Rivaltics

The idea was straightforward: G2, Capterra, and TrustRadius have thousands of reviews for every major SaaS product. Those reviews are publicly visible, moderately structured, and almost nobody was mining them systematically for competitive intelligence. Point the tool at a competitor, get back what their customers actually think, who those customers are, and how to reach the unhappy ones directly.

Building the intelligence layer

A Python scraping pipeline pulled reviews from all three platforms and ran them through Gemini for AI analysis. Reviews were bucketed by sentiment and theme using a vector database. The frontend surfaced patterns: what users love, what they hate, what they say is missing. Reviewer profile data aggregated across all reviews gave you a rough ICP picture, with company sizes, job titles, and industries. The skip-tracing integration let you surface contact info for reviewers to run outbound campaigns directly at your competitor's upset customers.

The G2 problem

G2 requires a logged-in user session and runs Cloudflare's strict anti-bot detection. Getting through it reliably required selenium-driverless, CDP patches, persisted Chrome profiles, and residential proxies through SmartyProxy. The profiles and proxies inevitably get flagged over time, so the system needed to detect that and cycle them automatically. This took a lot of iteration and was an ongoing maintenance burden. When something breaks in production on a Saturday it is not subtle.

Why I shut it down

I ran Google Ads to drive early users and got to $100 in revenue. The unit economics did not work. Ads cost more than revenue, and keeping the scraping infrastructure healthy was a meaningful ongoing time commitment for something that was not profitable yet. I made the call to stop rather than keep sinking time into it.

Not every startup idea works out. The technical work was real and the lessons from building and running live scraping infrastructure at that scale were valuable. I learned more about browser automation, anti-bot systems, and what it actually costs to run a SaaS product with real infrastructure from this project than from almost anything else.