PMax without a high quality feed is 5 to 7 times worse

Imagine: you launch PMax Google Ads expecting an explosive increase in sales, but instead you see PMax ROAS at 1-2 instead of the usual 8-10. From my experience since 2011, PMax without a quality feed really performs 5-7 times worse — this is not an exaggeration but a fact from real cases where Google’s PMax AI algorithm “goes blind” without the right input data. Why does this happen? PMax’s sensitivity to feed quality makes it vulnerable: without detailed feed attributes title, description and custom labels it generates junk conversions, training on cheap clicks and ignoring real deals. Are you ready to risk your budget on broad demand funnels where PMax degradation and a poor feed eat your ROI? I’ll break down step by step how to fix this, based on fresh insights from Google Ads reports of 2025 and our experience at BUSINESS SITE: from pharma companies to e-commerce. Read to the end: get a ready plan to optimize PMax that will lift your PMax e-commerce to a new level.

How Performance Max works in Google Ads

In my view, PMax success in Google Ads depends 70% on the Google Merchant Center feed: it provides the algorithm with data to match queries to products. BUSINESS SITE specialists have repeatedly seen how a weak PMax product feed forces the AI to rely on general signals, reducing relevance.

Algorithms are trained on input data — creatives and goals — where unique product identifiers and feed attributes like custom labels allow assortment segmentation. Without them PMax spends on non-targeted impressions, especially in PMax e-commerce. At BUSINESS SITE we integrated such a feed for an online clothing store; the result: CTR rose by 45% because the AI precisely selected visuals for queries like “summer dresses Ukraine”.

Why PMax without a quality feed delivers ROAS 5–7 times worse

Feed quality for PMax is fuel for the AI: full titles with keywords (brand + model + benefit) and descriptions with USPs raise the Quality Score. Google research 2025 emphasizes: a weak feed leads to a ROAS drop for PMax of 500–700%, because the algorithm focuses on cheap conversions without transactions, views or add-to-carts without purchases.

PMax without a quality feed performs 5–7 times worse, because PMax’s sensitivity to input data forces it to optimize for low-cost metrics rather than LTV. In our observations at BUSINESS SITE for a construction company where the feed had duplicated descriptions, ROAS started at 1.2 and after refining attributes jumped to 9.5. How does PMax work without a quality feed? It scales to cold traffic, generating junk conversions that distort training.

PMax errors without an optimized feed

Common PMax mistakes include skipping Merchant Center feed diagnostics, where 30% of products are rejected due to price or image mismatches. Duplicate unique product IDs in PMax cause internal competition, and rejected products deprive the feed of coverage. This provokes traffic cannibalization with Search campaigns.

The BUSINESS SITE team found a similar case for a travel agency: Merchant Center diagnostics showed 25% rejections due to non-unique IDs. After fixes the traffic stabilized, without overlaps with the Search core.

PMax Feed Optimization for ROAS

PMax optimization starts with optimizing the feed for PMax: add 15+ attributes, including promo_text for promotions with Nova Poshta. A correct PMax product feed raises ad rank by 2-3 points, expanding reach without increasing CPC. At BUSINESS SITE we apply feed transformation rules for automation; for a pharma company this delivered +60% to impressions.

Merchant Center feed optimization for PMax ROI focuses on relevance: integrate shipping data (Nova Poshta, Ukrposhta) and payment (PrivatBank, Monobank). The result — the algorithm matches Ukrainian queries better.

Issues in the PMax e-commerce feed

Run Merchant Center diagnostics weekly: filter rejected feed items by reasons (image_link, price mismatch). Fixing rejected items in the PMax feed is a priority: upload HD images 800×800+ and current prices.

Feed problem diagnostics in Merchant Center at BUSINESS SITE saved an e-commerce lab project: 40% of items were rejected due to expired stock; after fixing ROAS tripled. Use a comprehensive feed optimization methodology: audit + rules + A/B tests.

Feed attributes: , description, custom labels

The impact of title and description quality on PMax is huge: a title up to 150 characters with “buy [product] Kyiv delivery” increases CTR by 28%. Feed attributes title and description should include seasonality and brand.

Add custom labels (1-5) for product group segmentation – high_margin, bestseller. Segmenting product groups in PMax without overlaps allowed us at BUSINESS SITE to split products for the bank: ROAS for premium cards +12x.

Setting ROAS and PMax goals for scaling

PMax ROAS stabilizes with a PMax conversion goal (purchase, LTV >500 UAH) and a conversion volume of 50+ per week; below that the AI is chaotic. Setting conversion goals for a stable PMax requires ga4 integration with a CRM for tracking.

At BUSINESS SITE for an electronics store we set up offline conversion returns through the Monobank API: this recovered 20% of ‘lost’ deals, raising ROAS from 4 to 11.

PMax risks with fewer than 50 conversions

Risks of launching PMax when conversion volume is under 50, budget volatility and ROI risk management. The algorithm “learns” on noise, burning 30-50% of funds wastefully. To manage PMax budget risks, start with Search for the demand core and add remarketing.

Our experience confirms: for a new pharma company with 20 conversions we first increased volume through Prom.ua, then launched PMax: risks were reduced to zero.

PMax integration with GA4 and CRM for ROAS

Integrating PMax with GA4 and crm for scaling is key: import offline sales (returns after 30 days). Returning offline conversions for accurate PMax ROAS corrects the model, focusing it on LTV.

BUSINESS SITE specialists implemented this for builders: long-term ROAS scaling increased by 40%, taking into account payments after installation.

PMax Assets: mistakes, negative keywords, cannibalization

PMax assets enhance the feed: upload 5+ images, PMax video creatives (15-30 sec featuring products + a call to action). Disable Google’s automated video generation, it reduces quality by 15%. PMax negative keywords (competitors, “free”) and PMax brand exclusions protect the budget.

BUSINESS SITE practice for a travel agency: custom videos lifted conversions by 35% without auto-generation.

How to avoid cannibalization in PMax and Search

How to avoid cannibalization in PMax and Search for strategic growth: isolate the Search demand core (exact branded queries), leave PMax to expand reach with broad match. Strategically separate channels, monitor PMax insights.

We applied this for a bank: negative phrases removed overlaps, Search + PMax ROAS grew by 55%.

Limitations and best practices: URL, video, remarketing

Disable URL expansion and final URL replacement; they distort control. Best practices for video creatives in PMax: vertical 9:16 for mobile. Balance PMax remarketing with new audience (70/30).

At BUSINESS SITE for a store we disabled URL replacement: landing traffic increased by 25% with no losses.

Checklist of recommendations

• Check Merchant Center feed diagnostics weekly and fix deviations.
• Ensure a conversion volume of 50+ before scaling, integrating GA4-CRM.
• Test segmentation of product groups in PMax with custom labels for ROI.
• Add PMax video creatives, negative keywords, and brand exclusions.
• Monitor PMax insights reports for PMax e-commerce optimization.