Randatum is an offline-first desktop random data studio. Every computation runs locally on your machine β no network, no data upload. It can generate random numbers and colors in one click, mass-produce structured test data, run research randomization by explicit methods, and keep a reproducible record of every run.
1. Install & Launch
Randatum runs on macOS and Windows desktop.
- macOS: open the package, drag Randatum into your Applications folder, then launch from Launchpad.
- Windows: run the installer (.exe), follow the wizard, then launch from the Start menu.
The default window size is 1280Γ800, freely resizable (minimum 900Γ600). No sign-in and no network are required to use all free features.
2. Interface & Shortcuts
A fixed navigation rail sits on the left with six workspaces. On narrow windows the rail collapses to icons (hover to see names).
| Workspace | Purpose | Shortcut |
|---|---|---|
| π² Quick Random | Quickly generate random numbers or colors | β1 / Ctrl+1 |
| π Data Studio | Generate multi-field structured test data | β2 / Ctrl+2 |
| π¬ Research | Grouping, sampling, repeated simulation | β3 / Ctrl+3 |
| π Templates | Save and reuse generation schemes | β4 / Ctrl+4 |
| π Run History | View past task metadata | β5 / Ctrl+5 |
| βοΈ Settings | Theme, edition, and Pro features | β6 / Ctrl+6 |
3. Quick Random (Numbers / Colors)
Open "Quick Random"; the top toggle switches between Numbers and Colors.
3.1 Random Numbers
Fill in the rules in the left "Parameters" panel and click Generate (or press βEnter / Ctrl+Enter). Results appear as chips on the right.
| Parameter | Description |
|---|---|
| Min / Max | Integer range, inclusive of both bounds |
| Count | How many numbers to generate |
| Step | Value interval, e.g. step 2 yields β¦, 2, 4, 6, β¦ |
| Seed (optional) | The same seed reproduces the same batch; leave empty for a fresh draw each time |
| Exclusions | Remove specific numbers, e.g. 3,7,10-20 |
| Unique | When checked, no duplicates within one batch |
| Parity filter | All / odd only / even only |
| Sort | None / ascending / descending |
After generating, the Copy button at the top-right of the results copies them all at once (comma-separated).
3.2 Random Colors
Set the color count and an optional seed, then click Generate to get a palette of HEX swatches across the full RGB space. Click any swatch to copy its HEX value (e.g. #3A7BD5). The same seed reproduces the same palette.
4. Data Studio
Generate structured test data with many fields, then preview it, view column statistics, export to several formats, or save it as a reusable template.
4.1 Basic Flow
- Click Add Field, name each field, and pick its type;
- Click a field to edit its parameters (range, mean, probability, etc.);
- Optionally enable Data Pollution (missing / empty string / Gaussian noise) to mimic real-world dirty data;
- Click Preview for sample rows; click Stats for a per-column data-quality summary;
- Choose an export format and Export at the bottom, or Save to Templates.
Fields support rename, duplicate, drag-to-reorder, and delete.
4.2 Supported Field Types (17 total)
Basic types (Free): Integer, Float, Boolean, String, UUID, DateTime (ISO-8601), Color (HEX), IPv4, MAC, Uniform distribution.
Advanced distributions & weighting (Pro): Weighted Enum, Normal, Log-normal, Exponential, Triangular, Poisson, Binomial.
4.3 Data Pollution
On top of clean data, inject by ratio: missing values, empty strings, or Gaussian noise (add normal noise to numeric columns). Use this to test how your program tolerates dirty data.
4.4 Export Formats
- CSV, TSV, JSON (free)
- JSONL (Pro)
- Copy CSV to clipboard
Large volumes use streaming export, which shows the number of rows written and can be canceled at any time.
5. Research Randomization
A dedicated entry for study design: import a subject list, perform grouping or sampling by an explicit method, and keep reproducible records plus a methods draft.
β οΈ Input order is part of the reproducibility contract: reordering subjects may change the result.
5.1 Import Subject Data (CSV)
Click Import CSV and choose a file. The first column must be id; remaining columns can serve as stratification variables. Rows with empty IDs, duplicate IDs, or inconsistent column counts are rejected with a prompt.
5.2 Three Methods
- Group Assignment: Simple random assignment (free) β enter group names and allocation ratio. Permuted blocks & stratification (Pro).
- Population Sampling: Simple random sampling / SRS (free) β set sample size and with/without replacement. Proportional stratified sampling (Pro).
- Repeated Simulation (Pro): Draw an independent sample for each repetition. Set repetitions, sample size per run, and normal parameters. Stable sub-streams keep runs mutually independent while the whole run remains reproducible.
After running you get the assignment table / sampling result, a result summary, and a methods draft (Chinese + English) awaiting researcher review.
5.3 Export the Reproducibility Package
Click Export Reproducibility Package. Randatum creates a randatum_research_<timestamp> folder containing:
result.tsv: grouping / sampling result datarandatum.meta.json: metadata (engine, version, seed, time, etc.)summary.json: result statistics summarymethods.zh.md/methods.en.md: Chinese & English methods draftssha256.txt: SHA-256 checksum manifest of each file
6. Templates & Run History
Templates: after configuring fields in Data Studio, choose "Save to Templates". In the Templates workspace you can rename, duplicate, and delete templates. Templates can be exported to template.json files or imported from files.
Run History: after a data export or research randomization, the task's metadata is recorded automatically β including task type, size (rows / count), seed, and run time β for later reference and traceability.
7. About Reproducibility
Randatum's core promise is "same input β same output".
- The underlying engine uses Philox4x32-10, a counter-based random source with a stable streamId and counter partitioning.
- With a fixed seed, fields that rely on integer arithmetic only β integers, unbiased bounded integers, booleans, list picks, UUIDs β are reproduced bit-for-bit.
- Statistical distributions that use math functions (normal, exponential, etc.) are reproduced under a versioned algorithm with a published tolerance.
8. Free vs. Pro
Randatum uses a one-time purchase Β· lifetime Β· offline model β no subscription, no cloud dependency.
Free
- All Quick Random capabilities
- 10 basic field types + uniform distribution
- CSV / TSV / JSON export
- Simple random assignment & SRS
- Fixed seeds & metadata
Pro (one-time purchase)
- 7 advanced distributions + weighted enum
- JSONL export
- Permuted blocks, stratified assignment & sampling
- Repeated simulation
- Full reproducibility package export
Click the unlock button in "Settings β Randatum Pro" to purchase. After switching devices, use "Restore Purchase" to recover your entitlement.
9. Settings
In the "Settings" workspace you can:
- Theme: follow system / light / dark
- Randatum Pro: view edition status, unlock Pro, restore purchase
- About: view version info
10. FAQ
Q: Is my data uploaded?
No. Randatum is offline-first; all generation, import, statistics, and export happen locally.
Q: Why do two runs give different results?
Because no seed was set. Enter the same seed to reproduce the same batch.
Q: Why did my CSV import fail?
Make sure the first column is named id, with no empty IDs, duplicate IDs, or rows with inconsistent column counts.
Q: I reordered subjects and the result changed?
That's expected. Input order is part of the reproducibility contract, so reordering changes the random assignment.
Q: What is sha256.txt in the package for?
To verify that each file wasn't tampered with or corrupted in transit. It attests to provenance and integrity, not to randomness or fairness.
Q: Will Pro carry over to a new computer?
Click "Restore Purchase" in Settings to recover your entitlement.
11. Important Notices
- Randatum is for study design, pilot trials, teaching, simulation, and testing only, and does not replace regulated clinical randomization systems such as IRT / RTSM / GxP.
- Methods output is always a draft and must be reviewed by the researcher before use.
- Statistical checks are for supporting judgment and must not be presented as a "certification of randomness".
- The SHA-256 manifest verifies only provenance and integrity, and is not a proof of a fair draw.
12. Feedback
Found a bug, have a question, or want to suggest a feature? Email us at selfren@163.com.
When reporting an issue, including the seed, field/grouping configuration, and the metadata from your reproducibility package helps us track it down faster.