A baseline is the recorded starting value of a metric, used as the reference point for everything measured afterwards. Improvement, degradation and targets are all expressed relative to it.
A baseline is internal and historical. It is where you were.
A benchmark is external and comparative. It is where others are.
Confusing them produces a specific error: treating a peer figure as a starting point, then reporting progress against a number your own organization never actually measured.
The exact definition used, stored with the value rather than remembered.
The period covered, and whether the figure is a point-in-time reading or a period average.
The population in scope, including which entities, regions and asset classes were excluded and why.
The source systems the data came from.
Without the last two, the baseline cannot be recomputed, and a baseline that cannot be recomputed cannot survive a challenge.
Almost always because the organization changed and the baseline did not.
An acquisition adds four thousand endpoints. A divestment removes a business unit. A source tool is replaced and counts the population differently. A subsidiary that was out of scope comes into scope.
None of these are errors. Each of them means the current value and the baseline are describing different populations, and the improvement being reported is partly arithmetic.
Re-baseline deliberately when scope changes materially, and record the reason and the date as an event on the series.
Where possible, recompute the historical value under the new scope so the comparison stays honest. Where that is not possible, mark the break in the chart. See restatement.
Never let a scope change pass silently into a trend line. That is the single most common way a measurement programme loses credibility, and it is usually discovered by someone outside the team.
From the blog