eCommerce PPC Mistakes That Quietly Kill ROAS

Why does eCommerce ROAS underperform without obvious errors?

Most eCommerce PPC ROAS problems are structural, not tactical. The headline metrics look reasonable — impressions are up, clicks are flowing, conversions are recording. The problem is in what the algorithm is learning, which products it is prioritizing, and whether the attribution model is giving credit to the right campaigns. These are quiet problems that compound over months before they become impossible to ignore.

Product feed quality: the root cause most accounts ignore

Product feed quality is the single most impactful factor in Google Shopping and Performance Max performance. The feed determines which queries your products match, how your listings appear in Shopping surfaces, and whether the algorithm has enough signal to optimize toward the right buyers. Poor feed quality suppresses all downstream performance.

• –Title structure is wrong for Shopping. Shopping titles need the most important attributes front-loaded: brand, product name, key spec, size, color. A title that starts with a marketing tagline ("Experience ultimate comfort with...") will underperform a descriptive attribute-first title ("Nike Air Max 90 Men's Running Shoe Size 10 White/Black").
• –GTINs are missing or incorrect. Without valid GTINs on brand-name products, Google cannot match your listings to known product entities. This reduces impression eligibility and prevents price comparison features.
• –Images are too generic. Product images without clean backgrounds or with lifestyle context instead of the product itself underperform on Shopping surfaces where buyers compare multiple listings side by side.
• –Custom labels are not in use. Custom labels let you segment products by margin tier, seasonal priority, or bestseller status so campaigns can be structured around business value rather than just category.
• –Prices and availability are stale. A feed that does not update daily can serve ads for out-of-stock products or wrong prices, triggering Merchant Center disapprovals and wasted clicks.

Campaign segmentation mistakes that make optimization impossible

Poor campaign segmentation mixes products with very different margin profiles, conversion rates, and price points into the same campaign, where budget flows to whatever drives the most volume rather than the most profit.

Performance Max configuration errors that suppress results

Performance Max mistakes account for a significant portion of eCommerce ROAS problems in 2025 and 2026. The most common errors are not about the campaign type itself but about how it is configured alongside other campaigns.

• –No brand exclusions. PMax will serve on branded search queries unless brand terms are explicitly excluded at the campaign level. When PMax absorbs brand traffic, it claims attribution for conversions that would have happened through a brand Search campaign anyway. ROAS looks strong; it is not incremental.
• –Running PMax without a parallel brand Search campaign. If you exclude brand from PMax but have no dedicated brand Search campaign, brand queries go unserved. Set up a brand Search campaign first, then apply exclusions to PMax.
• –Poor asset quality. PMax needs high-quality images, video, and copy to compete across YouTube, Display, and Discovery placements. Minimum-viable assets produce minimum-viable results.
• –No audience signals. Customer match lists and CRM data give PMax a starting point for finding the right buyers. Running PMax without audience signals makes the learning phase slower and more expensive.
• –Too low conversion volume to optimize. PMax needs at minimum 30 to 50 conversions per month to run Target ROAS effectively. Below that, use Maximize Conversion Value without a ROAS target.

Attribution blind spots that inflate reported ROAS

Attribution errors make ROAS look better than it is, which leads to over-investment in campaigns that are not producing real incremental value. These are the most dangerous ROAS problems because the account appears to be performing well while actually wasting budget.

• –Last-click attribution on multi-touch eCommerce journeys. Buyers often touch multiple ads before purchasing. Last-click gives all credit to the final ad, usually a brand or remarketing ad, and zero credit to the earlier prospecting touchpoints. This undervalues prospecting campaigns and overvalues brand.
• –Duplicate conversion counting. If both a GA4 purchase event and a Google Ads conversion tag fire on the confirmation page, purchases may be counted twice. The reported ROAS doubles. The algorithm optimizes aggressively. Real ROAS is half what is reported.
• –Transaction value mismatches. If the purchase value passed to Google Ads does not match actual order value (pre-discount vs. post-discount, excluding returns), the ROAS target is set against the wrong revenue figure.

What to fix first

The sequence matters. Starting with bid strategy changes when tracking is broken produces worse results, not better ones.

1. Audit conversion tracking accuracy. Confirm purchase values and deduplication.
2. Fix the product feed: titles, GTINs, images, daily refresh.
3. Apply brand exclusions to PMax and set up a dedicated brand Search campaign.
4. Segment campaigns by margin tier or product priority using custom labels.
5. Add audience signals to PMax (customer match, website visitors).
6. Set ROAS targets based on validated historical data, not aspirational numbers.

For a deeper look at feed optimization specifically, see the Merchant Center feed audit checklist. For the full Performance Max vs Search campaign comparison, see the PMax vs Search guide.