This page walks through Adobe Analytics calculated metrics from A to Z, including:

• What are calculated metrics?
• Calculated metric governance
• Using calculated metrics, including limitations
• Calculated metric examples - basic, advanced, and with functions

Calculated Metric Governance

When creating & managing calculated metrics, keep these governance tips in mind:

• Name - Establish a consistent naming convention and ensure all calculated metrics adhere to it. Names should be easily & quickly understood by end users doing the analysis.
• Description - Include a description for every calculated metric to make it easier for end users to know when the metric should & should not be used. Note: Descriptions for events 1-1000 can be managed in the Analytics Admin console.
• Tag - Organize your calculated metrics into 1 or many groups. Tags are then searchable in the left rail of Analysis Workspace, by typing # or clicking the filter icon. Pro-tip: Add tags for the components used in your calculated metrics so that you can easily see which parts of your implementation power each calculation. Here's a helpful article by Adam Greco detailing this tip.
• Share - Standardize calculated metric definitions by sharing them with end users.
• Approve - Gives your end users the added assurance that they are using the right version of a calculated metric, if there are many that sound the same. This will apply an #approved tag to calculated metrics, which users can filter by.
• Unshare & Delete - Ensure your list doesn't become cluttered by pruning your metric list every so often. Adobe is also making strides toward helping you better understand where a calculated metric is being used & when it was last used, so you can have confidence that you aren't deleting something in use. But have no fear - even if you delete a calculated metric in use, we soft-delete it so that your reporting won't break.

Using Calculated Metrics

Calculated metrics are a helpful way to extend beyond the 1000 custom events that you get in Adobe Analytics. Combined with segmentation, there are unlimited metric definitions that you can create. For example, rather than using a custom event to capture Unique Visitors who watched video, you could instead create a calculated metric with a segment applied to the OOTB unique visitors metric.

Additionally, calculated metrics can help as a stopgap if an implementation detail is overlooked or incorrect. While correcting the implementation should be a top priority, you can use calculated metrics to derive a "stopgap" metric that you can use in the meantime.

IMPORTANT NOTE: When deciding whether to set a custom event or create a calculated metric, keep in mind that calculated metrics cannot be used in Analysis Workspace fallout, cohort, and histogram visualizations. They are also excluded from data feeds & data warehouse extracts.

Basic

Adobe has long provided calculated metric templates in the metric builder, denoted by the folder + magnifying glass icon in the left rail. These templates are also now available directly in the left rail of Analysis Workspace. They include some common calculated metric definitions that users create, and can be used as is or as a starting point for a more advanced metrics.

The easiest way to add new calculated metrics to a freeform table in Analysis Workspace is to highlight 1 or 2 metrics, right-click and select "Create metric from selection".

Metrics within metrics - You can use calculated metrics to create new metrics. For example, the denominator below is a calculated metric itself.

Applying segments to metrics - you can apply any segment to your collected events (or OOTB metrics) to create completely new custom metrics.

Can't find a segment you want to build a metric from? The easiest way to create a new segment is to drag in a dimension directly to the metric builder. A hit segment will be created for you by default, and can be modified as needed. The benefit of creating a segment in this way is that 1) you don't have to leave the metric builder, and 2) the created segment will be local to the calculated metric, and won't clutter up your left rail.

Advanced

There are 10 different attribution models that can be applied in Calculated Metrics, if fully provisioned for Attribution IQ (if not, 4 models are available). This allows you to compare attribution models, create hybrid models, or apply attribution to your important conversion metrics such as revenue or orders per visit.

Participation is helpful for attributing content on your site or app to downstream success. The participation model gives 100% credit to each touchpoint leading up to a conversion. Page is a common dimension to do a participation analysis against. Dividing by visits removes traffic variability so that every page can be assessed fairly. When sorting descending, lower visited pages that drive conversion will bubble to the top of your analysis.

Engagement scoring is the process of assigning numerical “scores” to key digital activities. Scoring helps you understand how varying engagement levels affect conversion and customer loyalty. Scores can be setup via calculated metrics (non-permanent approach) and used throughout your analysis as a secondary success measure, and indicator of valuable revenue opportunities.

Note: Engagement scoring can be taken even further with processing rules (permanent approach) that populate a counter eVar that aggregates a score for each visitor. The score eVar can then be classified into high/medium/low engagement groups. Additionally, there are several methods for intelligently determining the score weights, such as through propensity scoring in Data Workbench.

Intelligent Alerts can be setup for calculated metrics to tell you when a metric is behaving unexpectedly compared to historical trends. Alerts can be set to trigger on above expected, below expected, % change by, or anomaly detection (intelligence) thresholds.

Functions

Approximate Count Distinct returns the count of dimension items for a selected dimension. You can then use the metric in any report to understand the count of one dimension against values of other. For example, the count of unique customers or unique products by Marketing Channel.

Deriving Rolling averages using Mean - Rolling averages help to benchmark performance against recent trends. You can use the Mean function, coupled with rolling date range segments, to create rolling averages that will function irrespective to the panel date range selected.

Benchmarking using Cumulative Average - Cumulative will sum data across rows in a table. Cumulative average will average data across rows in a table. Both fluctuate as you move down the table, whereas "rolling average" is fixed to last X days. Dividing a metric against the cumulative average or rolling average gives you a daily performance benchmark.

Calculating top metric-drivers using If, Quartile and Greater Than - Quartile is an alternative to Percentile and can be used to return the minimum value, 25th percentile, 50th percentile, 75th percentile, or maximum value. For example, you can derive a metric that focuses on just the top 75th percentile of revenue or application completes.

Improved sorting using If, Percentile and Greater Than - 'If' statements allow you to set a floor and/or ceiling for any given metric. This is helpful for reports where you want to sort on metrics such as bounce rate & conversion rate, but don't want to see low volume dimensions. Percentile in this function ensures the floor is relative to actual behavior, rather than just an arbitrary threshold that you've set.

Calculating standard deviation using Z-Score - Z-score is the number of standard deviations an observation is from the mean and can be + or -, indicating whether it is above or below the mean and by how many standard deviations. This can be helpful for highlighting dimension items that are above the mean, such as marketing channels that drive above average revenue.

Ready for more content?

Visit adobe.ly/aaresources for a full list of Adobe Analytics Spark pages & other helpful resources.