Buttons.

Today's idea is a short and sweet one for you. Most of us are coming off the back of the holiday weekend (hello my favorite season), and I figured a tactical, high-leverage move you can run with today is the right beat.

And what's simpler than a button?

Nay, not even the button. Just the text in the button.

Too easy, right? But I bet that your button could be better. And if I do a good job, you’ll go look at your buttons in a few minutes realizing I was right—there’s almost certainly a winning AB test here for you.

So let’s unpack the data, psychology, and anatomy of a button so you get rid of your “cost” buttons and starting having growth-forward “pay” buttons.

Some people walk into meetings and reference precisely what was decided last time. They recall the numbers. They know who promised what and when. You wonder how they do it.

They're not better than you. They just have a better system than a messy Google Doc.

Granola is an AI notepad with a powerful Chat feature that turns months of meeting notes into an instant, searchable memory.

Granola captures your meetings automatically. Over time, you build up a library of every discussion and decision in fine detail. When you need to refer to something, simply open Chat and ask. It pulls from your personal notes, shared team notes, and private notes others have sent you. Every answer also comes with citations giving you extra confidence. 

You become that person in the room who never loses the thread. Who has everything at their fingertips. Not because you take endless notes but better notes. Notes that work for you after the meeting ends.

🥷 Rewrite the labels on your buttons so it names the payoff of the click instead of the task.

Picture this—you’re standing in front of your grocery store about to walk in. Right above the automatic doors you see a sign that says "Enter the grocery store here." Fine, not a bad sign post, technically accurate. Imagine the same sign: "Get what you need for dinner tonight." Same door. Different pull.

Or one more…you're in NYC on 5th Avenue passing a nice watch shop. One sign says "Talk to a watch salesman." The other, "Find the timepiece you'll pass on to your kids."

Nobody would ever hang the first sign. Even though you do take that path to enter and surely will speak to someone trying to sell you an expensive watch, that’s not what you want from doing that thing.

Yet that's exactly what most of us, and also the biggest companies on the planet, are doing in our products.

"Sign up." "Get started." "Start trial." "Book a call." "Continue."

That’s what YOU want me to do. I didn’t wake up this morning wanting to sign up for something or have a demo on my calendar. Neither did you.

I’d say this type of copy is a bad habit that’s been perpetuated through a bad game of Simon Says. Someone big like Airbnb or Apple did it, someone else copied it, and now we all keep using this standard.

Copy-paste buttons from every other SaaS site shipped as an afterthought.

I’m no different. I’ve built those generic-button homepages and onboarding flows too. We don’t know they are bad because we most of the time, don’t care to ask.

Our conventional wisdom, at least implied by all the examples around us, is keep CTAs short, lead with a verb, tell people what to do. Do this thing. And that semantic isn't wrong exactly it's just answering the wrong question.

Because like our made up person about to enter the watch store, that user hovering over your button isn't asking "what work do you want me to do for you?" They're asking "what do I get that is worthwhile if I press this?"

A button is like a price tag. It tells you what the click costs will cost and what it pays. And almost all buttons on the internet are just telling you what the cost it.

"Sign up" is a cost even if you have “for free” after. "Book a call" is a cost (that’s a meeting! on my calendar!). "Start trial" is a cost with a countdown. “Get started” is a cost…that sounds like a lot of steps!

But something like "Get matched with your first job" is a payoff.

My first startup was a job board for models, actors, and creators.

I remember our signup button for a long time was "Join for free."

Then I copied Airbnb’s “List your property” and tested our "List your profile." Felt more specific. It was marginally better. A profile is a chore. Someone opening up our app on the subway doesn’t want to list a profile right now do they?

Then we tested "Get matched with your first job."

I honestly can’t remember the exact number (this was like 2017), but I remember that it wasn't even close. Way better because it was just button that agreed with the way things are: Nobody wants to join a job board. They want the job.

But alas, the lesson was not engrained in me. I was recently looking at my app funnel for Good Times, and the prior 2 week data for Splash → Register CVR

The big button on my splash was just like all the other apps I was playing with at the time: “Continue”. So I changed it to “Find out what today’s question is”

Small sample so it’s not science but control converting at 55%, the payoff version at 72%.

So what separates a bad button from the best button?

❌ Bad: "Book a Demo" (yay a meeting with a salesperson.)
🆗 Better: "Get Started" (cheaper click and this exact swap is documented at +111%, more below)
✅ Best: "See it on your own data" (the actual payoff of a demo)

❌ Bad: "Sign Up" (pure price & zero prize)
🆗 Better: "Start Free Trial" (free helps; mind the verb. "Try Now" has tested 26% worse)
✅ Best: "Get your personalized diet " (that’s what someone goes to Noom for)

❌ Bad: "Subscribe" (the button on my own site and every newsletter)
🆗 Better: "Subscribe for free" (price lowered but payoff still missing)
✅ Best: "Send me the best idea" (immediate payoff and value where in my welcome email I link the best 3 posts)

That’s the anatomy of button copy that converts: a cheap verb + the thing they actually want + in their own words + under 8 words.

Before we continue with the breakdown and proof, a quick word from Attio. As always, thank you for supporting our partner who keep this all free.

Some of the fastest-growing startups—Granola, Modal, Wispr Flow—have moved to Attio, the agentic CRM. So did I. Attio gives you

George Loewenstein's information-gap theory (1994 but still the most-cited work on curiosity) says we feel curious and compelled when our attention lands on a gap between what we know and what we want to know. A payoff-stated button opens that gap and offers to close. "Learn More" tells you a gap exists but does nothing to close it. Learn something new about information gap theory

And even when we know what to expect when we go to a meditation app and click “Get Started” because it’s not rocket science, the point is that the science says its subconscious.

Also there's a newer study built on real click data. Upworthy Research Archive, Across tens of thousands of actual headline A/B tests, it shows that when the baseline copy is vague, making it more concrete increases clickthrough. And when copy is already concrete, piling on more specificity has a negative impact. An inverted U.

→ Pronouns win. An Unbounce AB test changed "Start your free 30-day trial" to "Start my free 30-day trial." One word had a +90% clickthrough lift. First-person copy makes you own the payoff before you've clicked.

→ Urgency loses. In 2007, the Obama campaign tested their splash-page CTA. "Sign Up Now," the loud and urgent choice lost big time. "Learn More" plus a family photo won, lifting signups 8.26% → 11.6% which was a $60M donation lift. People know clicking the button now means now, and spelling it out can cheapen it and create a false sense of urgency.

→ The wrong-verb. MarketingExperiments found "Try Now" reduced clickthrough by 26% against "Start Free Trial." "Try" reads as effort plus uncertainty.

→ The commitment tax. PartnerStack swapped "Book a Demo" for "Get Started" and saw a +111% lift. Nobody wants another meeting.

→ Context is a slam duck win. HubSpot analyzed 330,000 CTAs and found the ones personalized to where the person actually is (returning visitor sees "Continue your trial," customer sees "Refer a friend") converted 202% better than one-size-fits-all buttons. Passing on context to buttons can be at the user level, situation level, time of day level…lots of ways.

→ The invisible button. When Jiaona Zhang came into Linktree as CPO, a funnel audit found the little Linktree logo at the bottom of every profile was just decorative. Turning it into a "Create a Linktree" CTA almost doubled signups. Who knows, maybe your best buttons don't exist yet.

In short, lower the stated price, raise the stated payoff, and match it with context.

Onboarding steps. Activation moments. Paywalls. Empty states. Win-back emails. Every one of those has a button, and most of them live forever as task words. Price words.

But that’s backwards. The goal-gradient effect says people accelerate as they feel progress toward a stated goal. "Continue" states no goal, so there's nothing to accelerate toward. "See your first match" is a goal. Use buttons to increment people toward the payoff they want.

Button copy is a growth lever across your entire funnel. People always want the payoff. Even when they can't articulate it, even mid-flow, even on step 3 of 5.

Cheap tests that can compound. Go reprice your buttons.

I preloaded the whole framework (the price-tag rubric, the 2x2, the rewrite rules) into a prompt that audits your website’s (public) buttons and hands you rewrites plus one suggested A/B test to run first. One click, paste your site URL, done.

› The classic: Aagaard’s sign-up form copy tests, including the famous first-person button. → How to write high-converting sign-up copy (8 min)

› The $57M experiment: Optimizely’s write-up of the Obama splash-page test by the guy who ran it. → Obama’s $60 million experiment (5 min)

› The science: the Upworthy archive study on headline concreteness and clicks: when specificity helps, and when it backfires. → Nature Scientific Reports (15 min)

› The scale study: HubSpot’s 330,000-CTA analysis behind the 202% personalization number. → Personalized CTAs convert better (6 min)

🔨 A tool → GuessTheTest, a library of real A/B tests where you guess the winner before seeing the data.

📺 A video → The Death of Educational Content on YouTube. A great deep dive by humans on the slop problem