How to Clean a Messy CSV the Same Way Every Time (No Code)
If you receive the same messy CSV or Excel export every week or month — inconsistent dates, stray spaces, duplicate rows, blank cells, the occasional invalid value — you've probably cleaned it by hand more times than you'd like to admit. The problem isn't the cleaning. It's that you do it again from scratch every single time, and a tiny slip means this month's report doesn't match last month's.
The fix is reproducible cleaning: define the steps once, then replay the exact same steps on each new file. Here's how to think about it — and how to do it in one click.
Why “by hand” quietly fails
- It drifts. Manual edits are never identical run to run, so your outputs subtly diverge.
- It's unauditable. Nobody can see what was changed or why.
- It hides bad data. Rows that should be questioned get quietly fixed or deleted.
- It doesn't scale. Ten files become a lost afternoon.
The reproducible approach: a cleaning “recipe”
A recipe is just an ordered list of cleaning rules — step 1, step 2, step 3 — applied in sequence. Same input + same recipe = the same output, every time. A good recipe for a typical messy export looks like this:
- Trim spaces & fix case on text columns (e.g.
" john SMITH "→"John Smith"). - Standardize dates — read every format (
3/4/26,2026-03-04,4 Mar 2026) and rewrite them one way. - Remove duplicates by a key column like Email, keeping the first occurrence.
- Fill missing values in required columns.
- Flag invalid rows (e.g. emails without an “@”) to a separate list instead of silently dropping them.
- Convert types so numbers sort as numbers and dates as dates.
Do it in one click with Kramata
Kramata is a free desktop app built exactly for this. You drop in a file, add cleaning steps with a live before/after preview, then save the whole thing as a recipe. When next month's file arrives, you load it, click once, and get an identical clean output — plus a separate “Flagged” list of anything that didn't pass your checks.
- No code, no formulas. Every rule is a plain-language step.
- Reproducible. Same file + same recipe = byte-identical result.
- Batch mode. Run the recipe over a whole folder of files at once.
- Fully local. Your files never leave your computer.
Frequently asked
Is this like Power Query? Similar goal — repeatable transforms — but built for non-technical users: no formula language, plain-language steps, and a flagged-rows view so bad data is surfaced, not hidden.
Will it handle large files? Yes. Kramata uses an in-process database, so big CSVs load and scroll smoothly.
Does my data get uploaded? No. All processing is local.