Funnels

When to use it

• Checkout, sign-up or onboarding: “where do we lose people?”
• Marketing site to product: a funnel can start on one source and end on another, because people are stitched across your sources.
• After a change: did conversion between two steps improve?

Build a funnel

1. Go to Funnels and click New funnel.
2. Pick an event for each step. You need 2 to 8 steps, added with + Add step. The picker shows custom events from the last 30 days and autocaptured ones, and you can type a name.
3. Optional: click + Filter on a step to require a property value, such as Page is any of /pricing. Tick values or type one, then click Apply filter.
4. Set the conversion window, what to count and whether order matters (see below).
5. Watch the live preview on the right. It starts with the last 30 days; Widen to 90 days includes older events.
6. Click Save funnel and pick a board to put it on, or Save without a board to keep it in Funnels only. The funnel opens with your filter bar’s date range.

You don’t have to name it. Until you do, the title follows the steps, such as Viewed /cart → Order paid. Click auto-named · click to rename to type your own.

The Insight, Funnel, Flow and Retention switcher at the top of the builder turns the same events into another report type. It opens that report’s page with as much of the funnel as fits. Nothing is saved, and Undo on the toast brings you back.

In a hurry? A new funnel offers three templates under Or start from a template: Signup → activation, Browse → checkout and Trial → paid.

Window, count and order

• Conversion window: 7, 14, 30 or 90 days, with 14 as the default. Someone has this long from the first step to finish.
• Count: Unique people (the default) or Events. Events counts each person-and-session pair as a separate entry, and the results say Counts are sessions.
• Steps must happen in order is on by default. Other events can happen between the required steps. Turn it off and any order within the window counts.

Reading the results

At the top, four cards:

• Overall conversion: from the first step to the last.
• Converted: how many people finished, out of how many entered.
• Median time to convert, with the 90th percentile underneath.
• Biggest drop-off: the pair of steps with the most people whose window expired before they continued.

Under Conversion by step there are three views:

• Steps: bars for each step. Above each bar, a dashed ghost shows people who have not reached that step. Read the label to distinguish those who dropped from those who still have time.
• Over time: conversion by the day people entered the funnel. Click a day for who finished, who didn’t, and their events.
• Time to convert: how many finished within an hour, a day, a week, or longer.

The table below lists every step with people, conversion from previous and from start, how many dropped, and the median time to reach it.

Pending or dropped?

A 7-day window starts when someone enters the funnel. Someone who entered yesterday and has not finished is still pending. Someone who had the full week and did not finish has dropped within this definition. Changing the report dates changes how much of that week the report can observe.

• Still has time: the report runs through now and the person’s window is still open.
• Window runs past these dates: the report ends in the past, before the person’s window ends. A later completion is outside this report.
• Window expired: the report has observed the full window without the required continuation.

For example, 160 people have had the full week and 60 finished. Another 40 entered recently: 10 finished and 30 are pending. Overall conversion is 70 / 200 = 35%. Finished windows only is 60 / 160 = 37.5%. Of the 130 people who have not finished, 100 ran out of time and 30 are pending. These are illustrative counts; follow the activation walkthrough to check them against the sample data.

These pending and expired counts belong to the overall funnel. A drop-off card for an individual breakdown value does not yet have this split: its missing continuations can include people who still have time. Read the overall funnel’s window labels before interpreting a breakdown’s entire gap as an expired drop.

Who dropped off

There are three ways in, from quickest to fullest:

• Click a ghost above a bar. A card says how many dropped off, the median time to continue, and, in a breakdown, whether the gap to everyone else is significant. Its actions include Show paths between these steps, What did they do instead?, and lists of who continued and who didn’t.

  

• Click a step bar for its menu: see who finished or didn’t, their events, and what they did next.
• Click a drop percentage, or a row’s dropped count in the table, to open the drop-off drawer.

The drawer is titled by the drop, for example Dropped before step 3. It has three tabs:

• What they did instead: the last thing each person did, including Nothing — session ended.
• Where they were: the top pages and events they reached afterwards.
• People: the people in the selected cohort, most recent first. When the drop includes pending entrants, switch between Window expired and Still has time (or Window runs past these dates for a historical range). Click a person to inspect their event timeline.

Steps, ghosts and the drawer use friendly event names, such as Viewed page rather than $pageview, the same names the pickers show.

Export downloads the people as a CSV. View people jumps to the People tab. The drawer follows your date range, source, segment, step filters and counting settings. Where the pending/expired split is available, it starts with people whose window expired. If you count a company type, its label replaces People.

Breakdowns and segments

• Break down by splits the funnel by an event property, such as Country or Browser. The chart draws the top 3 values that have at least 500 people entering. Smaller ones are listed in the table only.
• Compare segments splits it by up to five saved segments, plus Other. The chart draws the first three with at least 500 people entering.
• A breakdown and a segment comparison replace each other. The headline cards then show a combined baseline and one card per group, with its difference in points. The table adds a Significance column: each group’s conversion tested against everyone else (or against the other group, when there are two).

Filters

The filter bar sets people (segment), dates, source and Compare. Add filter on a funnel filters one step: the builder asks which step has to match, such as Match on step 2 · Viewed pricing. Those step filters belong to the page you are looking at and aren’t saved into the funnel. To keep them, save them in the builder with + Filter.

With Compare on, Overall conversion shows the previous period next to it.

Saving and managing

• Funnels are shared with your whole team. The list shows each funnel’s step bars and conversion for the selected range, and it has its own compact filter row.
• Search matches funnel names and step events. Sort by Recently updated or Name A–Z.
• A card’s … menu has Open, Rename and Delete. Delete removes it for everyone, with Undo in the toast.
• On a funnel, Edit reopens the builder.
• On a saved funnel, Add to board offers Linked (follows the saved funnel everywhere) or Copy (a snapshot that lives only on that board). Only a copy keeps step filters from the page.

Tips and gotchas

• Across two domains? If your site and app are on different domains, check Identity first. If identity isn’t carried, people look like two strangers and the funnel under-counts. A funnel that spans two sources shows how well identity is carrying at the top.
• The preview uses its own period: the last 30 days, or 90 after widening. The saved report uses the filter bar’s dates, so its counts can differ.
• Want a funnel from real paths? In Flows, pick a path and click Open as funnel.
• Drop-off audiences can’t be saved as segments yet. Use View people or Export instead.
• Changing the range or a filter keeps the last result on screen, dimmed, until the new one arrives.

Recipes