Why do fashion brands struggle with Performance Max?

Many feed apps and hand-built feeds ship every size and color as its own product, with no item-group ID tying them together. One dress in 8 sizes and 4 colors becomes 32 separate listings, and each listing collects one thirty-second (1/32) of the dress's sales data. Google never sees one strong signal, so it bids timidly on all 32, and the budget fragments across variants that are really the same product.

None of this shows up in the campaign screen. The fix is item-group IDs, which compress the variants into one parent product so the conversions pool and the bidding firms up. Same dress, same ads. Only the feed grouping changed.

Smart Bidding learns a pattern and then fashion changes the catalog underneath it. A typical fashion brand runs around six collection drops a year, and every drop is new products with zero history, arriving right as the old season winds down.

That is a learning-phase problem on a loop. The way through is planning the feed and campaigns around the drop calendar instead of reacting to it: new-season products staged and launched ahead of demand, so they gather their first conversions before the sales window peaks rather than during it.

Fashion asset packs go stale on roughly a six-week cycle on trend-driven pieces. Run the same creative through two seasonal moments and click-through rate drops hard, around 30 percent, while the campaign settings look untouched. Most niches can ignore asset rotation for months. Fashion cannot.

Structure is the last piece: splitting PMax by margin tier and collection type stops Smart Bidding from cross-bidding cheap basics against hero pieces. Most fashion accounts we take over are stuck not because the products are weak, but because the feed is broken and the assets are six months old. The full fashion playbook, including the exact campaign structure, is on our blog.