The historical series

From the second study onward, Censuma stops showing you a photograph and starts showing you a film. The Statistics module cuts across studies: it compares between them, so it does not live inside any one of them.

What you can follow over time

• The indices, each with its own line.
• The categories, to see whether the one you attacked last year moved.
• A single question, with its acceptance percentage and the distribution of responses study by study.
• The targets: what you set against what came out.

In the By question tab you first pick the section (Satisfaction, Work climate, Recommendation or Additional) and then the question, just as inside a study.

Which years to look at

Statistics opens with the last 20 years. The «Year range» filter starts there, so in 2026 it reads «From 2006» even if nobody measured that year: it is the edge of the range, not a study.

• From and To only offer years with studies for the branch you picked, and they do not let you build a reversed range.
• To see the whole history, remove the range with the × in the filter strip.
• Reports you download from Statistics come out with the range and the segment applied.

Inside a chart there are two more aids:

• The range bar. When the series has four studies or more, a bar with the whole series in miniature appears below the chart. Drag its edges to zoom into a stretch, and the chart above shows only those studies.
• «Zoom in». Brings the vertical axis close to the series' values so small variations show. While it is on, the chart warns that the axis does not start at zero: a small change looks big, so read it with that in mind.

Gaps are information

When a point in the series does not clear the confidentiality threshold, it is drawn as a gap and the line does not cross it.

Joining the two neighbouring points and moving on would look tidier. It would also be a lie: that straight line would assert a trend across a period nothing is known about. The gap says exactly what happened —there was no publishable data there— and that is a fact in itself.

The same goes for periods with no study. An empty row is not a zero.

And for an index a study did not measure, because the organization had it turned off (see Satisfaction and Climate): when you look at that index, that study simply does not appear, and the line runs from the previous study that measured it to the next one. It is not a break, because nothing changed in the instrument.

When a question «breaks the series»

If a question changes what it asks, its previous results stop being comparable with the new ones. The line splits into two segments that do not join.

The question stays alive and the old responses are not lost. What is lost is the right to draw a trend between the two halves, because they are not the same question.

Fixing a typo should not break anything — but fixing it while a study is collecting is a different problem. If you need to correct wording, wait for the study to close. See Questionnaires and inheritance.

What ends a series without looking like it

Three changes feel administrative and are actually instrument changes. Worth knowing before you make them rather than after looking at the chart:

Changing theoretical model. The most expensive one. That index's history restarts from zero, because they are two different instruments, and the same happens when you duplicate a custom theory to change it. Statistics lets you choose which stretch to look at, one per model you have used. See Theoretical models.

Reorganizing the tree. Results are published per node. If you merge two branches or split one into three, the new nodes have no past: the old branch's series exists, but it is not any of the new ones'.

Changing the confidentiality threshold. It rewrites nothing backwards —each study freezes its own— but from now on points are published against a different bar. A series where half the points were measured under one threshold and half under another is not wrong, but you need to know it when reading.

Comparing against other companies

Cross-organization comparison exists because the fixed sections are identical for everyone — but only among those measuring with the same model. Two companies on different models are not comparable, for the same reason you are not comparable with yourself after switching. For the same reason, a custom theory is not compared with any other company: only with its own series.

One reading tip

Two points are not a trend. With two studies, all you know is that the second came out different from the first, and the possible causes include the survey going out at a different time of year, or different people answering.

A series starts saying something around the third point. Until then, treat each study as a photograph.