Aggregate Stats displays a configurable list of values: event counts, distinct customers, averages, sums, quantiles or the most common value. Use it for a concise summary above supporting charts.
Add and edit metric rows
- Add Aggregate Stats to a dashboard panel, or choose it as an investigation widget type.
- Open the widget's settings.
- Add a row, or select the edit control beside an existing row.
- Give the metric a clear Label.
- Select an Aggregate function and Attribute.
- Enter a Filter if this row should use a narrower population.
- Choose the formatting and colour settings.
- Reorder rows using their handles, then save the widget.
Choose the function
| Function | Meaning |
|---|---|
| Count distinct | Number of different non-null values of the selected attribute |
| Sum | Total of a numeric attribute |
| Average | Arithmetic mean of a numeric attribute |
| Minimum / Maximum | Smallest / largest numeric value |
| Quantile | Value at the selected fraction of the distribution; 0.5 is the median |
| Mode | Most common value |
For Count distinct, choose identifier to count events. Choosing the account customer token counts distinct accounts instead. Repeated logins by the same account increase the event count without necessarily increasing the account count.
Null values are excluded from Count distinct. An empty string or placeholder is not automatically a valid customer identifier; add an appropriate condition when your integration uses placeholder values.
Example: a login summary
Create three rows using the account and step names collected by your integration:
| Label | Function | Attribute | Filter |
|---|---|---|---|
| Login events | Count distinct | identifier |
step_name = 'login_submit' |
| Accounts logging in | Count distinct | identity['ACCOUNT'].customer_token.customer_token |
step_name = 'login_submit' |
| Rejected logins | Count distinct | identifier |
step_name = 'login_submit' AND outcome['CHAMPION'].decision_strategy.result = 'reject' |
On a dashboard these rows use the selected dashboard dates. In Investigations they also inherit the executed query's filter. A row condition narrows that population further; it does not replace the investigation filter.
Formatting and colours
Use a prefix or suffix for units, such as a currency symbol or “ms”, only when those units match the collected attribute. Math operation transforms this row's calculated result. The placeholder result refers to that result; for example, result / 100 can convert a stored minor-unit amount to major units when that is how your integration records money.
A math operation is not a reference to other rows. For a ratio requiring two separately filtered counts, use Signal Trigger Rate custom columns when the question is signal-specific, or calculate the two populations explicitly in a suitable analysis.
Choose Single Color for a constant label colour. Scale Color uses minimum, optional midpoint and maximum values and colours. Choose the font size to suit the board layout. These are display choices and do not change the underlying decision policy.
Check the result
Compare a Count distinct on identifier with an investigation using the same dates and condition. For account or device counts, check that those attributes are populated. A zero or blank result can indicate missing attributes, an empty population or an unsuitable calculation; it does not alone establish that there were no customers or no risk.