Data Provenance Log Tracker — Free Data Lineage Log Builder

This tool builds a provenance log for a dataset: a dated list of where each version came from, what changed, and who changed it. Provenance is the answer to the questions every dataset eventually faces: where did this number come from, who cleaned it, what was removed, and can the result be reproduced? You add entries as work happens — source received, duplicates removed, values imputed, file exported — and export the log as JSON to keep with the data. A dataset without a log is a claim without evidence; a dataset with one can be audited, reproduced, and trusted.

How to use it

Record the date, source, action, and actor for each step.
Click Add log entry after each processing step.
Review the log before sharing the dataset.
Click Export JSON and save the log beside the data file.
Keep logging when the data changes again.

Worked example

A survey dataset passes through four hands: Ana receives the export on March 15, Bo removes 23 duplicates on March 18, Ana imputes missing income values on March 20, and the team exports the analysis file on March 25. The log records all four steps with dates and actors. When a reviewer later asks why the analysis file has 977 rows while the raw export had 1,000, the answer is in the log — 23 duplicates removed on March 18 by Bo — instead of a panicked reconstruction from memory. That single answer is worth the minute each entry took.

Honest limits

The log lives in the page until exported: closing the page loses unexported entries, so export after every session and store the JSON with the data, not separately. The tool records what you tell it; it cannot verify that an entry is true or complete, so a log is only as honest as its writers. It also tracks dataset-level steps, not cell-level lineage — for cell-level history, use version control or a database with audit tables. Treat the log as a lab notebook: contemporaneous, specific, and never rewritten after the fact to flatter the result.

Who it is for

This tool is for research teams sharing datasets, students learning reproducible workflows, and anyone who has ever stared at a file named final_v7_REAL.xlsx and wondered what it contains. It is the cheapest reproducibility habit there is: one line per step.

Going further

Adopt a naming rule alongside the log: raw files stay read-only with their original names, processed files gain a version and date, and nothing is ever called “final” until it is archived. The log and the filenames then reinforce each other — the log says what changed, the name says which version you mean. When a collaborator asks for “the data,” send the versioned file plus its log, not a bare attachment: the recipient can then see exactly how far the file has travelled from its source and what assumptions it already carries. Datasets, like results, are only as credible as their paper trail. Link each log entry to the file it describes by recording a checksum or a versioned filename alongside the entry, so the log cannot drift away from the data it documents. A log that says “duplicates removed” without saying from which file version is only half a record. When datasets are shared, send the log in the same package and mention it in the cover note; a recipient who never opens the log gains nothing, while one who does gains the ability to verify every step. Provenance only works when it travels with the data. Review the log at project milestones, not only at the end: a mid-project review catches missing entries while memories are fresh and steps can still be reconstructed, while an end-of-project review finds gaps that nobody can fill. Make the review a standing agenda item — five minutes, log open, each person confirming their entries — and the log stays a living record instead of a reconstruction. Teams that review logs regularly also catch processing errors earlier, because describing a step aloud often reveals that it was done wrong.

Embed this tool on your website — free

Copy the code below to add this working tool to your own page. A small credit link stays with it — that’s all we ask.