Emergence Index
Live leaderboard of SaaS categories growing fastest now
Hey IH,
Just shipped The Emergence Index: a free, live leaderboard of which SaaS categories are growing fastest globally, automatically detected from certificate transparency logs. emergenceindex.com
The build
12 days from idea to live
Rails frontend, Go backend (my comfort stack)
Detection: I monitor Certificate Transparency logs for new SSL certs, which gives me a real-time signal every time a new website launches anywhere in the world
Categorisation: an LLM generates a problem/solution statement for each new SaaS, I embed those statements, and a clustering algorithm finds natural categories without any manual taxonomy
The leaderboard ranks clusters by 28-day launch volume and growth velocity, updated continuously
The economics
$370 in OpenAI credits during build (mostly burned re-running classification + embeddings until the cluster quality was acceptable
$180/month all-in to run (compute + LLM inference + infra)
$0 in ads
For comparison, $180/month wouldn't buy me 200 clicks on "Apollo alternative" keywords
The thesis
Paid ads stop working the moment you stop paying. A free tool compounds: through SEO, citations and word-of-mouth. Every category page is also a long-tail SEO landing page targeting searches like "[category] SaaS companies." The FAQs on each page are written for crawlers as much as humans, signaling Google that the site is a knowledge hub.
The bet: 6-12 months from now, this drives more qualified Letrics traffic than the equivalent ad spend ever could. We'll see.
What I'd love from this community
Anyone else run an Engineering-as-Marketing play? What worked, what didn't?
For those further along on free-tool-to-paid-product funnels: what was your conversion rate from free tool user to paid customer, and how long did it take to materialise?
Honest critique on the tool itself is also welcome: emergenceindex.com
Happy to go deep on the CT log pipeline, the clustering approach, or the unit economics in the comments.
Cheers.
1 Comment
Honestly this is a pretty clever engineering-as-marketing play. Using CT logs as a discovery source is smart too since you’re basically getting real-time startup creation signals without relying on scraped directories. Curious how painful the clustering quality/debugging became once the volume started increasing.