Change explorations

In this article:
In the web version of Analytics, the feature is available only in English. In the Analytics app, regardless of the language version you are using, you can ask questions and receive responses only in English.

About change explorations

When a metric value changes from one time period to the next, change explorations identifies segments of your data that are likely to explain the change. For example, if you see that Revenue dropped 15% last week, you can ask Intelligence a question like, “Why was Revenue down last week?” and see the segments that changed the most for that metric week over week (e.g., Default Channel Grouping = Organic Search, Keyword = backpacks).

See change explorations

You can see change explorations two ways:

  • Ask Intelligence "Why?" questions via the question box.
  • Use the Explore change option featured in some Intelligence Insights about anomaly detection, which runs a change-exploration analysis and surfaces it in the card.

How change explorations work

  1. Ask Intelligence a question about “why” a metric changed. Your question should include:
    • The word “why”, so that Intelligence knows you want a change exploration.
    • The metric you want to evaluate (subject to current limitations). You can evaluate filtered metrics.
    • The two date ranges you’d like to compare (e.g., May vs. June). If you do not specify an explicit date range, Intelligence selects a default.
  2. Next, Intelligence calculates the change across the date ranges. If the change is less than 1%, Intelligence does not provide a change exploration.
  3. Then, based on the metric, Intelligence explores changes across 3-5 dimensions that often drive changes in the metric. For example, for Users, the dimensions might be:
    • User Type
    • Default Channel Grouping
      Within a specific channel, change explorations explores closely related dimensions. For example, if the Display channel changed significantly, change explorations will scan the Campaign dimension to see if a particular campaign drove most of the change.
    • Country
  4. Finally, Intelligence creates thousands of segments based on each possible value for each of the exploratory dimensions (e.g., Default Channel Grouping = Direct, Display, Social, etc.). For each segment, Intelligence calculates a “score” for how much the segment changed (the score is a combination of absolute and relative change). Then, it presents the 3-5 highest scoring segments across the time periods and shows how much the segment changed.

Example

Let’s say you notice that your Users are down last week. You could ask, “Why were my users down this week?” and receive the following answer.

Week over week changes in metric values

Limitations

Metric restrictions

Currently, Analytics provides change explorations for only summable metrics like Users, Pageviews, and Revenue. Analytics does not provide change explorations for ratios, like E-Commerce Conversion Rate or Bounce Rate.

Change thresholds

The metric value must have changed by at least 1% across the two time periods in order for Analytics to show a change-explorations answer. Note that if the metric changed by less than 5%, change-explorations results may have limited explanatory value.

Explored dimensions

Currently, for each metric, we have a static, internally defined list of dimensions. You cannot yet choose which dimensions you would like to explore.

Was this article helpful?
How can we improve it?