- Step conversion is next-step entrants divided by prior-step entrants.
- The largest percentage loss and the largest absolute loss may occur at different steps.
- Fix instrumentation and compare eligible, mature cohorts before attributing a change to design.
Define the funnel before calculating a percentage
A conversion funnel is an ordered set of steps toward a defined outcome. Funnel drop-off analysis finds the transitions where observed entrants stop progressing. For a SaaS signup flow, choose product page → pricing → signup → confirmed account rather than every possible interaction.
Choose the measurement unit: visits, consented browsers, or connected accounts. Decide whether completion must occur within one visit or over several days. Also decide whether someone may enter at a later step. The same raw events can produce different counts under these definitions.
Use accepted events and a documented outcome. A rendered success message is weaker evidence than an account your server confirms was created. Do not quietly mix a button click with a paid or activated customer.
Use the step conversion and drop-off formulas
For one transition, step conversion equals the number who reach the next step divided by the number who reached the preceding step. Drop-off count equals the preceding count minus the next count. Drop-off rate equals one minus the step conversion rate. Multiply a rate by 100 to express it as a percentage.
Overall conversion divides the final-step count by the initial-step count. Do not average the step conversion percentages: multiplying the compatible step rates gives the overall rate. If any denominator is zero, the corresponding rate is undefined, not a meaningful zero-percent performance result.
Step conversion = next step / previous step
Drop-off count = previous step - next step
Drop-off rate = 1 - step conversion
Overall conversion = final step / first stepA worked SaaS signup example
These numbers are illustrative, not VisitLift customer results. Assume the same 1,000 observed entrants are followed through a closed, ordered funnel within the selected window. Four hundred reach pricing, 120 reach signup, and 30 create an account.
The signup-to-account transition loses 75% of its entrants, the largest percentage loss. The opening transition loses 600 entrants, the largest absolute loss. Those facts suggest different questions. Before changing either page, check traffic intent and whether each event is being captured consistently.
| Transition | Entrants → next step | Step conversion | Drop-off |
|---|---|---|---|
| Product → pricing | 1,000 → 400 | 40% | 600 (60%) |
| Pricing → signup | 400 → 120 | 30% | 280 (70%) |
| Signup → account | 120 → 30 | 25% | 90 (75%) |
Prioritize by potential impact and effort
A large drop-off percentage is a prioritization clue, not a diagnosis. Check the underlying signup interaction before changing its copy, and review the source mix before trying to send more visitors to pricing. A broken form and an uninterested audience need different fixes.
Model a concrete scenario. Holding all other rates fixed, improving signup-to-account from 25% to 30% yields 120 × 0.30 = 36 accounts instead of 30. Improving product-to-pricing from 40% to 45%, with the later rates unchanged, yields 1,000 × 0.45 × 0.30 × 0.25 = 33.75 expected accounts.
The fractional result is a model expectation, not a possible individual outcome or a forecast. Real changes may affect traffic quality and later-step behavior. Compare likely effort and user evidence before deciding which scenario to test.
Check the denominator and observation window
A visitor who entered an hour ago has not had the same chance to finish a seven-day journey as one who entered last week. Separate pending entrants from those whose window has closed. A recent incomplete path is not automatically an abandonment.
In VisitLift, within-visit funnels and purchase journeys answer different questions. Purchase journeys use an explicit seven- or 30-day window and end in verified paid Checkout. They depend on consented continuity or signed account identity; daily anonymous estimates are not a substitute for that population.
Compare the same timezone, source mix, currency, and Test/Live mode. An instrumentation deployment, changed consent coverage, or missing payment metadata can alter the denominator without changing the customer experience. Investigate those changes before crediting a redesign.
Turn the analysis into a repeatable audit
For each suspected leak, record the exact event definition, eligible audience, time window, count, observed friction, and one proposed change. Add the owner and review date. Confirm the path on an actual device and inspect journeys around the transition.
The free CSV worksheet includes measurement checks, source intent, signup verification, and a change log. It contains no benchmark or implied industry conversion target. Fill it with your own accepted-event counts, then compare a later cohort using the same definitions.
Use the funnel to decide where to investigate and journeys to understand what happened around that step. Stripe is only needed if your selected outcome is a verified customer payment; page and event funnels can answer useful questions earlier.