Choose the denominator that answers the question
Use the denominator that matches the question: purchasing sessions per session, requests per session or qualified requests per valid request. Show the counts and cohort dates beside each rate. The four calculations below are arithmetically related but should not share one ambiguous conversion label or benchmark interpretation.
Worked example — illustrative data.
| Metric | Illustrative count | Calculation |
|---|---|---|
| Session purchase rate | 20 purchasing sessions / 1,000 sessions | 2% of sessions purchased |
| Requests per session | 30 valid submissions / 1,000 sessions | 0.03 requests/session; 3 per 100 sessions |
| Request qualification rate | 6 qualified requests / 30 valid requests | 20% of requests qualified |
| Qualified requests per session | 6 / 1,000 sessions | 0.006 qualified requests/session; 0.6 per 100 sessions |
Write how repeat purchases, repeat submissions, internal traffic and invalid requests are handled. Select a period with consistent event definitions and comparable traffic conditions. For SaaS, allow the chosen cohort enough time to be reviewed before comparing qualification. Annotate outages or implementation changes instead of silently blending them into the average. Keep a copy of the definition beside the report so a later dashboard edit does not rewrite the baseline.
Annotate the period before comparing its rate
Place the denominator definition, date range and exclusion rules beside the baseline. If a release changed duplicate-event handling halfway through the period, split the period or rebuild a comparable series. In the demo example, record requests still awaiting qualification separately from those reviewed. Keep both the raw counts and the calculation so another reviewer can reconstruct the percentage.
Decision: Use the site’s own consistently defined baseline to evaluate a relevant change. The illustrative percentages are arithmetic examples, not benchmarks or expected results for a store or SaaS business.
Two orders in one session change the metric: A visitor can buy twice in one session. Orders per session and purchasing sessions divided by sessions are different calculations. Name the chosen metric rather than calling both conversion rate.
Can an industry average replace my baseline?
Use an external benchmark only as a separately labeled context source with its population and method. Your baseline needs your own consistently defined events, traffic and period. Another site’s rate cannot reveal whether your tracking or buying path changed.
Continue with Guides for stores, software teams and local businesses, Conversion rate optimization services and Trace a demo request from landing page to qualified opportunity.
For ecommerce event definitions, see Google Analytics ecommerce event documentation (checked November 2026).