Anomaly Detection overview

You can view and analyze data anomalies contextually within Analysis Workspace. Contribution Analysis works in conjunction with Anomaly Detection to help identify what contributed to the anomaly.

Anomaly Detection

Anomaly Detection provides a statistical method to determine how a given metric has changed in relation to previous data.

Anomaly Detection allows you to separate “true signals” from “noise” and then identify potential factors that contributed to those signals or anomalies. In other words, it lets you identify which statistical fluctuations matter and which don’t. You can then identify the root cause of a true anomaly. Furthermore, you can get reliable metric (KPI) forecasts.

Examples of anomalies you might investigate include:

Both Anomaly Detection and Contribution Analysis are core workflows in Analysis Workspace. You can run Contribution Analysis against any daily anomaly and embed the result in your Analysis Workspace project.

Analysis Workspace’s anomaly detection algorithm includes

Contribution Analysis

Contribution Analysis discovers hidden patterns within your data to explain statistical anomalies and identify correlations behind unexpected customer actions, out-of-bound values, and sudden spikes or dips for selected metrics across convergent audience segments.

Something happened. Why? Your Anomaly Detection report shows an unusual spike in orders and you want to know why. What happened out of the ordinary? Who is responding to what campaign or referral? Did something go viral? What are the specific factors that contributed to this anomaly? And perhaps most importantly: How can I capture important information about my customer and repeat this performance? (Or, if a dip in a metric or rise in a negative metric occurred, how can I avoid it in the future?)

Contribution Analysis helps you evaluate your data immediately to answer why an anomaly happened. It breaks down the contributions to an anomaly in seconds in what used to take weeks, providing patterns for audience segments and helping you develop a narrative for customer interactions. You can employ Contribution Analysis strategically to identify and capture meaningful associations to develop new audience segments, or use it tactically to identify out-of-bound or fraudulent activity that triggers an alert.

Anomaly Detection identifies data spikes and extreme statistical dips based on selected metrics and selected audience segments. It sets an historical norm based on a training period and then plots extreme offsets that correlate to specific events. It can report a precipitous rise in a positive Orders metric or a rise in a negative Bounces metric, or dips in both, capturing statistically relevant data points to be evaluated by Contribution Analysis. Once a statistical anomaly is identified, Contribution Analysis lets you drill down and evaluate relevant marketing and campaign variables across all anomalous data points. It runs advanced algorithms and machine-learning processes to evaluate associations that contributed to a significant spike or dip. These calculations are then displayed in interactive visualizations designed to give you varying perspectives to help answer why something happened, and what to do about it.

Contribution Analysis helps you develop a narrative to describe why an anomaly occurred and how to respond to it, capturing relevant metrics and identifying hidden points that give you an overall reason for audience interactions and trending customer interests. Sometimes an anomaly is easy to see and correct, such as an errant order for 2,000 kayaks. Sometimes it’s complex, like identifying an emerging trend across a time period in a region reacting only to a specific targeted campaign. Piecing together contributing items across metrics for various dimensions and their associations gives you an overall idea of your audience interactions and helps provide context for anomalous data points.

Here are a few use cases:

After a comprehensive analysis of an anomaly, a Contribution Summary is generated for Top Items ordered by total occurrences and the item’s percentage of the contributing values. A normalized Contribution Score lets you easily compare, contrast, and associate with other significant dimension items.

Contribution Analysis Tokens - overview section_3EF8D2BBCE6E4C309D753BCF04A453D0

All customers with a Contribution Analysis entitlement can run full Contribution Analysis a limited number of times per month in Analysis Workspace. This excludes point product (SiteCatalyst 15) customers, Analytics Foundation customers, and Analytics Select customers, who do not get Contribution Analysis at all.

The number of runs per company is limited by monthly tokens that are granted based on the Adobe Analytics product your company purchased. This includes the ability to restrict Contribution Analysis access to avoid token misuse.

Frequently asked questions section_11D0431AD2014B96AB9561CA66A367CE

Anomaly Detection and Contribution Analysis entitlements section_9278D58F21A840AA9B1ED1BD07A1EF0A

Below is a list of the detailed entitlements for Anomaly Detection and Contribution Analysis in Analysis Workspace.