Cohort Analysis with Analysis Workspace

[TanmayMathur]

What Is Cohort Analysis?

A cohort is a group of people sharing common characteristics over a specified period. Cohort analysis is useful, for example, when you want to learn how a cohort engages with a brand. You can easily spot changes in trends, then respond accordingly. (Explanations of cohort analysis are available on the web, such as at Cohort Analysis 101.)

After creating a cohort report, you can curate its components (specific dimensions, metrics, and segments), then share the cohort report with anyone. See Curate and share a project

Example Cohort Report

A cohort report returns visitors. Each data cell shows the raw number of visitors in the cohort who did the action during that time period.

Examples of what you can do with cohort analysis:

• Launch campaigns designed to spur a desired action.
• Shift marketing budget at exactly the right time in the customer lifecycle.
• Recognize when to end a trial or an offer, to maximize value.
• Gain ideas for A/B testing in areas such as pricing, upgrade path, and so on.
• View a Cohort Analysis report within a Guided Analysis report.

• Run a Cohort Analysis report
  Create a cohort and run a Cohort Analysis report in Analysis Workspace.
• Cohort Analysis Use Cases
  Use case examples for cohort analysis.

[JoeBlackwell]

Hi, I have been looking at the cohort analysis and it appears quite constraining. If i run it over a historic time period say 1st Dec 2014 - 15th Dec 2014 on a daily basis it doesn't populate all the following days i.e. 15th Dec 2014 has no data for Day 1 onwards. This data does exist as it is a historic date period. The only day you get the full day 1 to day 14 return data is for the very first day in your time period. Am i doing something wrong here? It is very frustrating.

[Jeremy_Brugger]

Is there a way to apply segments to either the Inclusion metric OR Return metric (instead of both)?

Also, it looks like it's not possible to use calculated metrics as cohort metrics. If it worked, it would be a nice workaround for my above question.