CTAs experiment
Experiment: CTA copy test
CTA copy tests are the lowest-stakes, highest-signal-per-effort indie SaaS experiment. Small change, isolated effect, and the result usually transfers across pages. The framework below names what to test, what NOT to test in the same experiment, and the patterns most likely to produce lift.
Min sample size: 500-1,000 visitors per variant for a 25%+ lift detection on click-rate.
Duration: 7-14 days minimum at adequate traffic.
Verified · editorial policy
Hypothesis structure
Changing the CTA button copy from [CURRENT] to [VARIANT] will increase click-through-to-next-step 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
Only the button copy changes. Same button position, same color, same surrounding context. Test ONE phrasing change per experiment.
Primary metric
Click-through rate (clicks / visitors who saw the button).
Secondary metrics (watch but do not decide on)
- Conversion-to-next-step (did the click eventually convert?).
- Time-to-click (proxy for buyer hesitation).
Procedure
Step 1
List the candidate variants
Four to test against control: (1) outcome verb (Send your first email), (2) free path (Try a free 50-form), (3) outcome + time (Get bookings in 5 min), (4) value (See the 60-day refund). Pick one to test first.
Step 2
Test ONE variant at a time
Multivariate CTA tests at indie SaaS scale produce noise. Sequential single-variant tests produce signal.
Step 3
Run for a full week-cycle
Same day-of-week boundary at start and end. 7 days is the floor; 14 days is better for the lower-volume sites.
Step 4
Decide and roll the winner across the site
CTA-copy winners usually generalize. Apply the winning pattern to other CTAs on the site before running the next experiment.
Self-deceptions to avoid
- Testing 4 variants at once. Even at 1,000 visitors per variant, that is 4,000 total; you do not have it.
- Calling a 15% lift 'a result' on 200 visitors. The noise floor is higher than the lift.
- Confusing 'looks better' with 'tested better'. A/B testing exists exactly because intuition fails on CTAs.
What success looks like
Variant wins by 20%+ on click-rate over 500+ visitors per side, statistically significant.
Related benchmark
See the directional range for landing page conversion rate to calibrate the expected lift in your hypothesis.
Frequently asked
- Do CTA-copy wins transfer across pages?
- Usually yes, especially the outcome-verb pattern. Pattern wins more often than specific phrasings; once the pattern proves itself, apply it across the site.
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.