AXON contribution is AT Internet’s new data science powered feature specifically designed to draw insights efficiently from your data. Seamlessly integrated throughout Explorer, AXON Contribution will do the heavy lifting so you can spend more time on higher value tasks.
This feature will filter through all your data points to identify and quantify significant trends in your underlying data.
AXON contribution’s power can be unlocked in two clicks: (1) simply activate Data Science mode by clicking on the light bulb icon in Explorer’s graph type selector then, if present, (2) click on an anomaly of your choice!
AXON Contribution can be activated by clicking on any timeseries graph within Explorer. To do this, activate the anomaly detection feature by clicking on the ‘lightbulb’ icon on Explorer’s graph type selector, if anomalies are present in daily granularity, you can click on any anomaly of your choice to launch AXON Contribution.
AXON Contribution is available on the daily graph granularity. The feature will be extended to other date granularities in the future.
AXON Contribution is available on any absolute metric of your choice (available on all metrics which are not ratios).
Due to the calculation power required to run the AXON Contribution algorithm, a limit of 100 uses per day per organisation has been put in place in line with fair use principles.
Our built-in anomaly detection algorithm identifies how your data should evolve within a tolerance area. If your curve leaves this tolerance zone an anomaly will be identified and highlighted with the positive or negative anomaly icons.
The difference between where your data should lie (baseline - the dotted line in the screenshot below) estimated by our algorithm and where your curve lies.
Dimension & dimension values
A dimension is an attribute of an event. Device, Traffic Source, Browser type are all dimensions. Dimensions have values, such as desktop, smartphone, tablet for the Device dimension.
The share of the total difference attributed to a specific dimension value.
Interpretation of the results
Each bar chart reprensents one dimension.
The navy-blue bar represents 100% of the anomaly that we are looking to explain (the gap between expected data and actual data of the overall anomaly).
The light blue bars represent each value’s share of contribution to the overall anomaly.
For exemple, in the exemple above, Traffic from the netherlands contributed to 100% of the anomaly. In addition, traffic from Google Chrome contributed to approximately 50% of the anomaly, Firefox 45%, and 15% of the anomaly can be explained by traffic from the Opera browser.
NoteWe show the top 3 values for each dimension with the same sign as the overall anomaly. For exemple, if we are analysing a positive anomaly, we will show the top three positive values. Conversely, we will show the top three negative values for negative anomalies.
NoteAlthough we only show the top 3 values for each dimension, the sum of all the values' contributions is equal to the overall contribution (the navy-blue bar).
Order of the results
Order of the bar charts
The charts are ranked in decreasing order on their top value's contribution. In the example below, the value 'Netherlands' contributed to 100% of the total difference. The values desktop and Corporate contributed respectively 75% and 70%. Therefore, the dimensions 'devices' and 'level 2 sites' are ranked second and third in the list of bar charts.
Order of the dimensions
For each dimension, within the graphs, the values are ranked in decreasing order on their contribution to the overall anomaly.
Which dimensions are analysed?
The contribution will look for significant results in the following dimensions: device, traffic source, country, OS, Browser type, Level 2, Pages.
I can't find a specific dimension or dimension value
The contribution only shows dimensions and dimension values with significant results for a given anomaly.
I'm analysing segmented data, will the contribution look within that context?
The contribution will look for significant results by keeping in mind your analysis scope and segments.
A value contributed more than the overall anomaly - (A light blue bar is bigger than the navy blue bar)
Sometimes dimension elements work against each other: As an example imagine a positive anomaly of 100 visits above what was expected, where desktop traffic increased 120 visits and smartphone traffic decreased by 20 visits. In this case desktop traffic over-compensated for the decrease in smartphone traffic resulting in a positive anomaly of 100 visits. These are all valuable insights in order to understand what happened on a given date!