Fix the leak at checkout
A small change in the checkout flow moves revenue more than almost anything else on the site. Test it safely, with payment success and refunds watched as guardrails.

What to test first
Guest checkout
Remove the account wall and measure completed orders and later repeat purchase.
Payment methods
Add wallets or local methods for a region and watch authorisation rate, not just clicks.
Shipping cost reveal
Show the full cost earlier and see if drop off at the last step falls.
Form length
Cut fields, use address autocomplete, and measure errors per session.
Trust and reassurance
Test returns copy and badges near the pay button against a clean control.
Promo code field
Collapse it and measure whether people stop leaving to search for codes.
Revenue, not only conversion
A checkout variant that raises conversion by 3% and lowers average order value by 4% is a loss. Kerf computes revenue per visitor with an interval, and keeps refund rate and payment failures as guardrails.
- Revenue per visitor as primary metric
- Payment error rate as guardrail
- Replay clips of failed payments by variant
A checkout test, start to finish
- 01
Instrument
Make sure start, payment attempt and order events reach Kerf or your warehouse.
- 02
Plan
Use baseline conversion and traffic to size the test. Expect one to two weeks to cover pay cycles.
- 03
Run behind a flag
Ramp to 50% with guardrails on payment errors and latency.
- 04
Ship or iterate
Read the revenue per visitor lift and the segments, then decide.
