Whoever asks this question usually already has a register somewhere, or has just started building one, and wants to know whether that effort pays off. That is a fair question. Recording lineage takes time: for each data point, finding out where the figure originates, through which systems and spreadsheets it travels, who adjusts it and on the basis of which rule it is correct or not. You don't want to spend that time on something that then disappears into a drawer.
What changes is not the figure itself. Lineage does not change the outcome of a calculation. It changes who can explain that outcome, and how long that takes.
Without lineage, every question about a data point is a search. Someone calls the previous controller, or searches through email exchanges, or opens ten tabs to reconstruct where a number comes from. That works, until that person is on holiday, changes roles or leaves the company. Then the question of who reads your data point register when you're no longer there is not hypothetical but acute.
With lineage in order, searching becomes pointing. An auditor, a new controller, a supervisory authority asks about the origin of a figure, and the answer is fixed: this system, this link, this assumption, this owner. That is no guarantee that the answer will be pleasing. It is a guarantee that there is an answer, and that it does not depend on who happens to be present that day.
It is more honest to also state here what lineage does not do. Recording lineage does not make a poor data point good. If a figure rests on an estimate that no one can substantiate anymore, then the lineage shows exactly that: an estimate without substantiation. That is progress, because previously no one knew it was an estimate. But the figure itself does not change as a result.
Lineage also does not automatically tell you which of your data points are actually needed. Some organisations record hundreds of data points because a template requires it, while part of them never appear in a report. Mapping lineage makes that visible, but the choice to drop or keep something still lies with the organisation afterwards.
And lineage does not resolve ownership that never existed. If a data point belongs to no one, it remains no one's, even if the origin is clearly recorded. What changes is that the absence of an owner is now noticed instead of going unnoticed.
How much changes depends on where you start. For CO2 figures, the origin is often already largely established, in invoices, consumption data or supplier statements, which is one of the reasons why CO2 is usually the first topic organisations start with. For social or governance indicators, the source is more often an estimate, a manual count or an assumption that was never recorded. There, lineage changes more, simply because there is more to discover.
How much time that takes depends in turn on how many data points already have an identifiable source. An organisation where most of the data points already have a source finishes faster than an organisation that still has to find out per data point where it comes from. That difference is also the reason there is no fixed answer to how long it takes to get a topic in order; it depends on what is already in place and what still needs to be reconstructed.
Three things typically do change once lineage per data point has been recorded.
First: questions from accountants or supervisory authorities take on a different character. Instead of a reconstruction after the fact, there is a reference to a recorded source. That does not change the outcome of the audit, but it does change the time the audit takes and the confidence the organisation can project in its own figures.
Second: changes in systems or reporting standards become less risky. If it is known which data point comes from which system, it is also known what changes when that system is replaced. Without that knowledge, an organisation often only discovers this afterwards, when a figure suddenly is missing or deviates.
Third: the conversation about quality shifts from individual doubt to a shared rule. Where previously someone decided by gut feeling whether a number felt reliable, there is now a recorded rule that explains when a data point is approved and when it is not. That makes the assessment less dependent on who happens to be looking at it.
Once lineage is in place, work remains that lies with people: assessing data points, handling exceptions, keeping sources up to date. Some of those steps are easy to repeat and therefore suitable to leave to a system, others continue to require judgement. Anyone who wants to know which part of that follow-up work can be transferred to AI and which part remains human work will find, in the work scan of FTE TO AI, a breakdown per task, separate from this topic but starting from the same question: what can a system take over, and what can it not.
Vraag maar waar een datapunt vandaan komt. Dat is meestal de hele vraag.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.