The ROAS decay was an inventory problem.

ROAS declined consistently after the first 72 hours of any campaign. With no centralised data warehouse, optimisation decisions were being made on intuition rather than evidence — and budget kept flowing to products that were nearly out of stock.

Architected a centralised warehouse in Google BigQuery that joined real-time inventory against blended Shopify and Meta Ads metrics, surfaced in Looker Studio. That made dynamic creative management possible: spend now shifts automatically with product availability instead of days after someone notices.

Sales up 74% year on year at Black Friday, ROAS up 49% and acquisition cost down 30%, plus automated inventory and item-trend reporting the team still runs off.