Every niche carries a status. Pick a bucket (chip), look, then advance it a stage and watch it run. Rejects are archived, never deleted — add a rule and re-run to thin the pool. No P&L here; that comes post-quote.
Pick a bucket above.
One corpus = one buyer opportunity, in any language. Translations and ordinary material/size/configuration variants belong to that corpus and are analyzed downstream. Only “Same corpus” merges. Both “Separate” choices always preserve a different product opportunity.
Type seed keyword(s), pick the market, and DataForSEO returns related search ideas → they land in "Raw ideas", each stamped with the exact seed + market that produced it. The full API call is saved (open any raw idea → "Stored calls").
This is the real text the "Classify" step sends to the AI for every batch of keywords (the keyword list is added at the end). Change it, then re-run on the Raw ideas — your version is saved and used from then on.
A rule fires as soon as its input exists (price/size stay inert until enrichment). Add one, then Re-run.