Pricing experiment
Experiment: pricing test
Pricing tests are the highest-stakes indie SaaS experiment. The variants are public and visible; the test changes real customer relationships. The framework below names the only honest way to test pricing on indie SaaS scale: grandfathered for existing customers, fresh prospects on the variant, and an exit criterion most founders skip.
Min sample size: 200+ paying customers per variant for a 25%+ revenue-per-visitor lift detection. Below this, the math is too noisy. Indie SaaS below 500 customers should rarely run pricing tests; qualitative + price-anchor research is more honest.
Duration: 30-90 days minimum. Shorter tests miss the price-anchor settling effect; longer tests start to confound with cohort drift.
Verified · editorial policy
Hypothesis structure
Changing the price from $[CURRENT] to $[VARIANT] will [INCREASE / DECREASE / MAINTAIN] revenue per visitor by at least [EXPECTED LIFT] because [SPECIFIC REASON].
If you cannot complete this template, you do not have an experiment — you have a guess.
Variant design
Existing customers grandfathered at current price (always). New visitors split into control ($CURRENT) and variant ($VARIANT). No mid-test pricing changes. No discounting either side to 'help' the variant.
Primary metric
Revenue per visitor, not conversion rate. A lower conversion rate at a higher price can still produce more revenue per visitor; the only honest metric is dollars per visitor.
Secondary metrics (watch but do not decide on)
- Conversion-to-paid rate (for diagnostic, not decision).
- Average order value (especially if testing tiered pricing).
- Refund rate per cohort (price-sensitive cohorts churn faster).
Procedure
Step 1
Lock in the grandfathering rule
Existing customers stay at current price forever (or for X months — be specific). Document this publicly so the test does not feel like a bait-and-switch.
Step 2
Set up the split at the checkout step, not the marketing page
Marketing-page tests measure conversion through several steps; checkout-step tests isolate the price effect. Use a server-side split keyed to a stable user identifier.
Step 3
Track revenue per visitor, by variant
Sum total revenue from variant cohort over the test window, divided by visitor count to that variant. This is the only decision metric.
Step 4
Run for full week-cycles, 30+ days minimum
Pricing tests have longer settling times than headline tests. Day-of-week effects matter; first-week novelty matters; second-week settling matters.
Step 5
Decide at 30 days; recheck at 60 and 90
Price-sensitive cohorts can churn in months 2-3. The 30-day decision is provisional; the 90-day decision is final.
Step 6
Roll out or roll back
Winning variant becomes the new price for new customers. Existing customers stay grandfathered indefinitely or until the next pricing event. Losing variant is rolled back; document the learning.
Self-deceptions to avoid
- Testing pricing without grandfathering. Burns trust with existing customers; even if the test 'wins', the brand cost outweighs the revenue.
- Reading 14-day pricing results as final. Pricing tests need 30+ days; pricing churn shows up in months 2-3.
- Confusing conversion rate with revenue. A pricing test that produces 50% the conversion at 3x the price wins on revenue per visitor.
- Testing pricing on warm-traffic referrals. Existing customer referrals have anchoring effects; test on cold traffic only.
What success looks like
Variant produces 20%+ higher revenue per visitor over 30+ days, sustained at 60 and 90 days, with refund rate within 1pp of the control.
Related benchmark
See the directional range for annual vs monthly discount to calibrate the expected lift in your hypothesis.
Frequently asked
- Should I tell prospects the test is happening?
- No, but document the pricing-change policy publicly so any prospect who reads it knows the rule. 'Existing customers grandfathered when prices change' is the policy that legitimizes testing.
- What if my pricing test reveals existing customers are underpaying?
- Keep them grandfathered. The lifetime-value of grandfathered customers usually exceeds the short-term revenue from a price-bump churn cycle.
Other experiments
Test on a page that is already pointed in the right direction
A/B tests on a misaligned page produce two losing variants. The diagnostic labels the alignment problem first; the test optimizes within the right alignment.