BlocInsights
A qualitative research platform that turns raw human input into a structured, attributed insight report. A brand either uploads its own interviews and stakeholder conversations, or asks a curated panel of senior marketing and creative leaders directly; either way, input runs through four structured fields, not a single line of free text, and AI synthesises it into a report with a genuine, attributed risk register, every raw verbatim staying visible next to the synthesis so the brand can check it against the source.
Livewww.blocinsights.comTwo Ways In, One Synthesis Engine
Most research and insight tools aggregate large volumes of structured data points. BlocInsights does the opposite: a small number of credible, individually attributed points of view on one sharp strategic question, structured well enough for AI synthesis to be worth trusting.
The platform ships two paths into the same engine. Self-serve: a brand brings its own stakeholder interviews or customer conversations and generates a report on demand, no external recruitment, no wait. Panel: a brand asks a curated panel of senior marketing and creative leaders directly, with a synthesised report delivered in 48 hours. Both paths answer four required fields, stance, reasoning, risk, and recommendation, rather than one open textarea, so every entry is structured enough for the AI to reason over directly rather than first having to extract structure from prose.
One Schema, Additive Not Forked
Next.js 16, Supabase with Row Level Security enforced at the database layer across every core table, not just hidden in the UI, verified twice: once by reading the schema, once by testing against a real authenticated session. Stripe handles payment, refunds, and failed-payment states explicitly, with idempotent webhook handling against Stripe's at-least-once delivery guarantee, across two separate price points for the two paths.
Self-serve was built as an addition to the existing schema, not a parallel product: the same responses and reports tables serve both paths, distinguished by a source and study_type flag rather than duplicated logic. A self-serve study skips panel dispatch entirely and exposes an on-demand "generate report" action instead of an automatic response-count threshold, since there's no fixed panel size to compute a threshold against. A self-healing synthesis watchdog still covers the panel path, recovering any study stuck mid-processing with zero added infrastructure.
Built to Survive Contact With Reality
A real production bug, a failed synthesis silently rolling a study back to 'open' status, making it look like it had never started rather than surfacing that it had failed, was found and fixed by re-verifying the system end to end after a feature change, not assumed clean because the code looked right. The fix included making failures visible on the page itself rather than silently reverting to a state with no signal anything had gone wrong.
The self-serve path was verified the same way: a real study created, a real structured response submitted, and a real report generated against the live production database and a live AI call, not a fixture, confirming the synthesis correctly attributes a self-serve respondent by their given label instead of misrepresenting them as an anonymous panel member.
A Report Worth Taking Seriously, On Either Path
The differentiating bet is structural: most competitors in this space, expert networks, marketing-research publishers, membership councils, either monetise access to a senior audience through community/content/events, or run live calls rather than structured written input. Nothing currently does on-demand, single-question, structured synthesis, for this specific audience, whether the input is a brand's own research or a paid external panel.
Verified end to end after every fix: a fresh study, a structured response, a correct synthesis, a report with a genuine risk register, complete with no manual intervention, on both the panel path and the newer self-serve path.