Asset Group Segmentation Decision at Thread & Loom
Thread & Loom is a direct-to-consumer apparel brand with 2,800 SKUs across women's knitwear, denim, and outerwear. Its single Performance Max campaign spends $6,500/day and produces about 90 conversions per month at blended ROAS 4.4; 60% of conversions come from 8% of SKUs, mostly core denim. The new head of growth wants to split the campaign into 12 asset groups - one per category-season combination - each with its own headlines, lifestyle imagery, and search themes, arguing that more asset groups give the AI 'more precise signals'. Creative production can supply roughly four distinct image sets per quarter. A separate brand Search campaign already exists. There is no standard Shopping campaign running alongside, and monthly conversion volume by product line ranges from 3 to 40. The proposal is scheduled for next week's budget meeting.
1. Evaluate the 12-asset-group proposal using the conversion math. Is the segmentation justified?
Read the worked answer
No - the math does not support it.
At 90 conversions per month split across 12 asset groups, each group sees about 7.5 conversions per month. That is well below the volume needed for stable bidding and for meaningful creative rotation, which in practice is in the tens of conversions per group per month, not single digits. Each group would spend most of its life in a learning phase, and budget would be spread so thin that no group accumulates enough evidence to improve.
Two conceptual errors compound this. First, asset groups are creative and signal containers, not catalog segmentation tools. Product coverage is governed by listing groups and feed labels, so splitting asset groups does not give finer control over which SKUs serve. Second, 60% of conversions coming from 8% of SKUs means the account already has a concentration problem; fragmenting the campaign dilutes the signal on exactly the denim SKUs that are carrying performance.
The 'more precise signals' argument confuses more inputs with better inputs. Twelve thin groups produce noisy signals, not precise ones.
2. What structure would you recommend instead, and how would you prove it was the better call?
Read the worked answer
Recommended structure: three asset groups, not twelve.
Align them to the three categories - knitwear, denim, outerwear - because those are the distinctions that matter for creative, margin, and seasonality, and because each can plausibly reach the volume threshold. Creative supply of four image sets per quarter comfortably covers three groups; it cannot credibly differentiate twelve.
Handle seasonality with search themes and feed labels rather than new asset groups. Use listing groups if the team needs SKU-level bid visibility, and keep brand traffic isolated in its existing Search campaign so it does not distort Performance Max reporting.
To prove the decision, run a campaign experiment: split traffic 50/50 between the current single group and the three-group build for four weeks, holding budget and tROAS constant. Judge on conversions and ROAS, not on CTR, and check that each group clears a volume floor before drawing conclusions. If a group stalls below roughly 30 conversions in the test window, consolidate it. Document the result so the next restructuring debate starts from evidence.