A standard Performance Max campaign uses all of Google's channels (Search, Display, YouTube, Discover, Gmail, Maps) and requires creative assets: headlines, descriptions, images, videos. Feed-only PMax strips this down. You run the campaign with only a product or service feed attached, no creative assets. Without creative assets, PMax can't serve on Display, YouTube, or Discover. It effectively becomes a smarter Shopping campaign that runs on Search and Shopping inventory.
For B2B companies with clearly defined, purchasable offerings, this structure has real appeal. You get Shopping-style placement without building out a full asset group. You keep PMax off the spam-heavy Display and YouTube inventory that generates junk leads in standard lead gen setups. You limit Google's targeting latitude to contexts where someone is actively looking for something that matches your feed.
When Feed-Only PMax Actually Works
The accounts where feed-only PMax delivers for B2B tend to share a few characteristics:
They have purchasable, well-defined products. This works best when your feed contains specific items with clear names, prices, and attributes, not vague service descriptions. B2B distributors, manufacturers selling direct, companies with configurable product catalogs. The more your offering looks like ecommerce, the better this structure performs.
They have enough conversion volume. Smart bidding needs conversions to learn from. The general threshold is 30 to 50 conversions per month per campaign. Feed-only PMax that's running on Target ROAS without enough transaction volume will underperform or underspend. Low-volume B2B accounts, where a "good month" means 8 to 12 conversions, often can't feed the algorithm what it needs.
The sales cycle is short enough for GCLID attribution. Google's click ID (GCLID) tracking has a 90-day window. If your B2B sales cycle regularly runs longer than 90 days from click to close, you'll lose attribution on won deals. The algorithm learns from attributed conversions. If it can't see the closes, it can't optimize toward the signals that drove them.
The account has a stable history. Feed-only PMax's performance is built on a learned model of who buys, in what context, from what queries. That model takes months to develop. It's sensitive to disruptions: budget changes, bid strategy switches, seasonal breaks where conversion volume drops to zero.
When Feed-Only PMax Fails in B2B
The failure modes are specific and, once you've seen them, predictable.
The seasonal break death spiral. B2B markets go quiet during holidays, summer slow periods, and fiscal year transitions. When PMax loses conversion volume during these breaks, especially if combined with a budget reduction or a Target ROAS adjustment. The algorithm loses confidence. It starts spending conservatively. Less spend means fewer conversions. Fewer conversions means the algorithm stays conservative. The campaign gets stuck.
"It feels like the algorithm 'forgot' who the buyers are because of the 2-week B2B winter break and is now stuck in a low-performing loop."
"Now it's stuck in a loop where low spend = few conversions = algo gets more conservative = even lower spend."
Campaigns that were performing at 500% ROAS before a seasonal break can come back at 180% and never fully recover. The training data isn't truly gone, but the model's confidence is shaken in a way that's difficult to manually correct. The recommended fix is to drop the tROAS target entirely, switch to Max Conversion Value without a target for one to two weeks, and let Google spend and relearn. This works sometimes. But sometimes the answer is to start fresh with a duplicate campaign and pause the original.
The budget and bid change double-trigger. A common pattern: you see performance softening, so you reduce budget and lower your tROAS target simultaneously. Both changes tell Google's algorithm that conditions have changed significantly. Both trigger a learning phase. Combined, they can push the campaign into a reset it struggles to come back from. Any change to bid strategy or significant budget change should be made one at a time, with two-week intervals to assess impact before making another adjustment.
The low-conversion-volume trap. Feed-only PMax for a B2B service with 10 conversions per month isn't going to work on Target ROAS. There isn't enough data for the algorithm to find patterns. The campaign either underspends or spends on the wrong signals. If your account doesn't have the volume, this structure will frustrate you.
The Practical Decision Framework
Before running feed-only PMax for B2B, ask:
- Do I have at least 30 conversions per month to train smart bidding?
- Is my sales cycle short enough that GCLID attribution captures my closed deals?
- Can I import offline conversions (qualified leads or closed sales) rather than relying on form fills as my conversion signal?
- Do I have a plan for seasonal breaks: seasonal adjustments applied proactively, and a budget that doesn't drop to zero?
- Am I selling specific products rather than vague "services"?
If most of those are yes, feed-only PMax is worth testing. If several are no, a well-structured Search campaign will likely serve you better and give you more control when things go sideways.
What to Do When Feed-Only PMax Breaks
If the campaign has gone into a death spiral after a seasonal break or major change:
Step 1: Drop the Target ROAS entirely. Switch to Maximize Conversion Value with no target. This gives the algorithm permission to spend again without the tROAS constraint holding it back.
Step 2: Hold that setting for two weeks without touching anything else. Resist the urge to make additional changes. Give it time to relearn.
Step 3: If two weeks produces no meaningful recovery (spending less than 60% of daily budget, conversion volume still collapsed), duplicate the campaign, start fresh with a new campaign, and pause the original once the new one is spending and converting.
Step 4: Before relaunching, add seasonal adjustments proactively for any known low-volume periods. This signals to Google that a conversion rate drop is expected, not a signal to decrease bids. It's the tool that prevents the death spiral in the first place.