How to Split a Large CSV Into Smaller Files

Somebody needs the data in pieces — because Excel will not open it whole, because each region should get their own, or because an import has a size cap.

The short answer

Open Data Converter, add the file, and choose split. Four rules are available — by row count, by output size, by sheet, or by a column's distinct values — and the header is repeated in every output part, which is the thing that makes the results usable.

The mistake almost everyone makes first

The obvious approach is to cut the file every n lines, with PowerShell or a quick script.

That gives you part1.csv with a header and part2.csv through part12.csv without one. Each of those opens with its first data row treated as column names — so you silently lose a row per file, and every column is labelled with a value from it.

Nothing errors. The files look fine. This is the single reason hand-rolled splits go wrong, and it is why "headers repeated in every part" is worth checking for in any tool you use.

Four ways to split, and when each is right

By row count. The one to use when the constraint is Excel — 1,000,000 rows per part keeps every file under the ceiling with room to spare.

By file size. For upload caps and email limits. Be aware the last part will be smaller, and that rows are never cut in half to hit the number exactly.

By a column's distinct values. Usually the most useful. One file per region, per client, per month, written in a single pass. The output is files somebody can actually be handed, rather than arbitrary chunks nobody can reason about.

By sheet. For a workbook — one output file per worksheet, which is a different job from splitting rows and comes up just as often.

Filter first, split second

Worth asking before you split at all: does the recipient need all the columns?

Most "this file is too big" problems are as much about width as length. Opening the file in Data Explorer, dropping the twenty columns nobody uses, and exporting the result frequently removes the need to split entirely.

The naming will matter later

Split output lands with dated, self-describing names rather than (1), (2), (3). That reads like a small detail and stops being one the moment you split the same export twice in a month and have to work out which folder is which.

Splitting by column value names the parts after the values, which is better still — region=West tells you what is in the file without opening it.

Putting it back together

The reverse is the same tool. Point Data Converter at the folder and choose combine: columns are unioned by name rather than failing if they disagree, and you can add a source_file column so each row records which part it came from.

That round trip — split, work on the parts, recombine — is much safer when the recombine step can tell you where every row originated.

---

Every other free way to do this asks you to upload the file first, which for a file you are splitting because it is too large is a particularly awkward requirement.