I built a free pre-launch website checker — dogfooded 88→91, looking for brutal feedback
Hey IH 👋
I kept shipping landing pages that looked polished but hid SEO / technical / CTA issues. So I built LaunchCheck: paste a public URL → 100-point launch readiness score across search, technical, and conversion, with evidence + fix prompts. No signup.
Dogfooded on my own site: started at 88/100, then after shipping structured data landed at 91/100 (Technical went to 100).
Still early (V0.1). Want honest feedback:
- Does the score feel fair or inflated?
- What’s missing before you’d run this before your next launch?
- Would you share a report link with a cofounder/client?
Try it: https://launchcheck-alpha.vercel.app/
Happy to scan a URL you’re about to ship and talk through findings.
The 88→91 change is useful, but the stronger signal is whether users actually change what they ship because of the score. Have any external users fixed something they otherwise would have missed?
Honest answer: not enough external proof yet. Most of the 88→91 move was me fixing my own site after the report (structured data). A few people have run URLs and told me the findings matched what they expected, but I don’t yet have a clean story like “user X changed Y before launch because of the score.” That’s the metric I’m watching next — not just scans, but whether a high-impact finding actually changes what ships. If you’ve got a URL about to go live, I’m happy to run it and follow up on whether the fixes were worth it.
Nice dogfooding loop. I’d separate the score into (a) deterministic checks with pass/fail evidence and (b) heuristic recommendations, then show confidence and the expected point impact for each fix. Before treating 88→91 as meaningful, run the same set of 10–20 sites twice and track false positives; a stable score plus a prioritized fix list will be more useful than a precise-looking number.
This is exactly the right framing. Under the hood I’m already splitting deterministic checks (HTTPS, robots, metadata, redirects, etc. with live evidence) from more heuristic conversion/copy signals — but the UI still presents one score, which can feel more precise than it is. Showing confidence + expected point impact per fix is on my list. Same for stability: I’ll run a fixed set of 10–20 sites twice and track flips/false positives before I treat small score moves as meaningful. A stable prioritized fix list > a precise-looking number. Appreciate you pushing on this.