Checkout funnel and its series

Facts about the result

The request as the server resolved it

Funnel steps in funnel order

Distinct visitors in the window, funnel or not - the denominator of conversion_rate. Counted over the same identity as the steps, so both halves of the rate describe one population. A visitor with a checkout event weighs one, every other visitor 1000 over the loosest sampling permille of their pageviews, so on a sampled application this is an estimate. Filters that select funnel PAGES (checkout_stage, the cart interaction metric ranges) do not narrow it - they would shrink the denominator towards the numerator - while audience filters (site, country, device, campaign) do. Where checkout events are read this holds the visitors of both tables, so a visitor seen only in the checkout counts here too and conversion_rate stays at or below 1. ABSENT when this half of the query failed (per response_model_exclude_none): the steps are still reported, the rate is not.

Visitors who reached the LAST step, over visitors_total. The classic e-commerce conversion rate: purchases over all traffic, the denominator GA4 ('session key event rate') and Shopify ('sessions converted') use, so a shop can hold it against the usual 1-3% benchmark. Never exceeds 1.0; absent when visitors_total is. NOT a step rate: how the funnel narrows from one step to the next is FunnelStep.step_conversion_rate, and those stay unclamped.

Seconds from funnel entry to the first success, over the visitors who reached it. Both ends are observed pageviews, so this is a measurement and not an estimate. count is the sample, so it equals the last step's reached. A p90 pulling away from the p50 is the checkout getting slow or getting in the way.

Visitors with at least one cart interaction (fetch/XHR cart mutation) over the window. A signal BESIDE the funnel steps: it only says an interaction happened, undercounts by design (fetch_xhr coverage), and is therefore not a conversion baseline.

cart_interactions over visitors_total - the add-to-cart rate, against a typical shop benchmark of 5-6%. Never exceeds 1.0; absent when visitors_total is. The underlying COUNT undercounts wherever the shop's cart calls are invisible to fetch/XHR collection (WooCommerce classic without the wc-ajax opt-in, OXID, anything posting the document) and never overcounts, so read the rate as a floor.

Visitors whose cart calls FAILED at least once over the window (transport error or an HTTP error status, including status 0). Read against cart_interactions: the share of shoppers who hit a broken cart. It moves before the conversion rate does, which is what makes it worth watching.

Kind -> visitors with >=1 interaction of that kind, where the shop system's endpoint reveals the kind: add | change | remove | other. 'change' includes Shopify's qty-0 removes; 'other' = the kind only lives in the POST body (never read). Only kinds that occurred appear.

Visitors who submitted payment in the checkout, from a Shopify Web Pixel event. A FACT BESIDE THE STEPS, not a fourth step: it is reported for the same run as the steps, and a visitor can carry it and a success at once. It can exceed the checkout step, because a visitor whose begin_checkout was lost still counts here. On Shopify a visitor with this and without a success most likely paid through a wallet (Shop Pay, Apple Pay) whose completion Shopify does not report, so the success step is a lower bound and the gap between the two numbers is where it is lost. ABSENT (per response_model_exclude_none) when no checkout event belongs to a run that ENTERED the funnel inside the requested window - an event of a buyer who entered before it counts for neither this field nor the steps. Also absent when none are collected (not a Shopify application, the events table is not installed, the identity is session, or a filter the event rows cannot honor such as URL, page type, campaign or a metric threshold), or when none occurred. 0 means events were read and nobody submitted payment.

One point per bucket that had a funnel entry OR any traffic. Step counts are bucketed by funnel entry time, visitors by the pageview itself, so a bucket with traffic and no funnel activity appears with zero steps and a real denominator.