For visual search to work, every sewing pattern had to be tagged with highly specific attributes — silhouette, structure, sleeve type, fabric behaviour — things like whether a garment is A-line or fitted, flowy or structured, sleeveless or bell-sleeved. At the time, this work was being done manually by reading pattern descriptions and applying tags by hand. As the pattern library grew into the hundreds, this approach became unsustainable. Manual tagging was slow, inconsistent, and heavily dependent on individual interpretation, making it difficult to maintain the accuracy required for search, filtering, and recommendations. At the same time, the underlying data lived on external websites as unstructured HTML, written in natural language with no standard format. Critical details were buried in long descriptions, and there was no reliable way to convert this information into clean, structured data at scale. Alongside this technical challenge, the business itself was scaling rapidly. PatternMatch was onboarding designers, running outreach, collecting beta sign-ups, gathering tester feedback, planning social content, and managing product readiness — all while building the app. These activities were spread across disconnected tools and documents, with no single system tying product data, designer relationships, and operational work together. The result was mounting complexity. Founder time was being consumed by admin, manual data clean-up, and context switching. Pattern ingestion couldn’t keep pace with demand, and the lack of a unified operating system made it difficult to move quickly, test assumptions, or prepare the business for launch. PatternMatch didn’t just need automation. It needed a scalable operating system that could power both the product and the business behind it.