Automatic identification of sources of web metric changes

A system, comprising:

• one or more processors;
• and a memory having instructions that are executable by the one or more processors to implement an anomaly-source detection module that is configured to:
• analyze values associated with a plurality of categorical variables corresponding to a web metric that indicates a measurable attribute associated with a website;
• determine deviations between expected values and actual values associated with respective categorical variables for a particular time duration, the expected values being estimated based on patterns identified in historical data corresponding to the plurality of categorical variables over a control period of time, the actual values being collected during the particular time duration;
• convert the deviations into comparable statistics for the respective categorical variables as a function of one or more standard normal distributions of the deviations;
• compare relative amounts the respective categorical variables contribute to a change in the web metric based on conversion of the deviations into the comparable statistics;
• use the compared relative amounts to identify the plurality of categorical variables that correspond to sources of the change in the web metric;
• and
• generate a user interface of a web analytics tool in part, by configuring the user interface to include indications of the sources of the change based on the identified categorical variables.

Non-transitory Computer-readable storage media comprising stored instructions that are executable by a computing device to implement an anomaly-source detection module configured to perform operations comprising:

• determining a plurality of dimensions of a web metric that indicates a 10 measurable attribute associated with a website, one or more said dimensions having a plurality of dimension elements;
• collecting data that indicates an actual distribution of values corresponding to the plurality of dimension elements of the one or more said dimensions during a particular duration of time;
• estimating an expected distribution of the values corresponding to the plurality of dimension elements of the one or more said dimensions that are expected during the particular duration of time based on historical data;
• applying a statistical algorithm to calculate deviations between the expected distribution and the actual distribution of the values;
• assigning scores to respective dimension elements that indicate a relative amount a respective dimension element contributes to a change in the web metric, the relative amount a respective dimension element contributes to the change being comparable to relative amounts other respective dimension elements contribute to the change in the web metric, and the scores determined based on comparable statistics for the respective dimension elements computed as a function of one or more standard normal distributions of the deviations;
• identifying at least one source of the change in the web metric based on a comparison of the scores and identification of which said dimension elements correlate to the scores;
• and
• generating a user interface of a web analytics tool in part, by configuring the user interface to include an indication of the at least one source of the change.