CVR is computed as orders / sessions (session-based) or orders / unique visitors (visitor-based), with numerator and denominator drawn from the same window. The denominator choice changes the number: a brand with a high return-visit pattern reports a lower visitor-based CVR than session-based, because the same visitor contributes multiple sessions. Analytics tools differ on which denominator they report by default, so identify which one a tool used before comparing CVR figures across dashboards.
CVR is the workhorse traffic-to-revenue ratio, distinct from ROAS (media efficiency) and AOV (order size). It sits between paid acquisition and revenue in the funnel arithmetic — sessions × CVR × AOV — and is where landing-page, checkout, and merchandising work shows up in the number. CVR varies meaningfully by category (considered-purchase categories tend to convert below impulse categories), and reported sitewide bands shift year-to-year and by data source, so a brand’s own baseline is the reference, not an external benchmark.
Two operator traps. First, sitewide CVR hides composition. It varies by traffic source (paid social typically converts below branded search), by device (mobile often lags desktop on considered-purchase categories), and by new vs. returning mix — the blended number obscures whichever segment is actually moving. Second, a CVR jump is not automatically good news. It can mean the funnel improved, or it can mean the traffic mix tilted toward higher-intent visitors — paid pulled back while branded search held, say — inflating the ratio with no storefront change. Segment by channel before celebrating.