Web analytics
Insights
Funnels, trends, histograms and cohort tables that answer questions spanning weeks or months, saved as cards below the day.
Insights answer questions that span weeks or months. Add them with Add insight below the day. Everyone on the domain sees the same cards in the same order, and each card has its own period. Edit, move or remove a card from its menu.
Your coding agent can build insights for you through the MCP server, such as "add a funnel from first open to first merge, and a weekly cohort table".
Funnel
Counts how many events reached each step, in order, with the conversion and drop-off between steps. Each conversion shows its 95% interval, the range that likely holds the true rate.
- Each step is an event with optional property filters. A filter for every step, such as
cohortbetween two dates, keeps the steps comparable. - Pick People instead to follow each visitor or signed-in user through pages and goals.
- Under Chart, pick Columns to draw a bar for each step with the drop between bars, or Rows to fit long step names in half the width.
Trend
Counts up to 5 events per day, week or month, each with its own filters, on one chart.
Histogram
Shows how often each value of a numeric property occurs, with its mean and median. Set the bar width to group values.
Cohorts
Groups a start event by the date in one of its properties, cohort by default, and shows the share of each group that sent each later event.
- A bar shows the 95% interval of each share, and a tick the typical share of the 4 groups before it.
- A green or red arrow marks a share above or below that typical share, beyond noise.
- A hollow dot marks a group too young for 95% of an event, so its share can still grow.
Measure a product without user IDs
Funnels, trends, histograms and cohorts count events, not people, so they work for apps that send no user ID.
| Question | What to send | Insight |
|---|---|---|
| Where do new installs drop off? | Each milestone at most once per install, such as first_open and then first_project, with a cohort property set to the date of the first open. | A funnel of the milestones. |
| How does each week of installs do? | The same milestone events. | A cohort table. |
| Daily, weekly and monthly users | One event per active day, with properties that say whether it is the first day of its week or month. | A trend with those filters. |
| How habitual are your users? | One event per month with the number of active days. | A histogram. |