Google Performance Max gets a bad reputation in DTC circles and the reputation is partly earned. Half the brands I audit who run PMax have it configured in a way that guarantees weak performance. The other half are running it correctly and getting good results. The difference is not spend. It is what signal the campaign is eating.
This is field notes on the specific inputs that separate PMax campaigns that compound from PMax campaigns that burn budget.
What PMax actually is
Performance Max is Google's automated campaign type that runs across Search, Display, YouTube, Discover, Gmail, and Shopping simultaneously. You provide a goal, a budget, some creative assets, a product feed, and (optionally) audience signals. Google's algorithm decides where each dollar goes. The pitch is similar to Meta Advantage Plus Shopping: hand over control, let the automation outperform manual structure.
For DTC brands with a product catalog, the catalog-and-creative pairing is the core of PMax. If your product feed is weak, PMax is weak. If your creative assets are sparse, PMax compensates with Display placements that rarely convert for DTC. If your conversion tracking is off, the whole thing runs blind.
The signal audit
Before touching the campaign, audit the six inputs PMax actually uses.
Input 1: Conversion tracking
Default GA4 ecommerce tracking is not enough. PMax needs enhanced conversions (hashed user data passed on conversion events) and ideally offline conversion import (post-purchase data from your CRM or Shopify admin). Without enhanced conversions, Google's attribution is weaker, match rate is lower, and the algorithm optimizes against a noisy signal.
A DTC brand running PMax with default GA4 tracking is feeding the campaign the same kind of weak signal that a Meta campaign would get from browser-only pixel. The DTC Stack Audit covers the enhanced conversions setup as part of the broader measurement stack.
Input 2: Product feed quality
PMax draws product data from Google Merchant Center. The feed quality drives Shopping placements, which for most DTC brands are the highest-converting placement type in the PMax mix. Weak feed means weak Shopping placements.
What matters in the feed: product titles that match search intent (not SKU numbers), product descriptions that include specification language buyers search for, correct GTINs where available, accurate product categories, high-resolution images, and variant structure that covers the combinations buyers actually search for.
Most Shopify brands use auto-generated feeds that pull product titles from the Shopify admin. If the Shopify admin titles are internal labels like "Performance Tee Mk2" rather than search-friendly like "Men's Performance Running Shirt Mk2, Quick-Dry," the feed is inheriting the weakness. Audit and rewrite.
Input 3: Audience signals
PMax lets you provide audience signals that guide (but do not restrict) the algorithm. Most brands skip this because "Google says PMax will find the audience automatically." That is technically true and operationally wrong. Audience signals speed up learning and focus early spend on likelier converters.
The signals that matter for DTC: your customer list uploaded and matched (these are your LTV definitions), high-intent site visitors from the past 30 days, in-market audiences for your category, and affinity audiences that overlap with your demo. Feed them all. Let PMax decide how to weight them.
Input 4: Asset group segmentation
Running one asset group that spans all your products is fine for a brand with a single hero product. For a brand with 40 SKUs across four product lines with different margins, one asset group groups everything into a single signal pool and lets PMax optimize for blended conversion without respecting margin.
Segment asset groups by margin tier or product line. Your 60 percent gross margin hero line gets its own asset group. Your 20 percent gross margin accessories line gets a different one. The campaign-level budget does not change, but each asset group can now be optimized against its own tROAS target.
Input 5: Creative volume
PMax wants 15 headlines, 5 long headlines, 5 descriptions, multiple images in 1:1 and 16:9, logos, and video. If you provide three headlines and one image, the algorithm has no assets to rotate through, creative fatigue hits fast, and Display placements (which PMax will use regardless) look generic.
Provide the full asset menu. If you do not have video, PMax will auto-generate video from still images. These auto-generated videos are bad. Supply real video.
Input 6: Bid strategy at the right time
Day one of a PMax campaign should use Maximize Conversions (no target CPA cap). This gives Google room to explore. Once you have 30+ conversions and stable performance (usually 2-3 weeks), switch to Maximize Conversions with target CPA or target ROAS. Going to tROAS from day one starves the campaign because it has no baseline to optimize against.
Where PMax compounds
When the six inputs above are strong, PMax typically pulls 15-25 percent of a DTC brand's total paid spend at a blended MER that holds its own against Meta. The specific case where PMax shines: brands with catalogs of 50+ SKUs where search-intent matching is a big part of the buyer journey. The search + shopping signal PMax aggregates is hard to replicate in a pure social channel.
Where PMax underperforms: brands with narrow catalogs (fewer than 15 SKUs), brands with weak search volume on their category, and brands whose product is discovered through social video rather than active search. For these brands, PMax is rarely more than 10 percent of blended spend.
How to measure it honestly
Do not measure PMax on the platform ROAS alone. PMax's attribution claims tend to be aggressive, especially around branded search conversions that would have happened anyway. Run the same blended MER discipline you run on Meta. MER over platform ROAS covers the blended math.
The other honest measurement is incrementality. If you pause PMax for two weeks and your overall revenue drops by less than what the PMax platform ROAS implied, PMax is claiming more credit than it deserves. If revenue drops by more, PMax is doing more than it reports. Most brands I audit see a pause-test result somewhere between those two extremes.
Brand search carve-out
Always exclude brand search from your PMax campaign. By default, PMax will happily spend on branded search queries and claim the conversions as PMax wins. They are not PMax wins. They are branded search conversions you would have gotten for a fraction of the cost in a dedicated branded search campaign.
Set up a separate branded search campaign with brand-only keywords. Add negative brand terms to your PMax campaign. This single change often improves PMax-reported ROAS by 30 percent while simultaneously making the ROAS claim more honest.
“Most PMax campaigns look broken because they are fed broken signal. Fix the signal diet and the campaign gets obvious.
”
Should I run PMax alongside traditional Google Shopping campaigns?
No, usually not. PMax overlaps with Shopping and the two cannibalize each other. Pick one. For most DTC brands, PMax with strong feed quality outperforms manual Shopping. If you need granular control over bidding per product, manual Shopping is still useful.
How much budget does PMax need to learn?
PMax needs roughly 30-50 conversions in the first 2-3 weeks to stabilize. At a $100 AOV, that is $100-200/day minimum for a campaign to be worth running. Below that, the campaign oscillates.
Can I see where PMax is actually spending my budget?
Partially. PMax gives placement reports at the asset group level but not at the keyword or exact placement level. Use Asset Group Insights and Search Terms reports together to piece together where spend is going.
Related reading
This piece is part of the paid social for DTC operators hub. The parallel discussion for Meta's automation is the Advantage Plus Shopping decision log. For the blended measurement framework that makes PMax ROAS honest, MER over platform ROAS is the companion read. For the conversion tracking and feed quality work that makes PMax function at all, the DTC Stack Audit runs the upstream checks.