Insights and dashboards

An insight is a question you ask of your events in Analytics → Explore: how a number moves (a trend), how many people get through a sequence of steps (a funnel), how many come back (retention), what people do in order (paths), how often they repeat something (stickiness), and who is new, returning or gone (lifecycle). Add any insight to a dashboard to keep it.

Build an insight

Open Analytics → Explore. The answer is in the middle; the panel on the right sets it up, top to bottom:

1. Insight → Type: what kind of question it asks. Trends is the default.
2. Period: the Range, and for answers drawn over time, Group by day, week or month.
3. What it measures: a trend's Series, a funnel's Steps, retention's Start with and Come back with, or a path's start and end.
4. Filters: narrow everything at once. Each series or step also has its own filters, under it.
5. The type's own options, and for trends, Display.

Everything you choose is kept in the page's address, so you can bookmark or share a view. To keep it on a dashboard, click Add to dashboard.

A filter reads property, comparison, then values. Comparisons are is, is not, contains, doesn't contain, starts with, ends with, matches regex, doesn't match regex, is set, is not set, greater than and less than. Funnels, retention, paths, stickiness and lifecycle filter on event properties; a dashboard filter they can't apply is marked Not applied.

Insight types

Trends

Each series measures MRR, revenue or an event. An event counts as a Total count, Unique people, Unique sessions, First-time people, Weekly active people, Monthly active people, Count per person, or the sum, average, minimum, maximum, median or a percentile of a Property value.

• Formula: click Add formula under the series and combine them by letter, such as A / B * 100, with +, -, *, / and parentheses. The formula is drawn in place of the series. Where a series has no value, or a division is by zero, the formula has none.
• Display: the Chart (Automatic, Line, Stacked bars, Stacked area, Number, Table, Pie or Bars by total), Smoothing over 7 or 28 days, Cumulative, a logarithmic scale, values on the points, stacked charts as a percentage of their total, and goal lines.
• Cumulative adds each period to the ones before. Unique people, unique sessions, and weekly or monthly active people are counted once each from the start of the range, so a person who comes back isn't counted again.
• Compare to draws the Previous period of the same length, or the Previous year: the same dates a year earlier, with weeks lined up on the same weekday.

Funnels

• Window: how long after the first step the last may come.
• Breakdown: split everyone by an event property. Attribution says which value counts: the one on their First step, on the Last step reached, only a value All steps share, or the one on A specific step.
• Exclusions: people who do an event between two steps leave the funnel.
• Show: Conversion steps, Time to convert (how long people took, with the average and median), or Historical trends (conversion over time). Relative to shows each step as a share of the first or the previous one.

Retention

People join a cohort when they do the start event, and count as returning in each later period they do the return event (the start event again unless you choose another).

• Starting: Recurringly puts people in every period they start in; For the first time only in the period of their first start.
• Returns on or after: a period counts everyone who came back in it or any period after.
• Relative to: each period as a share of the Cohort size, or of the Previous period.
• Breakdown: one curve per value of a property on the start event. In the table, a value expands to its cohorts.

A period that hasn't ended yet is hatched in the table and left out of the curve.

Paths

A path is one session's events in order, so paths count sessions. Click a step to start or end paths there, or to exclude it. With a Breakdown, each session takes the property's value from its start event (or its end event when only Ends at is set), and tabs above the chart show each value's paths.

Split by cohort

Compare up to five cohorts in one insight. In a trend's Split by, or a funnel's or retention's Breakdown, choose Cohorts, then pick them. Each cohort gets its own line, bar or row, named after it, however few people it holds.

The variants of your A/B tests are in the same list, named with their test, so a test's variants compare on any insight like cohorts do.

Someone in two cohorts counts in both, and people in none of them count only in the total, so the cohorts don't add up to the total. A split by cohorts therefore never stacks: Chart offers Bars and Area side by side instead of stacked, and no shares of the total are shown.

A person's cohorts are the same at every step, so a funnel broken down by cohorts has no Attribution. Funnels counting people or sessions, and retention counting people, can be broken down by cohort; paths, and insights counting groups or a property's values, can't.

Who is counted

Funnels, retention, stickiness and lifecycle count people unless you change Count. Funnels can count Sessions, so one person with two sessions counts twice. All four can count A property's values, such as a company ID, so each company counts once however many of its people took part.

See the people behind a number

Click a number on a chart, in Explore or in a dashboard tile you've opened, to list who it counts:

The list opens beside the chart, titled with what you clicked, such as "Dropped off after step 2 · pro", and says how many it holds. Where a click covers more than one number, the menu at its top switches between them: each line's point on that day, a step's Reached and Dropped off, a period's Converted, Entered and Dropped off, or each lifecycle status.

Each row opens the person's profile and shows what put them there: how many times they did the event, what they paid, their MRR, how long they took to convert, or in how many periods they came back. Search finds people by name, email or ID, and Load more reads the next 50. The list always matches the number you clicked. People who were never identified are shown by their anonymous ID, and people deleted since as Deleted person. When an insight counts companies or other groups, the list shows the groups, each opening its page. A formula's result isn't any one series', so it lists no one.

To keep the people for later, choose Save as cohort and name it. It makes a static cohort of everyone behind the number, not only those listed or found by a search, up to 10,000 people. Visitors never identified and people deleted since can't be in a cohort, so they're left out and Studio says how many. Seeing the people and saving them need access to People.

Dashboards

A dashboard arranges insights, built-in charts and notes as tiles, viewed over one range.

1. Open Analytics → Dashboards and create one, or click Add to dashboard in Explore and choose New dashboard.
2. From a dashboard's More menu, choose Add insight, Add built-in chart or Add note.
3. Choose Edit layout to arrange the tiles.

Click a tile to see it larger and change it. Each tile's menu can rename, duplicate or remove it. The dashboard's range, Compare to and filters apply to every tile. You can change them for yourself, or save them as what everyone sees.

If something's off

• An insight shows nothing: the event name doesn't match what your app records. Check Activity → Analytics events.
• Previous year shows no data: your project has no events from a year earlier.
• A funnel counting sessions is empty: your app doesn't send session IDs with its events.
• Fewer than five people: numbers narrowed to fewer than five people stay hidden to protect them.

Next steps

• Query with VoidQL
• Reports
• Events (developer docs)