Data import brings a spreadsheet into the app. Each kind of record has its own importer; this walkthrough uses Clients.

The shape of the job is: download the blank template, fill it in offline, upload it, review what it is about to do, then commit.

Uploading imports nothing. It validates the file and shows you what would happen — created, updated, archived, and the rows it cannot write with the reason. Rows can be fixed in the review screen rather than back in the spreadsheet, and Re-validate re-runs the check over your edits.

Commit is the step that writes. If the import would delete anything, or change more than a hundred rows, it asks you to type the word COMMIT first — a deliberate speed bump on the two cases that are hard to undo. Smaller, additive imports do not ask.

Start with a few rows of test data. An import creates and updates real records, and the blast radius of a wrong column is the whole file.

Step 1
Open the Clients importer
Admin Settings > Data Import. Each thing you can import has its own importer — pick Clients.
Step 2
Start from the template
Download the blank template so the headings are exactly the ones the importer expects. A renamed or reordered column is the most common reason a file will not validate.
Step 3
Fill it in offline, then upload
Upload the completed file back into the same importer.
Step 4
Read the review, not the file
Uploading only checks. Nothing has been imported — what you are looking at is what WOULD happen, row by row.
Step 5
Fix what is flagged
A row that cannot go in is flagged with its reason, and you can correct it here instead of editing the spreadsheet and starting again.
Step 6
Re-validate
Runs the checks again over your edits, so you can see the flags clear before committing anything.
Step 7
Commit
This is the step that writes. Read the counts before you press it.
Step 8
The typed confirmation
If the import would delete anything, or change more than a hundred rows, it asks you to type COMMIT first. A small additive import will not ask — that is not a sign something is wrong.
Step 9
Read the result
It reports what was created, updated, archived or deleted, and how many rows could not be written.
Step 10
Check a record
Open one of the new clients and compare it against the row you sent. Do this every time — a column that landed in the wrong field validates perfectly well.
Step 11
What validation cannot catch
It checks the shape of the file: required fields, formats, references it can resolve. It cannot tell you that the right data is in the wrong column, that a date is the wrong year, or that you have imported last month's export. The review screen shows you what each row will DO — read a few of them rather than skimming for red.
Step 12
Importing twice
An import matches rows against existing records, so re-uploading the same file updates rather than duplicates. That is what makes correcting and re-running safe — and also what makes a file with a wrong identifier column able to update the wrong people. The review screen is where that shows up, as updates on records you did not expect.