A call-to-action that nobody clicks has failed in exactly one of three ways. It was unseen, it was unclear, or it was unearned. These look identical in your analytics — a low click rate — and they have almost nothing to do with each other. Changing the button colour when the problem is "unearned" is a way of spending a week to learn nothing.
Work out which one you have first.
Unseen: the button never entered their attention
The visitor's eye moved down the page and the button wasn't where the eye was.
This happens when a button sits below the fold on the screen size most of your traffic uses, when it's styled like everything else around it, when it's stacked among four other buttons of equal weight, or when it appears before the visitor has any reason to look for a next step.
The tell: scroll depth is short, or people who do scroll past it convert at a normal rate. The button works fine — it's just not being met.
Fixes are physical, not persuasive. Put it where the eye lands after the value is delivered. Give it the strongest visual weight on the screen and take that weight away from everything else. Repeat it after long sections rather than trusting one placement. Check it on the phone, at the real screen height, not a resized browser window.
Unclear: they saw it and didn't know what it does
The visitor read the label and couldn't predict what happens next. Hesitation, in the absence of information, resolves as "not now."
"Get started" is the canonical offender: started doing what, and at what cost? "Learn more" is worse — it promises reading, which is work, not value. "Submit" describes what your database does, not what the visitor receives.
The strongest labels name the outcome the visitor gets, and imply the cost. "Get my free Friction Score" tells you what arrives. "Start my free trial" names the thing and its price. "Book a 20-minute call" tells you exactly how much of your life this consumes, which is why it outperforms "Contact us" so reliably.
Two smaller sources of unclarity: the button that goes somewhere different from what its label implies, and the microcopy gap — the unanswered question sitting next to the button ("Do I need a credit card? Can I cancel? What happens after I click?"). A single line under the button, answering the objection you know they have, is often the whole fix.
Unearned: they understood, and they weren't ready
Nothing is wrong with the button. The page hasn't yet given them a reason.
This is the diagnosis nobody wants, because it means the fix isn't the CTA. They don't believe the claim, they haven't seen proof, their obvious objection went unaddressed, or the ask is disproportionate to what they've received so far — you're requesting a demo booking from someone who has read one paragraph.
The tell: they scroll the whole page, they may even hover the button, and they leave. Attention was there. Conviction wasn't.
Fixes live upstream. Put the proof next to the ask, not in the footer — a visitor deciding at the button shouldn't have to go looking for a reason to trust you. Answer the biggest objection immediately before the button. And consider whether the ask is sized to the moment: a lower-commitment action ("see a sample report") sometimes converts more people into the higher one than the higher one ever would directly.
The competing-CTA trap, which is its own disease
One specific pattern deserves separate mention because it's the most common and most fixable friction on the web, and it doesn't fit neatly into the three categories.
Two buttons of equal visual weight, pointing to different destinations, with no explanation of the difference. "Talk to an expert" beside "Start now." "Book a demo" beside "Try free."
Each button may be well-designed. Together they force the visitor to make a decision they didn't come to make, on information they don't have. The most common resolution to that decision is to make neither.
The fix takes minutes: make one primary, make the other visibly secondary, and add one line explaining who each is for. If you can't say who each is for, you've found a strategy problem wearing a design problem's clothes.
Telling the three apart
Look at behaviour, not the click rate:
- Short scroll depth, low clicks → unseen.
- They reach the button, don't click, don't leave immediately → unclear (they're reading, deciding, looking for the answer to a question).
- They reach it, linger, and leave → unearned.
- They click one of two buttons at roughly equal rates and complete neither → competing CTAs.
Defrixa's free scan reads the structural version of this: whether one action is unmistakably primary, whether proof sits near the ask, whether the label names an outcome. It names the single biggest issue instead of listing all four.
Common questions
Contrast matters; the specific colour rarely does. A button needs to be the most visually dominant element in its region. If two elements compete for that role, neither wins it.
Yes, after each section that delivers a reason to act. Repetition costs nothing and meets people at the moment their conviction crosses the line.
Long enough to name the outcome and imply the cost. "Get my free Friction Score" is better than "Get score" and better than "Get started" — length isn't the variable, specificity is.
On any given page, yes. Two co-equal primaries is the same as none. Choose the action that serves the majority of that page's visitors and demote the other — it can still be there, just quieter.