Subflare is lead scoring software for B2B companies, SaaS teams, agencies and online shops in the EU. Analytics show that someone signed up. Subflare shows who it was and what they did first: which pages they read, for how long, and when they came back.
When a visitor leaves an email, AI turns it into a lead profile: an intent score, a persona, a short summary and a next step for sales. The founder came with the idea and a Figma file, but no team. He hired us as his CTO and dev team instead of giving away equity to build one.
The whole product rested on a technical bet that was still unproven. Could one small script plus AI work on any website, tell what the business sells and score every visitor by behavior alone, with no custom setup per site? The hardest part was linking a visitor's anonymous visits to the person once they typed in an email, so the score comes from what they did on the site.
Money mattered as much as tech. The founder needed something to demo and sell soon, not a finished platform before the first customer. So we agreed on speed and breadth for demos first, and depth later.
We split the work into steps, so the founder never paid for the next step before the last one proved itself.
How it went:
The demos let us change priorities as we went. Once it was clear where the sales value was, we moved the dashboard and the lead detail page forward and pushed lower-value work back. Specs lived in Notion and tasks in Jira.
The Rocketech team:
Each client gets a snippet with its own site ID and pastes it into the site once. It records page views, clicks, scrolling, navigation and form submissions. Clients can redesign the site or add forms without installing it again.
From a site's address and content, AI works out what kind of business it is (online shop, SaaS, content or services) and writes buyer personas for it. A new client gets useful results with no manual setup.
We took on the riskiest questions first, in the 2-week test build. The team built the snippet, event capture, AI site mapping and the AI step that turns an email into a persona and a score. It all worked with no blockers, so the full build started in mid-November.
The architecture changed on purpose. Early drafts used Supabase Auth and MongoDB. The build moved to AWS Cognito for sign-in, PostgreSQL for app data, DynamoDB for the stream of visitor events, and Redis with BullMQ for background jobs.
Tell us what you're building, and we'll help you find the fastest path from idea to launch.