Building a chart
A chart is one or more series plotted over a shared time axis. Every choice below stays editable after saving.
Series
Pick a measure from any dataset's concept list on the left to add it to Y. Series may come from different datasets — they all align on the same time axis. Each series has its own aggregate (sum, avg, min, max, count).
Split by
Splitting one series by a dimension (or CATEGORY) automatically fans it out into one sub-series per value — you don't need to add a series per value yourself. On a stacked bar each value becomes one segment of the pile; on a plain bar each value is its own grouped bar; on a line each value is its own separate line.
Stack
Only matters for stacked bars. Series sharing the same stack name pile into one column. Give a series a different stack name to put it in its own column next to the others — e.g. two separate stacked totals side by side on the same chart.
Axis
Left or right Y axis. Left blank, series sharing a unit share an axis automatically and the first unit takes the left. Set it explicitly when combining measures with different units, e.g. Kg on the left and R-million on the right.
Bucket, period & filters
Bucket sets the X-axis granularity (minute, hour, hour of day, day, month, year). Filters restrict a dataset to a dimension's values — use them to isolate the segment a question is about instead of charting everything.
Tags are multi-valued, so a tag filter matches a row carrying ANY of the picked values, and splitting by a tag counts a row under each of its tags — per-tag figures are right, but they add up to more than the ungrouped total.
The "Peak in a…" aggregates plot the busiest single minute/hour/day instead of the total, which is how a burst that a summed timeline flattens out shows up. Put a dimension on the X axis (e.g. operator) and the tallest bar is whoever concentrated the most activity into one window.
The gap aggregates measure the seconds between each observation and the previous one in the same series — median, fastest, and how many arrived under a minute apart. A steady machine-like cadence and a person working in clumps give the same totals; only the spacing tells them apart. Needs a dataset whose period kind is "Date & time", and a short gap is a question to ask about the source system, not a conclusion on its own.