Analyze8 min read
Funnels
A funnel measures how many visitors moved through an ordered sequence of steps, and where the rest dropped off. Every number on this page comes from a single walk over each visitor's events within the window you set: there's no separate query per tab.
Building a funnel
A step is either a pageview (a path) or a named event, matched by an operator: is, contains, starts with, or matches (regex). A funnel needs at least two steps before it computes anything.
Conversion window
Every funnel has a conversion window: how long a visitor has, from their first qualifying event, to complete the rest of the steps. It defaults to 14 days and can be set in minutes, hours, days, or weeks. The window opens at the visitor's *first* matching event, not their best-fitting one, so entrants is a stable count of distinct people rather than of attempts.
Order modes
| Mode | What it means |
|---|---|
| In order | Steps must happen in sequence. Unrelated events in between are fine. |
| Strictly in order | No *other funnel step* may occur between two consecutive steps. |
| Any order | Counts visitors who completed the steps in any sequence. Step N means "completed at least N of them." |
Some analytics tools define strict order as disqualifying a visitor who did *anything else* between two steps. Mrkr's funnel query only fetches events that match one of the funnel's own step or exclusion predicates: pulling every event a visitor ever produced, just to check for intervening noise, would multiply the read by the site's entire event volume, and since every pageview is mirrored into events, it would disqualify nearly every real session on any funnel that includes a pageview step. So here, strict order means no *other step in this funnel* may fire between two consecutive steps, which is the strictness that actually catches people jumping back or skipping ahead, without being useless in practice.
Exclusion steps
An exclusion is an event that, if it fires between two steps you specify, removes that visitor from the funnel entirely, not just halts their progress at the step they'd reached. Use it for things like "exclude anyone who filed a support ticket between signup and purchase."
Breakdowns
You can split the funnel by a session dimension (device, source, country, and so on) or by an event property. Each visitor is attributed to a single group based on the value observed on their *entry* event (first-touch attribution), so the grouped bars always sum to the funnel's totals. The largest 12 groups are shown individually; everything past that folds into an _other bucket rather than rendering hundreds of series.
Steps, Time to convert, and Trend
Three views on the same computation:
Steps- The funnel chart itself: visitors reaching each step, drop-off counts, and conversion rate both from the previous step and from the top.
Time to convert- A histogram of how long converters took, in fixed buckets from under 10 seconds to over 7 days, plus the median and 90th-percentile duration.
Trend- Entrants and converters per day, bucketed by the day a visitor *entered* the funnel (not the day they converted), so a cohort's conversion rate stays attached to the day it started.
Drop-off inspection
Clicking a step that lost visitors opens a sample of up to 100 people who stalled there, with enough session context (source, device, country, entry page) to spot a pattern, and a link into session replay when a recording is available. On a very large range, the underlying event read is capped; if that cap is hit, the funnel says so rather than presenting a truncated count as a final one.