How to trace where a data field came from
Somebody questions a number in a report. Lineage is how you answer them with a trace instead of an opinion.
What you will do
You will follow a column backwards from the dataset in front of you, through each transform applied to it, to the source it originally came from.
This answers both directions of the same question: where did this come from, and what breaks if the source changes.
When this is useful
- A number in a report looks wrong and you need to find where it went wrong.
- Somebody asks what a field means and the answer depends on how it was produced.
- A source system is changing and you need to know what depends on it.
- An audit or handover requires you to show provenance rather than assert it.
Before you start
- A dataset with some history — Lineage is most useful on a dataset that has been through transforms. On a freshly loaded file it will show you a source and very little else.
Steps
Open Govern → Quality & Lineage → Lineage
Select Govern, then Quality & Lineage, then the Lineage tab. Select the dataset you are investigating.
A lineage graph for that dataset.
Read the graph
The graph runs from the original source, through each transform step that was applied, to the current staged state of the dataset.
Each step is one of the operations that was actually applied in Prepare — not a summary, but the sequence itself.
A chain from the source, through the transform steps, to the current dataset.
Follow a specific column back
Pick the column you are questioning and follow it backwards through the steps. Where it was cast, split, derived or renamed, you see that.
A column that was derived from two others shows you both — which is usually where the explanation for a surprising number lives.
Reach the source
Keep going back to the original source: the uploaded file, the database table, the API endpoint, the CRM object.
That is the answer to "where does this come from" — a specific origin rather than a department.
The original source at the start of the chain.
Use it forwards for impact analysis
The same graph read the other way tells you what depends on a source. Before changing or retiring one, this is what you check.
Add the business meaningPro
Lineage tells you where a column came from. The Dictionary tells you what it means. Together they are what a new person actually needs.
What happens next
Where lineage exposes a transform that should not have happened, Prepare is where the sequence gets corrected.
For a written record rather than a screen, the document generation screen produces architecture and mapping documents that capture provenance.
Example
A revenue figure is 8% higher than finance expects. Lineage on the reporting dataset shows a Fill Nulls step using the mean applied to the amount column three steps back. The transform is doing exactly what it was told; nobody had realised it was told to.
Tips
- Check lineage before defending a number. It converts an argument into a fact.
- Read it forwards when planning a source change — it is the cheapest impact analysis you will get.
- Lineage plus a dictionary entry is what makes a dataset comprehensible to someone who did not build it.
- A surprising number almost always has a transform behind it. Look at the steps before doubting the source.
Limitations
- Lineage covers what happened inside DataLens. Transformations applied before the data arrived are outside its view.
- A dataset loaded and never transformed has a short lineage — a source and a current state.
- The Dictionary that adds business meaning to lineage is a Pro capability.
Related how-to guides
Profile a dataset and find PII
See what is really in each column, and find the personal data before it ends up somewhere it should not.
Explore a source model
See the tables, columns and relationships of a system you did not design — as a diagram you can read.
Clean a messy dataset
Drop nulls, fill gaps, remove duplicates and cap outliers — with a live preview before anything is applied.
Generate a data document
Produce an architecture document, a model spec, a mapping or a runbook from what you have already built.
Related questions
More of these on the DataLens FAQ page.
Can I trace where a data field came from in DataLens?
Yes. The Lineage tab under Govern shows a graph from the original source through every transform step applied to the dataset, to its current state. Following a column backwards shows exactly where it was cast, split, derived or renamed.
Can I see what would be affected if I change a source system?
Yes — read the same lineage graph forwards. It shows what was built from a given source, which is the impact analysis you need before changing or retiring one.
Does lineage cover transformations done before the data reached DataLens?
No. Lineage records what happened inside DataLens. Anything applied upstream — in an ETL job or the source system itself — is outside what it can see.
Try this in DataLens
DataLens is in private beta. Request access and work through this guide on your own data.
Request beta access