Help · Importing data

Import Excel, CSV and JSON

Excel, CSV, JSON, a photograph of a printed table, or a live connection to a database. Importing has no row cap on any plan — that part is free, always.

Files: Excel and CSV

Upload file on the home screen creates a new table from the file. Inside an existing table, Import → Upload file appends to it instead.

Either way you get a preview before anything is written, and columns are matched by name — so next month's export lands in the right columns even if someone moved one. A column in the file with no matching column in the table is skipped rather than guessed at.

A few things worth knowing before you import a real workbook:

• The first row is treated as headers. A merged title row above the real header is the classic trap: a workbook whose row 1 is a merged “Q3 report” imports with columns named Q3 report, Untitled, Untitled(1), and your actual header row lands in the data as a row. Delete title rows in Excel before importing — faster than fixing it afterward.
• Merged cells don't survive as merged cells. The value goes to the first column of the range and the rest come through empty. Tables here are databases, not canvases.
• Formulas in the file come across as values. If you want the calculation to keep happening, rebuild it as a formula column — it'll then apply to every row you add later, which is usually what you wanted.
• One worksheet per import. If the workbook has several, you pick which one; run the import again to bring in another. They don't all arrive at once.

JSON

Import → Import from json takes an array of objects and turns the keys into columns. Nested objects are flattened one level; deeply nested arrays are better handled by picking the array you actually want first. Handy for an API response you already have on disk.

A photo or screenshot of a table

Import → Import from image reads a table out of a picture — a scanned invoice, a screenshot of a report someone sent you, a photo of a printed page. This one runs through the AI model you configured, which means it needs your own API key first (setup) and the image is sent to that provider. If that isn't acceptable for the document in front of you, type it in instead.

Check the result before you build on it. OCR is good, not infallible, and the numbers it is least sure about are the ones that look most plausible.

Databases and APIs

Data Connections in the top right connects to a database or an HTTP endpoint and keeps a table in sync with it, on a schedule. Reading a connection works on the free plan; creating and editing connections is a paid feature, as is writing back out. See pricing for the split.

Paste from the clipboard

Copy a range in Excel or a browser table and paste into a TableDI table. Small transfers don't deserve a file.

If an import looks wrong

• A whole column came in blank — the column name in the file doesn't match the column name in the table. Rename one to match the other and import again.
• Numbers came in as text — the source has currency symbols or thousands separators inside the cell. Import as-is, then add a formula column to clean it; the raw column stays untouched so you can always check the original.
• Dates came in as text — the same story. Keep the raw column, derive the clean one. That way the next import is already handled.
• Nothing happened at all — check the file isn't open and locked by Excel.

Still stuck? Email us — one inbox, read by the people who build TableDI.