> ## Documentation Index > Fetch the complete documentation index at: https://docs.nebuly.com/llms.txt > Use this file to discover all available pages before exploring further. # Retention charts Retention charts measure how well users return to your AI application over time. Users are grouped into **cohorts** based on when they first interact with your application, and Nebuly tracks what percentage of each of them remains active in the following time periods. A retention chart answers questions such as: * How many new users come back after one week? * Which groups have the highest long-term engagement? * Has user retention improved over time? Each row represents a **cohort** of users, while each column represents a period after the cohort was created. Every cell shows the percentage of users from that cohort who returned during that period. For example, if a cohort contains **100 users** and **45** of them return one week later, the **Week 1** retention is **45%**. ## Retention settings ### User groups Optionally, select one or more **User Groups** to calculate retention only for a specific audience. If no user group is selected, retention is calculated across all users.

### Retention type The **Retention type** determines how Nebuly calculates whether a user is retained. * **Anchor** tracks users from their **first interaction**. Once a user enters a cohort, Nebuly measures whether they return in each subsequent period (Week 1, Week 2, Month 1, etc.). This is the classic cohort retention analysis and is ideal for measuring long-term engagement. * **Rolling** measures retention using a rolling time window. Instead of anchoring users to their first interaction, Nebuly continuously evaluates whether users remain active from one period to the next. This is useful for understanding ongoing user engagement without focusing on acquisition cohorts. ### Retention frequency Choose how cohorts are divided over time. Available frequencies include daily, weekly, monthly For example, **weekly retention** creates one cohort per week and measures retention week by week. ### Starting date The **Starting date** defines the earliest cohort included in the analysis. Only users whose first interaction occurs on or after this date are considered. ### Filters Apply report filters to calculate retention on a subset of users or interactions. For example, you can measure retention only for: * Users from Germany * Conversations tagged as **Support** ## Understanding the retention chart The retention chart consists of two complementary views: * **Retention curve** (top) – Displays the retention trend for the selected cohort(s). Each point represents the percentage of users who returned during that retention period, making it easy to visualize how engagement changes over time. * **Retention table** (bottom) – Displays the exact retention values for each cohort and period. Each **row** in the table represents a cohort of users grouped by the selected retention frequency (daily, weekly, or monthly). The **Users** column shows the number of users in that cohort. Each **column** represents a retention period: * **Month 0** (or Week 0 / Day 0) is always **100%**, as it represents the users when they first entered the cohort. * **Month 1**, **Month 2**, and subsequent columns show the percentage of users from the original cohort who returned during those periods. The **Average retention** row summarizes the average retention across all cohorts included in the chart. You can select one or more cohorts using the checkboxes on the left. The retention curve updates automatically to display only the selected cohorts, allowing you to compare how different cohorts retain users over time.