A SiteLine funnel is an ordered list of the same rules you use for goals: exact path, path contains, or event name. It does not fire new events. It re-reads pageviews and custom events you already send. Defining a funnel does not touch quota.
What you get back is not a marketing cartoon of "the customer journey." For the selected date range, SiteLine counts distinct cookieless visitors who matched step 1, then the subset who also matched step 2, and so on. Each row shows that visitor count and the percentage of starting visitors (step 1), not the percentage of the previous step.
Read that twice. 40% on step 3 means 40% of people who hit step 1 also hit step 3 in the range — not 40% of step 2.
What a step actually matches
Same types as goals:
- Exact path — pageview path equals a value (
/thanks). Query strings are not part ofpath./thanks?plan=growthstill matches/thanks. - Path contains — pageview path includes a substring (
/checkout/). Easy to over-match./blog/checkout-tipscontains/checkoutif you get sloppy with the substring. - Event name — a custom event whose
nameequals the value (signup_completed). Pageviews are not custom events.pageviewis the wrong match type.
Paths are for pageviews. Event names are for track(). Mixing them in one funnel is allowed and usually what you want: /pricing → /signup → signup_completed.
Honest limit: identity is daily
Unique visitors are a hash of truncated IP + user agent + a daily salt. The hash does not survive midnight. A person who views /pricing on Monday and /thanks on Tuesday is two visitors. They will not complete a two-step funnel across that night.
Read multi-day funnel reports as "how many same-day progressions happened in this window," not "how many humans finished the sequence this month." If the conversion is supposed to happen on the same visit, this model is fine. If your sales cycle is a week, a SiteLine funnel will undercount completed sequences. That is the cookieless trade, not a chart bug.
Funnels also intersect visitors inside the selected range. They do not require step 2's timestamp to be after step 1's. Someone who hits /thanks then later /pricing in the same day still sits in both steps. Keep steps in the order people usually travel, but do not treat the analyzer as a session recording.
1. Send the underlying events
Install the script or React SDK. Confirm Live shows pageviews. If a step is an event name, fire that event with a stable name before you create the funnel. Goals and funnels cannot invent traffic.
2. Open Funnels
In the analytics header: Funnels. Date range lives in the same header as Overview. Visibility soft-caps apply here the same way they apply to pageview charts: after you hit the plan entitlement, newer data is hidden until you upgrade. Ingest keeps accepting events. See quotas.
3. Create the funnel
The form is named Create funnel.
- Name it something you will recognize in a month (
Signup, notTest). - Add at least two steps. Each step has a type and a value. Placeholders in the UI are
/pricingstyle paths. - Use Add step for more. Remove last step appears once you are above the minimum of two.
- Submit Add funnel.
No redeploy. If the matching pageviews or custom events are already flowing, counts fill in for the selected range.
A useful default:
- Exact path
/pricing - Exact path
/signup - Event name
signup_completed
Prefer a dedicated thank-you path or a named event for the last step. Homepage → "path contains /" is a funnel that measures nothing.
4. Read the rows
Each saved funnel lists its steps as position. match_type: match_value with a visitor count and a percentage of starting visitors.
- Step 1 at 100% is tautological. If step 1 is huge, you started too wide (
path contains /). - A cliff between
/pricingand/signupis a product problem, not an analytics one. - A cliff on an event-name step often means the event is not firing, the name does not match, or it fires on a hostname that is not allowlisted.
- Percentages can sit at 0.0% with a handful of visitors when the start volume is large. Trust the count first.
Delete exists on each funnel. There is no separate "edit steps" flow in the form you used to create it — delete and recreate if the sequence is wrong.
5. Change the range on purpose
Narrow to a day when you are debugging instrumentation. Widen when you want volume. Remember the daily hash: a 30-day range is thirty identities per returning human, and cross-day completions do not chain.
Compare with Goals on the same site. A single goal is one rule and a conversion rate against all visitors in the range. A funnel is a sequence. They answer different questions. Neither emits extra billable events.
FAQ
Do funnels cost extra events?
No. They are derived from pageviews and custom events you already send. Optional outbound-link and file-download tracking does emit custom events and therefore costs quota.
Why is step 3 lower than I remember from last week?
Date range, daily visitor rotation, the visibility cap, and instrumentation drift. Check Events for the custom name, Install for the allowlist, and whether you upgraded after hitting a plan cap. Hidden post-cap events do not appear in funnel counts until the entitlement covers them.
Can I funnel on UTMs or countries?
Not as step match types. Steps are path or event name. UTM, country, device, browser, and OS are separate breakdowns on the analytics site. Filter the date range, then read the funnel for that window.