Private by architecture, not by policy
The free tools never route your data through a server: there's no upload step to trust, and no privacy policy to take on faith. Parsing, profiling and transforming all run in your browser's own memory.
Client-side data profiling & harmonisation
Clean, profile and harmonise your data without it ever leaving your computer.
Columnwise runs entirely in your browser: drop in a CSV or TSV and get instant delimiter detection, type inference and statistical profiling. No upload, no account, no server wait.
Runs entirely in your browser for every free tool on this page, including the full pipeline workspace.
A quick taste, right on this page: no signup, no install
This reader previews the client-side approach. The full pipeline workspace (joins, aggregations, an expression builder, a re-runnable spec) is the actual product, one click away. Everything below runs on your device via the browser's File API; nothing is sent anywhere. Here's how to verify that yourself.
Drag & drop a .csv or .tsv file here
or
✓ Processed in your browser, nothing uploaded
Skeptical? Open your browser's DevTools → Network tab before choosing a file. If you've accepted analytics below, you'll see one small Google Analytics ping for this page view, and nothing else. Choose a file and you won't see a single request carrying it: delimiter detection, type inference and the whole demo run in memory, in this tab.
- Detected delimiter
- –
- Rows
- –
- Columns
- –
- File size
- –
Detected columns: confirm or correct the type
Columnwise guesses a type per column. Pick the right one if it guessed wrong; the quarantine count below updates immediately, using the same logic that drives the downloadable rejects file and transformation spec.
| Column | Type |
|---|
Transform summary
A preview of what a downloadable, re-runnable transformation spec records: the delimiter used, the type applied per column, and every row that doesn't conform, quarantined rather than silently dropped.
- Delimiter used
- –
- Rows kept
- –
- Rows quarantined
- –
Why rows were quarantined
| Row # | Column | Value | Expected type |
|---|
Preview (first 20 rows)
This reader is a taster. The full pipeline workspace adds joins, aggregations, an expression builder and a downloadable, re-runnable transformation spec, still entirely in your browser.
Open the workspaceWhy Columnwise
Reproducible
Every cleaning and transformation step is captured as a downloadable, re-runnable spec, not a one-off manual edit you can't retrace. Hand it to a colleague, re-run it next month, or audit exactly what changed.
Researcher-grade
Distribution-aware statistical profiling, quarantine-don't-drop handling of non-conforming rows, and cardinality-checked joins that warn you before a merge silently duplicates rows. Built for people who need to trust their numbers.
Everything Columnwise does with your data
Most of this is live today in the pipeline workspace; the quick reader demo above only shows the first two.
Delimiter, encoding and header auto-detection
Columnwise inspects a sample of the file to detect the delimiter (comma, tab, semicolon, pipe), text encoding and whether the first row is a header: the same logic powering the demo above.
XLSX/XLS reading, with a sheet picker
Open Excel workbooks directly: pick a sheet if there's more than one, and blank leading rows before the real header are detected automatically. Feeds the exact same pipeline as CSV.
Column data-type inference, with quarantine for non-conforming rows
Each column is classified as numeric, date, boolean or string from its values. Rows that don't conform to the inferred type are quarantined to a rejects file instead of being silently dropped, so nothing disappears without a trace.
Cardinality-aware joins with pre-merge duplication warnings
Assert one-to-one, one-to-many or many-to-one before merging two files. Columnwise checks real key uniqueness against that assertion and warns you, with the exact extra-row count, before a mismatched cardinality quietly inflates your data.
Merge two CSV files → Find duplicate rows → Remove duplicate rows →
Derived fields and expression builder
Build derived columns with a safe, whitelisted expression syntax: no arbitrary code execution (no eval, ever), so formulas stay auditable and sandboxed.
Aggregations with group-by: sum, mean, median, count, count distinct, 95% CI
Standard aggregation functions with group-by support, including a confidence interval on the mean, for summarising a table without leaving the browser.
Downloadable, re-runnable transformation spec
Every cleaning step is recorded as a portable XML spec you can download, version, share and re-run, so your pipeline stays reproducible and auditable, not a memory of what you clicked.
How Columnwise compares
| Approach | Where your data goes | Effort | Reproducible? |
|---|---|---|---|
| Excel / spreadsheets | Stays local | Manual, error-prone at scale | Rarely (edits aren't tracked) |
| Heavyweight ETL platforms | Depends on deployment | Expensive, complex to set up | Yes, but with a steep learning curve |
| Cloud AI data cleaners | Uploaded to a third-party server | Low effort | Varies by vendor |
| Columnwise | Runs entirely in your browser (free tools) | Low effort, intelligent defaults | Yes (downloadable, re-runnable spec) |
Pricing
Building and running your own pipeline, on your own files, is always free, with no file-size limit. A license is only needed to re-run a pipeline spec someone else exported to you, or one you saved for later.
Free
£0
Every tool and the full pipeline workspace, no signup. Build and run your own pipeline on any file.
Open the workspace7-day trial
£5
Re-run an imported or saved pipeline spec for 7 days.
Start the trialOne year
£39
Re-run imported and saved pipeline specs for a full year.
Get the one-year licenseThe trial and the one-year license are separate purchases; the trial price isn't credited toward the annual price.
Try it on your own file, right now
No signup, no upload. Open the full workspace, or try the in-page reader first.