A disapproved product in Merchant Center doesn't lower your ROAS. It removes the product from auction entirely. The ROAS number in your dashboard only reflects products eligible to serve. The ones that weren't? Invisible. No impression, no click, no conversion, no line item. You're optimizing a denominator that's already been quietly shrunk.
The Feed Problem You Can't See in Campaign Metrics
Google's Merchant Center documentation is clear: disapproved items are ineligible to show on the affected Google destination until the issue is resolved. Disapprovals fire when product data fails requirements, conflicts with landing-page information, or violates Shopping policies. They can hit individual products or the whole account.
There's a subtler cousin. Products with "limited" or "warning" status may still serve, but with reduced reach. And Google notes that warnings can become disapprovals if left unresolved. The product that's "sort of working" today could vanish from Shopping tomorrow.
For B2B catalogs with complex buying conditions (quote-required items, minimum order quantities, account-specific pricing), the feed-to-landing-page consistency problem is structural. A price mismatch between feed and checkout page triggers a disapproval. An availability mismatch does the same. These aren't edge cases for B2B distributors; they're the norm unless someone has built a reconciliation process.
The diagnostic move: pull the product-status report from Merchant Center, filter by issue type, and cross-reference with historical revenue contribution. That number is your at-risk pipeline, and it tells you where to start before you touch a single bid.
Performance Max Signals Are Suggestions, Not Fences
B2B teams often treat Performance Max audience signals like targeting rules. They aren't. Google's own documentation states that audience signals are guidance to Google's automation and are not hard targeting restrictions. You can upload your best-fit account list, layer in custom segments, and add first-party purchaser data. All of it helps the algorithm find patterns faster. None of it prevents the algorithm from serving ads to someone outside those lists.
Real consequence for B2B: you can't rely on audience signals alone to keep consumer traffic out. The feed itself is your first filter. Front-load titles with the terms B2B buyers actually search: part numbers, specifications, compatibility, certification. A title that leads with "Industrial Grade 304 Stainless Steel Flange, 4-inch ANSI 150" attracts a different click than one that leads with a brand name and a marketing tagline. Meanwhile, upload account lists of high-LTV customers and let the algorithm find similar patterns, while accepting that it will also explore beyond those lists.
The Conversion You're Optimizing Toward Might Be Wrong
This is where most B2B Shopping programs leak value. Performance Max optimizes toward whatever conversion action you tell it matters. If that action is a form fill or a shallow lead event, the algorithm will find you more of those. Google's own lead-gen guidance emphasizes accurate conversion tracking and, where possible, deeper-funnel signals such as qualified or converted leads to avoid optimizing to low-value actions.
For B2B commerce, "deeper funnel" might mean a completed quote request with a valid business email, a purchase above a minimum order threshold, or an offline conversion imported back into Google Ads at the qualified-opportunity stage. The hypothesis: if you shift your primary conversion action from form fill to qualified lead or completed purchase, then cost per qualified opportunity should decrease, because the algorithm stops rewarding volume and starts rewarding signal quality.
Your stop-loss should be a floor on total qualified conversions per week. If qualified volume drops below that floor for two consecutive weeks, widen the conversion definition back one stage and retest.
Asset Groups Are Architecture, Not Decoration
For a large B2B catalog, segment asset groups by product family or use case. Don't dump everything into a single group and hope the algorithm sorts it out. The practical constraint: split too aggressively and each asset group starves for data. Split too loosely and your creative can't speak to the buyer's actual problem. Google's guidance recommends providing a variety of creative assets, including up to 15 headlines and 5 descriptions per asset group, along with images and videos in different orientations. For most B2B teams, starting with three to five asset groups organized by top product categories, each with full creative coverage, is a more stable architecture than fifteen thin ones.
Use custom labels to encode margin tier, lifecycle stage, and inventory depth into your feed. These labels become the levers for bidding decisions and reporting cuts. Maintain a shared doc that defines what each label means so paid media, finance, and merchandising interpret the values the same way.
The Ongoing Review That Actually Protects ROAS
None of this is a one-time setup. Feed health degrades as products change, prices shift, and landing pages get updated without feed synchronization. Merchant Center reviews can take up to seven business days; after approval, products may become eligible within 24 hours. That's a week of lost impressions for every disapproval you catch late.
Build a weekly review cadence: check disapprovals and warnings, audit search terms for consumer bleed, compare conversion value by asset group, and flag any SKU where margin data is stale. The teams that treat Shopping optimization as a recurring ops process, not a campaign launch, are the ones whose ROAS numbers actually hold up under scrutiny. The ones that don't are reading a dashboard that's already missing half the picture.