Does social proof below the submit button lift lead form leads?

Our lead form gave visitors no proof near the button that others trusted the brand. We tested social proof below the button, alone and with a percentage progress bar. Over 37 days, the combined version moved conversion 0.3%, short of our confidence bar.

Not significant Responsive 37 days Not called

SURFACE

Lead form

DEVICE

Responsive

VERTICAL

Lead Gen (CPL)

PRIMARY METRIC

Conversion rate

VARIATIONS

2

The hypothesis

By adding social proof below the submit button, we will make visitors more confident in the brand, resulting in a higher conversion rate.

The button is where visitors decide whether to go on, so proof that others trusted the brand belongs close to it.

A percentage progress bar had trended up in the round before, so we paired it with social proof in a second version to see whether the 2 changes would add up.

What we changed

Social proof below the button, alone or with a percentage bar

Var A added social proof below the submit button. Var B added the same social proof and replaced the step-count progress bar with one based on percent complete. We paused Var A partway through the test.

VAR A: SOCIAL PROOF

Social proof below the submit button

VAR B: VAR A + % PROGRESS BAR

Social proof below the submit button
Progress bar shows percent complete

The result

Conversion moved 0.3% with social proof and a percentage bar

CONVERSION RATE

+0.3%

LEAD VALUE PER VISITOR

+14.1%

VAR A CONVERSION

-6.9%

Over 37 days, the combined version (Var B) converted 0.3% better than the original and lifted lead value per visitor 14.1%, but the result did not reach our confidence bar or hold up in our data check. Social proof alone (Var A) converted 6.9% worse, but it got much less traffic than the other versions after we paused it, so that number is less reliable.

What we would do next

Rerun the combined version with even traffic before acting

Lead value per visitor rose 14.1% with the combined version, but lead volume barely moved and the result did not hold up in our data check. A rerun with even traffic across versions would show whether that gain is real.