How to Clean CSV Tables in the Browser

ToolComb’s CSV Editor runs in the browser. You paste or open a file, edit cells, add or remove rows and columns, and download. No account. The cleanup does not require uploading your export to finish the edit. After the table is stable, CSV to JSON is the usual next hop - or Chart Maker if you only need a PNG (CSV to chart workflow). The roundup of related tools: Best free CSV, chart, and file inspector tools.

Step 1: Open the first twenty lines with suspicion

Before any tool, look at the raw text if you can. A title like Q3 Sales Export above the real headers is a classic. A blank line between header and data breaks naive parsers. A sep=; hint means Excel wanted semicolons even if the extension is .csv.

In CSV Editor, scroll the top. Delete promotional rows. Ensure row one is only column names - sku, qty, price - not merged report titles.

Step 2: Fix the delimiter before you fix the data

Wrong delimiter looks like “one giant column.” European exports use ;. Warehouse dumps use tabs. US Excel often uses commas but wraps quoted fields that contain commas.

If every row appears as a single cell, change how you import or split - match the separator the file actually uses. Cleaning values inside cells does not help when the split never happened. Full failure catalog: Why CSV to JSON fails.

Step 3: Strip footers and subtotal rows

Exports love a final row: Total,,,48291. Charts treat that as another category. JSON arrays get an object with empty keys and a string total.

Delete summary rows from the data section. Keep totals in Excel or compute them downstream. If you need totals in JSON, compute in code after conversion, not as a fake data row.

Step 4: Normalize headers for machines

Headers become JSON keys. Spaces are fine; leading/trailing spaces are not. Duplicate header names (Date, Date) collide in object form. Rename to date_start and date_end.

Avoid punctuation you will regret in code (%, #, /). Lowercase snake_case is unfashionable in spreadsheets and friendly in APIs.

Step 5: Handle encoding and weird characters

Smart quotes, em dashes, and currency symbols survive in UTF-8 CSV. Mojibake (é) means the file was read with the wrong encoding. Re-export UTF-8 from the source if possible.

The BOM problem hides in the first header cell. If lookups miss on sku but work on a copy-pasted key, inspect \uFEFF.

Step 6: Sample before you ship the whole file

Copy ten data rows plus headers into CSV to JSON. Inspect one object:

• Are numbers still strings? Often correct for IDs and money.
• Are dates still strings? Usually yes until you parse with locale.
• Are empty cells "" or missing keys? Know which your converter emits.

If the sample is clean, run the full file. If not, return to CSV Editor - do not stack fixes in three different cloud tools.

What people break

They convert first and edit JSON by hand for two hundred rows. CSV was the right edit surface.

They delete columns in Excel without updating formulas, then export broken #REF! cells into CSV.

They “clean” by sorting in Excel and saving, accidentally sorting header with data or splitting blocks that must stay aligned.

They trust auto-detect delimiter on a file that mixes tabs and commas in quoted notes.

What a browser CSV editor will not do

It will not join to a second database, run pivot tables, or validate against a JSON Schema. Multi-megabyte sheets may slow or fail in the tab - that is memory, not a prompt to upload PII elsewhere.

ToolComb processes the table locally for the edit. It does not replace backup policy, access control, or redaction before you paste regulated data.

When cleaning is enough

If the goal is a slide chart, clean through Step 5 and open Chart Maker. If the goal is an API payload, clean then CSV to JSON. If the goal is understanding why the last conversion exploded, read Why CSV to JSON fails after you fix the obvious header row - sometimes the file was never CSV at all.

Keep the original export on disk. Download the cleaned CSV with a new name. Ship the cleaned copy. The thirty-second diff between old and new headers prevents an entire bad deploy.