sku_mapping.csv whose product group is —, and their child ASINs. Add or drop a SKU there and the dashboard follows on the next build; the group list itself is the COLLAGEN_GROUPS setting at the top of collagen_dashboard.py. Collagen Booster (capsules), TriCollagen and Beauty Collagen are not included.Business_Reports_-_ASIN_Totals_by_Day__Historical_.csv, the Performance Dashboard's Amazon sales feed). Units are units ordered from the same report.Amazon_Ads_Performance__Historical_.csv (daily Nucleus pull, from —). Spend is as reported by the Ads API; the monthly "Extra AMZ Spend" true-up in Historical Spend Gaps is not applied._fulfilled_shipments/raw, history from —). Two consequences. First, the report trails orders by one to two weeks, so the newest weeks keep filling in as items ship: a week's shipped buyers are compared with its distinct Collagen orders (Amazon_Orders_by_Day__Historical_.csv) and the week is treated as final once 90% of orders have shipped rows — until then its new-customer and buyer counts are italic and flagged "still filling in". Second, "new" is only as good as the history behind it — a week with fewer than 12 months of buyer history behind it overcounts (everyone looks new at the start of history). Those last-year cells are italic. "Buyers" in the table is distinct Collagen buyers that week, new or returning.