Free-form charts and anomaly detection

The Free form technique has six visualisations; all read the same Rows/Values slots, so you can prototype in a table and flip to a chart.

• Donut chart: share of one metric across one dimension; the breakdown dimension is the first row dimension. Good for channel mix or device share, limited to a handful of slices with Other.
• Line chart: a metric over time. Granularity is day, week or month; the breakdown dimension (optional) draws one line per value, capped at a few top values. This is the only visualisation with Anomaly detection: tick it in Tab settings, set a training period (how many prior days to learn from, 7 to 90) and sensitivity (the probability threshold). Points outside the forecast band are circled, which is exactly the signal behind the Insights cards in Reports, now under your control.
• Scatter plot: two metrics as x and y, one point per dimension value, optionally a third metric as bubble size. Use it to find pages with high views but low engagement, or campaigns with high cost but few key events.
• Bar chart: horizontal bars for one metric across one dimension, with an optional breakdown.
• Geo map: a metric by Country, Region or City on a world map.

Line and bar charts honour segments as separate series, so applying two segments gives a direct visual comparison.

Two gotchas. First, charts use the same date range as the tab, and the line chart's granularity cannot exceed what the range allows. Second, (not set) and (other) rows appear in charts too; add a filter to drop them before presenting.

Export a chart as PDF, or screenshot it; charts do not export to Sheets, only tables do.