SHOTA 推出 43 款 Chrome 擴充功能，撤下 22 款並選出 2 款測試

Disclosure: I'm the developer. This post links to Chrome extensions I build and sell.

A note for English readers: this was first published in Japanese on Zenn, a Japanese developer blog (linked as the canonical source). I'm a solo developer in Japan. I don't write code myself; I build with Claude Code by giving it instructions in Japanese. Some of the extensions below target Japanese sites, and some revenue is in yen and some in US dollars — I left both as they are, without converting.

The numbers first

I build and sell Chrome extensions on my own. By the end of September 2026 I had shipped 43 extensions to the Chrome Web Store, taken 22 down, and 21 are live now. Of those, I'm testing only 2 for four weeks to see whether they can sell.

Every number here is measured. Revenue comes from Stripe, excluding purchases I made myself to test checkout. User counts are the Chrome Web Store's "weekly users" (not cumulative installs). Where I couldn't get a number, I say so.

Why I built so many

Before extensions, I built SaaS products and apps. I had almost no followers, so nobody saw them, no one came from search, and they faded away. I figured a solid platform would get me seen, so I moved to Chrome extensions.

There wasn't anything in particular I wanted to build. I had AI list candidates that looked easy to enter, and picked from those. When they didn't sell, I made more. I thought enough shots would land one. None did.

At some point, "operations" — running several terminals at once — became the main job. The more I built in parallel, the more each one's defects stood out, and I was busy fixing them. In reality, I kept fixing things that weren't selling.

Live extensions (20, leaving out 1 not listed)

Sources (by column):

• Weekly users: the "N users" figure on the Chrome Web Store (value on 2026-09-29, pulled daily into my own database). It's weekly active, not cumulative installs
• External revenue (total): Stripe payments (2026-01-01 to 2026-09-30, succeeded, refunds excluded, my own 3 purchases excluded). Yen and dollars are left as they are. Total for the extensions in the table = ¥3,480 and $106.75
• Active subscriptions: active Stripe subscriptions (as of 2026-09-30)
• Next: the categories decided on 2026-09-28 and 2026-09-30

The 22 I took down

"Before 2026-03-15" = the extension already showed a user count in the store at the time of my 2026-03-15 analysis. The exact launch date isn't available.

The reasons and dates come from my own records (decision notes and my Chrome Web Store submission history).

• The "batch takedown of 17" was a decision on 2026-09-01. The reason: "lifetime revenue $0.00 (and for 2 of them, the pricing description didn't match reality)"
• Breakdown: 21 live + 22 launched and then taken down = 43 (1 of the 21 live ones isn't in the table above)
• On top of that, 4 were never published (2 still rejected in review, 1 stopped at upload, 1 duplicate registration of the same extension)
• For 2 extensions, my records disagree on whether they were ever published (TabVault, ReShapic). I can't say either way, so they're not in the 22

Lessons from taking down 22

It sold once, so I thought it would sell again. It didn't. After switching to monthly plans, I stopped having $0 months, but it hasn't grown.

Installs go up, but surprisingly few people actually use it. Most install and forget. If it isn't used, it means nothing.

Around the time I'd built close to 40, billing stopped working and the upgrade button threw errors. When I checked on a real device, it was broken. I had it fixed, and sometimes it still wasn't fixed. How far do you trust "I fixed it"?

The shallow research wasn't wrong, but it gave me too little to decide on. If I'd dug into whether there were competitors, and whether they were paid or free, things would have turned out differently.

I leaned on AI too much. There was almost no judgment of my own in it. I didn't know much, so I thought handing it over was the right call.

As for deciding to shut things down, there was nothing I didn't want to let go of. I didn't even have a good grasp of whether shutting them down was OK.

Why I narrowed to 2

The 2 I'm keeping (Procshot / SnapReply) are not "the ones selling now." Procshot has 1 subscription; SnapReply has 0.

It started as 3, but I dropped DataPick right before launch. The feature I meant to sell (diff against the previous extraction) — when I tested the per-site readers on real pages, only 4 of 12 sites returned named rows. That's not something I could promise. For the other 8 sites (3 with 0 rows, 4 with empty names, 1 page that couldn't be read), I'm hiding them from the options, taking that feature out of the paid promise, and shipping again.

The deciding factor was competition. If free extensions with the same feature dominated, I dropped it — you can't beat free. If the only competitors were paid, I kept it — that tells you people will pay for it. I also looked at install counts: more installs pushed toward keeping, fewer toward dropping.

I'll judge after 4 weeks. The criteria are written down in advance, and I won't change them after seeing the results.

These are the 2 I kept (both are my own products):

• Procshot: turns your browser actions into a step-by-step guide with screenshots.
• SnapReply: inserts canned replies with variables into the Gmail reply box.

Lesson

Build one at a time, properly, before putting it out.

Until now, I pushed several things forward in parallel all at once. I ran them side by side with half-baked plans, and both I and Claude Code got confused. There was too much to check, and each check got sloppy. Development reached "done" and things took shape, but defects kept surfacing afterward. Even with those defects, they passed my own checks, passed Chrome Web Store review, and went live. Passing review was no guarantee that things actually worked.

So I'm switching to one at a time. I'll plan carefully, then focus on one and finish it fast. Doing everything at once actually took longer. In the end, it's all connected. Spreading out to 22, and running so much in parallel that checks couldn't keep up — the root was the same.

What's next

In 4 weeks, I'll post the results for the 2 in the same table format. If they don't sell, I'll say so.

I'm the developer, and every extension in this post is my own. The ledger behind these tables is updated every Monday, and from now on I'll post short "this week's numbers" updates.

This is a 7-part series; the next one is about "installed but not used." Following me on X (@ktg_s0215) or here on dev.to is the way to catch the next post.