SaaS pricing page A/B
A step-by-step recipe for running high-signal pricing page experiments that move ARPU without tanking conversion.
Hypothesis lock
Define the single variable you are testing — price point, anchor tier, or discount structure — before touching any code.
Traffic split
Route 50% of new visitors to variant B via edge middleware. Keep existing customers on control to avoid noise.
Guardrail metrics
Track trial starts, checkout abandonment, and support ticket volume — not just revenue per visitor.
Implementation checklist
- ✓Deploy a
middleware.tscookie-based split with deterministic hashing. - ✓Instrument checkout funnel with custom events — do not rely on pageview-only analytics.
- ✓Run for minimum two full billing cycles before calling a winner.
Common pitfalls
Changing too many variables. If you alter copy, layout, and price simultaneously, you will not know which lever moved the needle.
Peeking early. Statistical significance requires sample size. Calling a test at day three guarantees false positives.
Ignoring LTV. A lower price that doubles conversion but halves retention is a net loss. Measure cohort retention, not just week-one revenue.