A register usually starts as a list. Someone lines up all the data points from the reporting standard, finds a responsible party for each one, and calls that a register. That is the starting point, not the end point. The question "when is this finished" has a concrete answer, and that answer has nothing to do with how many rows the list has.
A data point register is complete when four things have been recorded for each data point. First, the definition: what the data point precisely measures, in which unit, over which period and what scope. Second, the source: the system, the document or the person the figure comes from. Third, the owner: who is responsible for the accuracy of that specific data point, not for the report as a whole. Fourth, a quality rule: a check that lets you see whether the value is plausible, for example a range, a comparison with a previous year, or a sum that must add up.
If any of these four is missing for a data point, the register is not finished at that point. This also applies if the data point itself has already been reported for years. A figure that has been in the report for three years can still lack a recorded source or owner.
The order in which you fill the register determines how much work it takes and how reliable it becomes. Starting with the reporting standard is logical, but the standard does not say what your organization can actually measure. That is why the first step is often the distinction between what is mandatory and what you already record somewhere: which data points you actually need if you work with multiple business units determines the scope of the register before you start on the content.
Next comes the question of where each data point already exists. At organizations with multiple locations or business units, the same data point sometimes sits in multiple places, in a slightly different form. Where a data point already exists if you work with multiple business units is the step that prevents you from collecting the same data twice under two different names.
For the structure of the register itself, including the order of work and the division of roles between who fills it in and who checks it, there is a separate page that describes step by step how you set up a data point register if you work with multiple business units.
In almost every register, a data point turns up for which no clear source can be found. The figure appears in the previous report, but no one can point to where it came from. This is not a reason to reject the register or bring it to a halt. It is a separate category to be handled: data points without a source get a status and a plan, not a guess that gets written down as the source. What you record in that situation is described on the page about what you do with a data point without a source if you work with multiple business units. A register with ten data points in the status "source missing, action assigned" is further along than a register in which those ten data points have silently been assigned a source that no one can verify.
Pointing to a source is not the same as knowing how the figure gets from that source to the report. Between the two there is often a series of operations: adding up, converting, correcting for double counting, aggregating across units. If no one has recorded these steps, the same source data can lead two people to two different report figures, without anyone noticing. This link between source and report line is called source-to-report mapping, and the logic behind it is explained on what source-to-report mapping involves. For more complex data points it is also useful to know exactly which operations sit between source and report, so that a check does not only verify the final figure but also the steps in between.
A register can be fully filled in and still not be finished, if the quality rules have never been tested. Only once someone has actually applied a quality rule to a value, and that value has passed or been rejected on that basis, do you know whether the rule works. So a register is only truly finished after at least one round in which the rules have been used, not merely written down. That is also the moment when it becomes clear which owners can actually answer questions about their data point, and which owners are owners only on paper.
Once the register, the sources and the quality rules are fixed, a different kind of question arises: who carries out the checks, who fills in the missing sources, and which part of that work is repetitive enough to automate. That is a question about capacity, not about structure, and that is what the work scan from FTE TO AI is for: it calculates per task which part of the work can be taken over by AI, so that you know where people remain necessary and where they do 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.