75% bounce rate — rebuilding the landing page in a day

I opened Vercel Analytics. 139 visitors in the past week. Not bad, I thought.

Then I saw the bounce rate: 75%.

That means 75 out of 100 visitors leave without doing anything. They don’t click the download button. They don’t even scroll. They just leave.

Looking at the data

Here’s what Vercel Analytics showed:

• Visitors: 139 (one week)
• Bounce rate: 75%
• Referrers: Instagram 28, Facebook 5, Google 1
• Devices: Mobile 55%, Desktop 43%
• OS: iOS 33%, Android 24%, Linux 25%, Windows 11%

A few things stood out.

Almost all traffic comes from Instagram. People tapping links in an in-app browser. These users have short attention spans — they decide in 3 to 5 seconds. They don’t scroll much.

Linux at 25% is suspicious. Likely Instagram’s in-app browser reporting its User Agent as Linux. There’s no way 25% of our visitors are actual Linux users.

Mobile is 55%. Good sign for an app download page. The problem is that these people aren’t making it to the download.

What we changed

We rebuilt the landing page over the course of a day. Not all at once — one change at a time, build, verify, repeat.

Simplified the Hero

Before:

• Badge (“Start with templates”)
• Headline
• Subhead
• 2 lines of description
• 2 download buttons
• 2 QR codes

After:

• Headline
• Subhead
• Ratings (★ 4.9 App Store, ★ 5.0 Google Play)
• App demo video

Instagram visitors only see the first screen. If there’s too much information, they read nothing. We stripped the badge, description, and download buttons — leaving only the essentials.

Download buttons aren’t needed in the Hero because the mobile Sticky Bar is always visible. On desktop, the bottom CTA section has buttons with QR codes.

Simplified CTA labels

The download button labels in the CTA section were shortened. “Download on the App Store” became just “App Store”. “Get it on Google Play” became “Google Play”. With icons next to the buttons, the long labels were redundant.

Unified the subhead

“Write it down. Focus. Finish.” became “Write it down. Organize. Finish.”

This flow — write → organize → finish — was then unified across the entire landing page and blog posts. When the same app is described with inconsistent language, the message gets diluted.

Added social proof

App Store 4.9 and Google Play 5.0 ratings in the Hero. We left out download counts and review numbers since they’re still small. The ratings alone communicate “this app is worth trying.”

App demo video

Screenshots don’t convey what the app actually feels like. We recorded a short video of adding a task in the app and placed it in the Hero. On desktop it sits next to the text; on mobile it appears below.

The mov file went from 4.4MB to 108KB after ffmpeg conversion. Virtually zero impact on web performance.

Screenshot carousel

The first carousel item was replaced with the video — autoplay, loop, muted. We also reduced screenshot sizes on mobile where they felt too large.

Screenshots were split into ko/en directories too. Same images for now, but when English screenshots are ready, we just swap the folder.

Bullet lists for Problem/Solution

Paragraphs became bullet lists. Easier to scan. People from social media don’t “read” — they skim.

What we didn’t do

Some ideas were considered but left out.

Exit-intent popup. Showing a “Wait, try it free” popup when someone tries to leave could work, but it could also backfire. Especially in Instagram’s in-app browser where the experience is already poor — adding a popup on top might push people away faster.

Mid-page CTA after Features. Adding another download button after the features section is reasonable, but too many CTAs can make the app feel pushy.

Results?

We don’t know yet. Changes were made today — results come next week.

Honestly, I don’t expect landing page changes alone to dramatically improve conversion. The fundamental question is whether “why should I download this app” gets answered in 3 seconds, and that’s a deeper problem than copy and layout.

But we did what we could. Look at the data, form hypotheses, change one thing at a time. We’ll open Vercel Analytics again next week.