Dev.to WebDev 🛠 Dev 👁 0 📖 2 min read

OpenKeyword is live: 389k keywords, 5 markets, and what I cut to ship it

OpenKeyword opened today. It's a free keyword research tool: you type a keyword and get a 0 to 100 popularity score, the search intent, a 12-month trend and up to 50 related ideas. No account for 30 lookups a day. I wro

OpenKeyword opened today. It's a free keyword research tool: you type a keyword and get a 0 to 100 popularity score, the search intent, a 12-month trend and up to 50 related ideas. No account for 30 lookups a day.

I wrote earlier about seeding 305k keywords. This is the launch-day
follow-up.

What shipped

  • About 389,000 scored keywords in the US, UK, Germany, France and Spain. The German, French and Spanish sets are smaller (20 niches each instead of 42) and cost about $3.50 per market.
  • The website in 8 languages, with a market picker. Picking a country changes the data, and the language setting changes only the interface.
  • Optional accounts with email-code login and no passwords. An account raises the limit from 30 lookups a day to 200.
  • The Chrome extension, approved by Google yesterday. It's contributor-only: it shares the words you search after you opt in, and shows you nothing while you browse.

What I cut

  • Exact search volumes. The raw Google Ads numbers stay on my home server. The public score is a relative 0 to 100 rank against the other keywords in the same market.
  • Intent outside English. The intent labels (how-to, best, vs, question) come from English patterns, so most German, French and Spanish keywords say "Other" for now.
  • Canada, Australia and Romania. They're in the market picker as "coming soon" because there's no data yet, and I'd rather show that than an empty result.

The last-hour bugs

I did a final check an hour before deploying. Every test and build passed: 124 API tests, 59 web, 77 home server, 71 extension. Then I opened the site at phone width and found two pages that scrolled sideways.
The first was a tooltip. The "Estimate ?" badge on the score card opens a 293px box anchored to the badge's left edge. On a 375px phone the badge sits 165px from the left, so the box ended at 458px. It was hidden with visibility: hidden, but hidden elements still count for scroll width, so the whole page could be dragged sideways. On small screens the tooltip now anchors to the full label row instead of the badge.

The second was a CSS grid classic:

.layout { grid-template-columns: 1fr; }

1fr means minmax(auto, 1fr), and auto won't shrink below the content's minimum width. A three-column table with min-width: 130px cells wanted 390px, so the column grew to 390px, even though the table already sat inside its own overflow-x: auto wrapper. The fix:

.layout { grid-template-columns: minmax(0, 1fr); }

Neither bug shows up in tests. I only found them by loading every page at 375px and comparing scrollWidth to clientWidth, which I'll keep as a pre-deploy script.

How it deployed

Both Workers went up with wrangler versions upload first, so the new code sat on Cloudflare without traffic, and a timer promoted them with wrangler versions deploy <id>@100% at the planned minute. Rolling back is the same command with the old version id.

What I'd like to hear

If you do keyword research for small sites, which market or feature would make you come back? And if you've grown a dataset from user contributions, how did you get the first thousand people to share?

📰 Read the original article on Dev.to WebDev

Originally published by Dev.to WebDev. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.