songdle.win
songdle.win ↗
A daily rap song-guessing game, Wordle-style. Three tracks a day, six stages per round, artist and genre modes, and an unlimited mode for people who want more.
213,689 visitors and 1.1M page views in the 12 months to September 2026, all organic. Numbers are from Vercel Analytics.
It's good at getting players and bad at keeping them. Two big spikes, in March and August, both faded, and without event tracking I couldn't see why.
How it's built
I kept it small enough to run alone between school and work, and cut anything that didn't feed the daily habit.
- No framework, no build step, no database. Plain JavaScript, HTML and CSS using ES modules.
- Same puzzle for everyone, nothing to store. The daily challenge is a seeded random pick from a frozen JSON snapshot of songs, run in a Vercel serverless function.
- Unlimited mode goes through a small serverless wrapper around the iTunes Search API, and audio previews stream from Apple's CDN.
- No accounts. Game state lives in the browser's local storage, so you can start playing in one click.
Traffic pushed it past Vercel's free plan, so it runs on Pro now, and I'm weighing a move to Cloudflare to keep hosting costs down.
The numbers
- Best month: August 2026, with 87,786 visitors and 518K page views. The peak day was August 27, with 5,926 visitors.
- About 90% of people who land on the home page go on to start a game.
- Visitors who come in directly view about 4.5 pages per visit. Visitors from Google view about 1.2.
- Players come from everywhere. After the long international tail, Australia and the US send the most, then Canada.
What broke in August
In early August I shipped 11 production deploys in two days, aimed at growth. Search traffic did grow, from about 4,200 Google visitors a week to almost 15,000. Then it collapsed, down more than 90% from the peak in under a month. The causes, in order of damage:
- I moved domains the wrong way. I moved from playsongless.win, which had 14 months of search history, to songdle.win using canonical tags instead of 301 redirects. Both domains stayed live with the same content, so they split the ranking between them.
- I targeted the wrong searches. I optimized for the name of a well-established game with a similar name. People searching for that game landed on mine and left. Tens of thousands of search visitors never started a round, and Google demoted the page.
- A banner leaked into search results. A "play fullscreen on songdle.win" banner, meant for people playing inside other sites' iframes, showed for everyone, including Google's crawler. It became the first line of my search snippet.
- A referral source dried up. A Wordle-games directory that had been sending thousands of visitors a week dropped to zero the same week the banner shipped.
The deeper problem
The same curve happened twice: a spike in March, then decay, and a spike in August, then decay. The game acquires players well and doesn't hold them. I had zero custom analytics events, so I couldn't see how many people finished a round, came back the next day, or quit at a particular stage.
The fix list:
- Remove the banner.
- Pick one domain and 301-redirect the other.
- Rank for what the game actually is (rap, three tracks a day, artist and genre modes) instead of another game's name.
- Add events for round completion, return visits and the stage where players drop off.
- Tag the share button so I can see which shares bring players back.
The short version of what this taught me: distribution is designed, not hoped for.