Creators are paid commissions on attributed sales, so attribution is a payments problem. Heading into 4x growth, the service ran at ~100% CPU, DLQs hit six figures, and event tables outgrew the DB.
A first-party JS pixel + Shopify webhook attribution engine on Redis Streams into Postgres; migration to AWS Aurora; plus an AI pipeline: Whisper, GPT-4o matching, Gemini analysis, pgvector search.
Events table shrunk 146GB to 10GB with zero data loss on ~100M visitor rows, served reliably while prepared for 4x traffic growth. 3 AI providers in production.
Have a similar product challenge? Let's talk about the right technical path.
Discuss a Similar Build