Removing a CTA’s wobble animation increased the measured next-page conversion rate from 50.66% to 54.91% in this seven-day experiment. The result supports the quieter button in this particular funnel. It does not establish that visitors trusted the business more, or that every animated CTA performs worse.
The question: did the animation help the next step?
The client already had an established funnel. The CTA used a repeated wobble intended to draw attention. Our hypothesis was that removing it might make the experience less distracting and improve progression. That explanation motivated the test; the experiment measured behavior, not distraction or trust directly.
| Element | Control | Treatment |
|---|---|---|
| Button design and placement | Existing design and position | Same design and position |
| Motion | Wobble animation enabled | Wobble animation removed |
| Measured outcome | Reaching the next page | Reaching the same next page |
The implementation targeted the button with a CSS selector and changed the animation class for the treatment. The important constraint was to preserve the label, destination, position and click behavior. A change to any of those would make it harder to interpret this as a motion comparison.
Results and what the metric means
The historical experiment ran for seven days with 1,974 participants. Its results display, transcribed during the original review, records 987 participants in each arm, with 500 next-page conversions in control and 542 in treatment. These are aggregate historical results, not a new experiment.
| Measure | Animated control | No-animation treatment |
|---|---|---|
| Participants | 987 | 987 |
| Next-page conversions | 500 | 542 |
| Conversion rate | 50.66% | 54.91% |
The difference is approximately 4.26 percentage points using the counts, or an 8.40% relative increase. Subtracting the rounded displayed rates gives 4.25 points; that small difference is rounding. The historical display reports a 97.09% win probability and a 95% credible interval for relative lift of approximately +2.24% to +14.48%. These are reported model outputs, not a newly reconstructed analysis with independently verified model settings.
The primary metric used a page view on the next page as an opt-in proxy. It therefore supports a conclusion about progression to that page. It does not by itself verify a completed backend registration, a qualified lead, a purchase or retained revenue. In a replication, check whether reloads, direct visits or an incomplete submission can trigger that event, and deduplicate conversions using the intended participant identity.
Use the A/B testing metrics guide to choose the business outcome before running a similar test. PostHog’s Bayesian statistics documentation explains the product’s probability-based interpretation; avoid replacing that with a generic claim that significance proves the hypothesis.
What this experiment did not establish
The no-animation experience performed better on the recorded proxy, but the test did not ask visitors about trust or measure why they progressed. Reduced distraction, a visual preference or another response to motion remain possible explanations. The earlier interpretation that “trust beats tactics” went beyond what this experiment measured.
The record also does not establish a universal seven-day stopping rule, long-term persistence or an effect across every device and traffic source. The duration and sample belong to this experiment. A new test should use its own baseline, decision threshold and follow-up window.
How to reproduce the comparison without changing the question
First identify the exact CSS class or animation declaration responsible for the wobble. Remove only that motion in the treatment, retaining the button’s other styling and behavior. Confirm the correct variant persists through reloads and that the animation does not briefly appear before assignment is applied. Test mobile layout, keyboard focus and form validation in both versions.
Next verify that exposure is recorded when the tested experience is actually available, and that the next-page event follows the intended action. Where possible, pair the proxy with a successful registration event and a downstream qualified action. A higher next-page rate is useful only if it represents the desired user progress.
The motion comparison described here is separate from choosing a control type. For that different problem, see the date-selector experiment, which varied selection format, waitlist availability and instructions.
A bounded decision
For this funnel, the measured result favors removing the wobble. The reusable lesson is to test whether attention-grabbing motion helps the action it is meant to support, rather than assume more attention means more useful conversions. Preserve accessible focus and interaction cues while testing decorative motion; they serve a different purpose.
