the way to its validation result.
Meet the source
The example contains three synthetic customer records. Customer 001 has whitespace around the name; record 3 has no identifier. The workflow renames fields, trims whitespace, preserves identifiers as strings, and checks three deliberately small validation rules.
Three synthetic source records. One identifier is missing; one name has surrounding spaces.
[
{
"Customer ID": "001",
"Name": " Ada ",
"Country": "GB"
},
{
"Customer ID": "002",
"Name": "Milo",
"Country": "US"
},
{
"Customer ID": "",
"Name": "Sam",
"Country": "FR"
}
]Ready to inspect. All processing stays in this page.
What the mapping changes
Customer ID becomes external_id, Name becomes name, and Country becomes country. Surrounding whitespace is removed and country labels are uppercased. The value 001 remains a string; the mapping does not turn it into the number 1.
What validation checks
The identifier and name must be nonempty. Country must be GB, US, or FR for this specific exercise. Those three allowed values are a teaching choice, not a complete country standard or a rule for your application. Each result also carries a one-based source record number.
Read the result
The complete output contains evaluated: 3, two accepted records, and one rejected record. The error points to source record 3 and the missing external_id. No records are saved. A production write, authentication, full CSV parsing, and arbitrary file processing are outside this local example.
Turn it into your own test
Download the source, result, and example schema. In your own project, agree on identifier behavior, accepted values, and missing-field semantics before expanding the example. Add duplicate checks and destination authorization where your workflow requires them.
Continue with the example reference for exact shapes or the mapping and validation guide for the broader design decisions.