The decisions, in order:

Numerator: what's a 'conversion'?
- First purchase? Any purchase? A returned purchase excluded?
- A canceled order still in progress?
- A subscription started (and immediately refunded)?
Denominator: what's a 'visit' or 'user'?
- Sessions? Users? Unique devices?
- Include bot traffic? Internal employees? Repeat visits from the same user in a day?
Window: when?
- Same-session conversion rate is very different from 7-day conversion rate. Which answers the business question?
Segmentation: new users vs. returning, free vs. paid traffic, desktop vs. mobile.

A defensible retail / e-commerce definition:

"Session-level conversion rate = (distinct sessions where a purchase was completed and not refunded within 14 days) / (distinct authenticated+guest sessions), excluding bots and internal traffic, measured per calendar week in UTC."

Why 'session-level' vs. 'user-level'?

Session-level is the industry standard for retail — measures shopper intent per visit.
User-level (purchases per unique user per month) is better for measuring acquisition funnel performance.
Both have their place. Be explicit about which.

The common mistakes:

Counting refunds as purchases. Inflates conversion rate, hides UI problems (e.g., 'users finish checkout but refund immediately because the UI confused them into buying the wrong thing').
Including bots in the denominator. Bots rarely purchase; denominator is inflated; conversion rate looks low.
Counting multi-item orders as 'one conversion.' Fine as long as you're consistent — same number from numerator.
Wrong grain for the question. Measuring conversion at the click level for a subscription business makes no sense; the decision-useful grain is user or trial.

Stakeholder framing: "Before I pull the number, let me confirm: session-level or user-level? And do we count refunds? Those two choices alone change the answer by 15-30%."

What's your conversion rate definition, exactly?

Draft your answer

Stuck? Peek at a hint

More product metrics questions