Writing a call to action people act on

Button copy gets treated as a finishing touch. It is usually the last thing anyone writes, chosen from a short list of things other websites say, and then never looked at again.

It deserves more attention than the headline. The headline decides whether someone keeps reading. The button decides whether the visit was worth anything.

What makes a call to action work?

Three ingredients. Skip any one and the button gets weaker.

The verb the person will actually perform. Call, text, book, email, download. Not "start," not "explore," not "learn more," which describe nothing anybody does.

The thing they get. A quote, an appointment, an inspection, a callback, a menu. This is the part that answers "why would I."

The cost, when it is low. Free, no card needed, takes two minutes, no obligation. If the commitment is small, say so, because the visitor is assuming it is large.

Put together, you get "Book a free onsite estimate" or "Text us a photo for a rough quote." Both are longer than "Contact Us" and both work better, because length is not what makes a button hard to click. Uncertainty is.

A button is a promise. Write the promise, not the invitation.

Why does "Get Started" fail?

Because it hides the commitment, and visitors know it.

Started on what. A form, a phone call, a sales sequence, an account signup. The visitor has been burned before by a friendly button that turned into a fifteen minute process, so they hesitate. That hesitation is the whole loss.

The same problem runs through most of the standard set. "Learn More" promises reading rather than progress. "Submit" describes what the software does, not what the person gets. "Click Here" says nothing at all and is unusable for anyone on a screen reader.

There is a subtler failure too. When every button on the page says the same thing, the page stops guiding anyone. A visitor at the top of a service page and a visitor at the bottom are in different states of mind, and giving both the same generic prompt wastes the difference.

Write the button last, after you know what the section above it just argued. If the section explained your emergency response, the button says "Call now for emergency service." If the section explained pricing, it says "Get a written quote today."

How many calls to action should a page have?

One action, repeated, rather than several competing ones.

The rule people misremember is that pages should have one call to action. What they should have is one destination, offered several times. On a long service page, the visitor decides at unpredictable moments, and asking once at the bottom means missing everyone who was ready at paragraph three. Repetition of the same ask is not nagging, it is availability.

Where it goes wrong is offering three different actions side by side. Call us, fill this form, book online, chat with us, download the guide, all in one block. That is not helpfulness. That is a decision you handed to someone who came here to avoid making decisions.

Pick a primary action and make it the loud one. Offer a secondary in plain text underneath for people who will not do the first. Phone as primary and a form as the quiet alternative works well for trades. Booking as primary and phone as the alternative works well for clinics.

If you are considering adding a chat widget as a third option, check first whether it competes with the action you want. There is a real question about whether chat helps a small business or just adds another thing to answer.

Should the button be a different color?

It should be obviously a button and obviously the loudest thing in view. The specific color matters far less than the category pretends, and there is a good source for saying so.

You will eventually be shown the "red beats green" case study. It comes from a single published test by Joshua Porter, and it has been repeated in conversion decks for over a decade. Ron Kohavi, Alex Deng, Roger Longbotham and Ya Xu name it directly in their 2014 KDD paper, written out of Microsoft and LinkedIn, as their worked example of a result that does not generalise. Their words: since they do not see a lot of sites with red call to action buttons, they believe this is not a general result that replicates well.

That is a rule with a name in the paper, and the rule is that your mileage will vary. The same authors are equally blunt about the genre those case studies come from. On published A/B results they ask: was it peer reviewed, was it properly run, were there outliers, and was the p value low, noting that they have seen tests published where the p value was above 0.05, which is normally considered not statistically significant.

A color result from someone else's site is a fact about their site. What actually matters on yours is contrast against everything around it, size large enough for a thumb, and enough space around it that it does not read as part of a block of text. A button that fails any of those will underperform in every color.

Two things break buttons more often than color. They do not look tappable, so people scroll past them, which shows up clearly in session recordings as visitors moving past the exact element you built for them. And they load late, so the visitor sees a page with no visible action for a moment and starts to scroll. That second one is the honest part of the speed debate, where page speed and conversion are worth separating from the myth.

Can I just test my button copy and find out?

Almost certainly not, and you can settle that in four lines of arithmetic rather than by argument.

The standard sample size formula for comparing two conversion rates, at the conventional 5% significance level and 80% power, is 16 times your conversion rate times one minus your conversion rate, divided by the square of the improvement you want to be able to detect. The 16 is two times the square of 1.96 plus 0.84, the z values for those two conventions, which comes to 15.68.

Say you convert at 3% and you would be delighted with a fifth more inquiries. That target rate is 3.6%derived, so the difference you are chasing is 0.6 percentage points. The formula gives roughly 12,900 visitors in each version, so about 25,800 visits in total. At 500 sessions a month you are running one button test for four years.

The threshold is not a traffic number, it is a function of your rate and the size of the change you want to see. That is why nobody can hand you a single visitor count without asking what you are trying to detect, and why the round figures in circulation should be checked rather than repeated.

The same Microsoft paper puts realistic effect sizes in perspective. At Bing, most experiments fail, and those that succeed move key metrics by 0.1% to 1.0% once diluted to overall impact. If a search engine running thousands of experiments a year is finding effects that small, a promise that new button copy will transform your inquiry volume deserves a raised eyebrow.

So write the button on judgment, not on evidence you cannot generate. That is not a compromise. It is what everyone below real traffic is actually doing, whether or not they admit it.

What is a better button worth to me?

Here is the arithmetic, and the point of it is that you can price a ten minute edit.

Start with your baseline. Say 500 sessions a month and 15 inquiries, a rate of 3%.

State the assumption out loud. Suppose rewriting the main button from "Contact Us" to "Book a free estimate, no obligation" moves you a fifth. That is an assumption, not a finding, and nobody has published a lift figure for local service button copy. Say so when you write it down.

Convert it to leads. A fifth more than 15 is 18 inquiries, so three extra a month.

Price the leads. LocaliQ and WordStream's 2026 search advertising benchmarks put the median cost per lead in Home and Home Improvement at $90.92 and in Restaurants and Food at $30.57. Those are campaign level national medians for paid search, not page benchmarks, and LocaliQ sells advertising services. Use your own cost per lead if you have it. Three extra inquiries a month at the Home and Home Improvement figure is about $273 of equivalent lead value, every month, from an edit that took ten minutes and cost nothing.

Then accept that you cannot prove it. Four years, per the arithmetic above. What you can do is measure direction over quarters and stop treating a two week comparison as evidence.

That is the whole honest case for button copy. It is cheap, the reasoning is sound, the downside is zero, and the proof is unavailable. Anyone who offers you the proof at this traffic level is offering you something else.

What to do this week

Write down every button and link on your main service page and what it says. Most owners are surprised by how many say the same three words.

Rewrite the main one using the formula: verb, thing they get, cost if it is low. Ten minutes of work.

Then repeat that button after each major section of the page, using the language of the section it follows.

Run the four line cost calculation above with your own numbers, so the edit has a dollar figure attached to it rather than an opinion.

Look at whether the change moves anything over a quarter, using inquiries rather than clicks, since a healthy traffic number sitting next to a flat inquiry number is the exact situation button copy is meant to fix.

And if someone has told you the whole site needs rebuilding because it does not convert, try the copy first. A redesign is often the wrong answer to a problem that lives in twelve words.

Questions about your button copy?

Email me at eric@seod.com with the exact text on your main call to action button and a link to the page it sits on. I will send back three rewritten versions and tell you which one I would run and why. No form, no call, and it takes me a few minutes.

If your button already says something specific, I will tell you it is fine rather than inventing an improvement, because rewriting a working button is a way to feel busy.

There is more on conversion here.