Exploration

Explore your data using tables and graphs.

This help center article is part of the App + Web Property Beta.

App + Web properties are not currently supported in the Analytics app.

The exploration technique presents your data in a flexible table layout. You can arrange the rows and columns as you like, create data groupings, and apply segments and filters to refine your view. Add the metrics you're most interested in. If you spot a significant data point, create a segment from it and use it in other analyses.

The example below explores the relationship between Device Category and Screen Resolution by Country, measured by number of Users and Revenue. The table reveals that while the screen resolution for the majority of desktop users is 1366x768, the bulk of the revenue is generated by users on 1440x900 screens. You could create a segment from this data point and use that to dig deeper into this audience's behavior.

Exploration example

You can also display the data using a number of different visualizations: this example shows a line chart comparison of the Mobile vs. Organic Traffic segments applied to the previous data table:

Line chart example

Configure the exploration

Set up the exploration with these options:

Common Options Description
Visualization

Switch between chart types.

Segment comparison Apply up to 4 segments to the analysis.
Filter Restrict the data shown in the analysis according to the conditions you provide. Filter clauses are applied using AND logic.
Table options
Pivot Display segments in the table as rows or columns.
Rows Display up to 5 dimensions as rows in the table.
Start row Select the starting row in the table.
Show rows Set the number of rows to show in the table.
Columns Display up to 2 dimensions as columns in the table. Using multiple dimensions creates column groups.
Start column group Set the starting column group in the table.
Show column groups Set the number of column groups to display in the table.
Values Display up to 10 metrics in the table.
Cell type Display metric values as plain text, bar charts, or heat maps.
Pie chart options
Breakdown The dimension used to provide the breakdown data series for the visualization.
Row limit Set the number of data series displayed in the visualization.
Values Display a single metric in the chart.
Line chart options
Granularity Set the date interval for the chart. The week interval starts on Sunday. The month interval starts on the 1st day of the month.
Breakdown The dimension used to provide the breakdown data series for the visualization.
Lines per dimension Set the number of data series displayed in the visualization.
Values Display a single metric in the chart.
Anomaly detection Turn anomaly detection on or off. See below for more information.
Training period (last days) Increase or decrease the timespan used to examine your data. Longer training periods can increase accuracy.
Sensitivity Set the probability threshold value, below which anomalous data will be reported. A higher sensitivity value may result in reporting more anomalies.
Scatter Plot options
Breakdown The dimension used to provide the breakdown data series for the visualization.
Y Axis The metric used on the vertical axis
X axis The metric used on the horizontal axis
Geomap options
Geo breakdown The location dimension used to provide the breakdown data series for the visualization.
Points per dimension Set the number of data point to show in the visualization
Values Display a single metric in the chart.

Anomaly detection

You can identify outliers in your data by using anomaly detection in a line chart. This option is on by default, and you can configure the detection model with these two settings:

  • Training period (last days) determines how many days prior to the beginning of the the currently selected date range are used by the anomaly detection model to calculate the expected value of the metric displayed.
    • For example, if your currently selected date range is the first 10 days of the month, and you set the training period to 7 days, the anomaly detection model will use the data from the 7 days prior to the beginning of the month.
  • Sensitivity sets the probability threshold below which anomalous data will be reported. Sensitivity doesn't affect how the model "thinks": rather, it just specifies how data should be labelled. The probability of a point occurring at a particular value is predicted by the model and is not influenced by sensitivity.
    • For example, a sensitivity of 5% means that any point occurring with a probability of less than 5% is considered anomalous. Therefore, a higher sensitivity model might result in reporting more data as outliers.

Once the Anomaly detection model has been defined, Analysis applies a Bayesian state space-time series model to the training data to forecast the value of the metric displayed in the time series.

Finally, Analysis flags the datapoint as an anomaly using a statistical significance test with p-value thresholds based on the selected sensitivity.

Was this helpful?
How can we improve it?