Funnel
Track how actors move through an ordered sequence of events, with a conversion window, so you can see where people drop off.
A Funnel widget shows step-by-step conversion for an actor across a set of ordered events. You list the steps you expect someone to go through, in order, and the funnel tells you how many actors reached each one. That makes it the right tool for questions like "of everyone who started signing up, how many finished?" or "how many people who added an item to their cart actually paid?".
Because a funnel follows one person or server through several events, it is built
on top of actors. Every event you want to count in a funnel
needs an _actor attached when you send it. Events with no actor are anonymous,
so they can never be tied together into a journey and are left out of the funnel.
Funnels rely on actors
Attach an _actor to every event in your funnel. If your signup and purchase
events do not carry an actor, the funnel has no way to know they came from the
same person, and no one will appear past the first step.
How a funnel works
A funnel is a list of 2 to 8 steps, in order. Each step is one event, with
optional filters to narrow it down (for example, only
purchase events where plan equals pro).
The order matters. An actor counts toward a step only if they fired that step's event after the previous step, and within the conversion window. So the funnel is not just "how many people did each of these things" - it is "how many people did them in this sequence, quickly enough".
The conversion window
The conversion window is the amount of time an actor has to move from one step to the next. You can set it anywhere from 1 hour up to 30 days (1 to 720 hours). If someone fires a step's event too late - after the window has passed since the previous step - they do not count toward that step, even though they technically did it.
Pick a window that matches the journey you are measuring. A checkout funnel might use a short window like 1 hour, because people who buy usually do it in one sitting. An onboarding funnel might use several days, because new users come back over time.
Settings
Prop
Type
Example: a signup funnel
Say you want to know where people give up while creating an account. You send four
events from your backend, each with the same _actor (the new user's id), and
build a funnel with these four steps in order:
signup_started
The user opened the signup form. This is the top of the funnel - everyone who gets counted here is a potential new account.
email_verified
The user clicked the link in their verification email. To count here, they must
have done this after signup_started and within the conversion window.
profile_completed
The user filled in their profile. Again, only actors who reached the previous step, and did this in time, are counted.
first_purchase
The user made their first purchase. This is the bottom of the funnel - the actors who made it all the way through.
With a conversion window of, say, 7 days, the funnel now shows how many actors
reached each step. If a lot of people hit signup_started and email_verified
but the count falls off a cliff at profile_completed, you know exactly where to
focus: the profile step is losing people.
The same idea works for a purchase funnel. Chain events like view_item,
add_to_cart, checkout_started, and purchase, attach the buyer's _actor
to each, and set a short window so you only count people who bought in one
sitting.