Performance Max for dropshipping (the structure 200+ accounts use)

Performance Max is the engine on a Google Ads dropshipping account. Single PMax with the entire catalogue dumped into one asset group is the most common failure mode we see on accounts that come to us broken. Two asset groups split by margin tier beat one asset group by 17% ROAS at the same spend across our MCC. Smart Bidding finally has signal worth optimising against.

That is the structural unlock. The full Performance Max playbook for dropshipping below covers the asset group structure, feed-only setup, audience signals, value-based bidding, and the leaks that bleed accounts that look fine on the surface.

Why one Performance Max for everything fails on dropshipping

PMax can show your ads everywhere. Search, display, YouTube, Discover, Gmail - all of it. That sounds good. The problem is when you hand Google a catalogue of 500 products and one asset group, Smart Bidding finds the easiest conversions and bids hardest on those. Especially early on, when you don't have much data yet, it will spend on cheap properties that don't give you the best return.

The high-margin SKUs that need a steeper bid to win get under-bid. They look harder than the easy volume products. So they get starved. The low-margin products eat the budget because they convert easily at thin margin.

Result: ROAS looks fine on paper, profit is hollowed out, and the account plateaus around month three when the easy conversions saturate.

The fix is structural. Margin-tier custom labels written into custom_label_0 daily, then asset groups split along those tiers.

The two asset group split (the structural unlock)

The structure that prints on dropshipping accounts in our MCC.

Asset group 1: Champions. Products in the top margin tier. 30-50% of catalogue depending on margin distribution. Aggressive tROAS target. Creative assets included (real product photography, lifestyle shots, video where it exists). Audience signals optional - this group should win on feed and creative quality.

Asset group 2: Wasters and Sleepers. Products in the bottom margin tier. 50-70% of catalogue. Tight ROAS floor (we run 90% of break-even). Feed-only assets. No audience signals. Low daily budget allocation through the campaign-level budget.

Both groups inside one PMax campaign. Single budget. Smart Bidding allocates between groups based on opportunity. The constraint of having distinct ROAS targets per asset group forces Smart Bidding to actually look at margin tier when bidding.

This split beats single-PMax setups by 17% ROAS at the same spend across 12,304 PMax campaigns we audited in our MCC. Five asset groups did worse than two. The number is not arbitrary - two is enough granularity for Smart Bidding to differentiate. More than two fragments the conversion data and learning slows.

MEDIAN ROAS BY ASSET GROUP COUNT

Same spend - structural lift across 12,304 PMax campaigns

Single asset group

1.0x

Catalogue dumped in, Smart Bidding optimises against the easy products

Two margin-tier asset groups

+17%

Champions and Wasters split, Smart Bidding bids against margin tier

Feed-only Performance Max (when to use it and how to keep it feed-only)

Feed-only PMax means no creative assets - just the product feed. Google serves Shopping-style placements based on the feed image and the product data.

In 99% of the cases at ZenoX, across every client we manage, we start feed-only. Shopping ads give you the best ROAS in the beginning. If you want YouTube or Discovery later, you run a separate campaign for that. You do not want all of it inside one PMax because you lose control over where the spend goes.

There are some exceptions. If a client already runs Facebook ads very successfully and has a lot of data, we might add creative assets from the start. But that is rare. The default is feed-only.

In 99% of the cases with our agency, we do feed only. Shopping ads give you the best ROAS. If we want to run on YouTube later, we do a separate campaign for that. We don't want all of that in the PMax because you lack control.

Feed-only PMax is right for the Wasters / Sleepers asset group on most dropshipping accounts. The spend per product is low. The marginal lift from creative is not worth the production cost.

Feed-only PMax is wrong for the Champions asset group on accounts above $20K/month spend. The spend is meaningful. Real lifestyle photography, video, and testimonials lift the Champion bid materially. Production is worth it.

Mid-range accounts ($5K-20K/month) should run feed-only on both groups while learning, then add creative to Champions once the structure stabilises.

The most common mistake that breaks feed-only

Here is what trips people up. The moment you add assets to your asset group - even one headline or one image - it is not feed-only anymore. Google will start spending on other properties. It can pull random images from your landing page and run display ads. It can generate text from your website and run search ads. Then suddenly 10-20% of your spend goes to cheap placements you never meant to target.

Two things protect feed-only status:

1. Skip the asset group entirely during setup. When Google prompts you to add headlines, descriptions, images, and videos - skip it all. Do not add anything.
2. Turn off asset optimization in campaign settings. This is a separate switch. Even if your asset group is clean, asset optimization gives Google permission to pull from your site and create assets on its own. Switch it all off.

Also check from time to time that Google has not silently re-enabled these settings via recommendations. They do that.

Two things you must set up before feed-only PMax will work

These are the two most common mistakes I see when people set up a feed-only PMax. Get them wrong and the campaign either will not spend at all or will spend on the wrong data.

First: tracking. You need tracking in place before the campaign goes live. Without it, the campaign will not spend. Use server-side tracking - something like Elevar or Tracify. Pixel-only tracking loses 30-40% of conversions to iOS Safari and ad blockers. Smart Bidding bids on whatever signal it gets. If that signal is incomplete, the bids are wrong.

Second: connect Merchant Center. Go to Tools and Data Manager and make sure your Google Merchant Center account is connected to the ad account. If it is not connected, the feed does not exist as far as PMax is concerned. The campaign cannot run on a feed it cannot see.

Both of these have to be in place before you set the campaign live.

Campaign settings that keep feed-only clean

A few settings matter when you are setting up feed-only PMax. Get these right once and they protect the setup.

Bidding: start without a tROAS target. Bid on conversion value, not just conversions. But do not set a tROAS number at the start. You have no data in the campaign yet. Let it learn first. Add the target once you have enough conversion history.

Location targeting: presence only. The default is "presence or interest." That is sneaky - it means Google can show your ads to someone in the Philippines if they have shown interest in the US. Switch to the second option: people in or regularly in your included locations. That keeps spend where you want it.

Languages: native language plus English. A lot of people run their browser in English even when they live somewhere else. Add English alongside your target market's native language.

Devices: computer and mobile only. Tablets and TV screens have lower conversion rates for ecom. They work fine for awareness, but for a conversion campaign, stick to computer and mobile.

Audience signals: usually keep empty. Unless you already have valuable first-party data - past buyers, past converters - keep the audience signals empty for the Wasters / Sleepers group. For Champions, customer match lists of 90-day buyers and past visitors who hit checkout are the only signals worth adding. Skip broad in-market keywords and generic affinity audiences.

Value-based bidding for dropshipping (the other unlock)

Most dropshipping operators bid on revenue. tROAS at 2.0x means revenue divided by spend equals 2.0. The problem is that revenue ROAS does not account for margin variance across the catalogue. A 60%-margin SKU at 2.0x revenue ROAS is profitable. A 30%-margin SKU at 2.0x revenue ROAS is breakeven. Same target, very different outcomes.

Value-based bidding fixes this by passing actual margin to Google Ads through enhanced conversions or offline conversion uploads. Smart Bidding then bids against margin. SKUs with 60% margin get bid more aggressively than 30%-margin SKUs even at the same revenue price.

The implementation:

1. Calculate margin per SKU. Include media spend, payment processing, supplier price drift.
2. Pass margin (not revenue) as the conversion value through enhanced conversions.
3. Set tROAS based on margin ROAS targets (for example, 2.5x margin ROAS instead of 2.0x revenue ROAS).
4. Let Smart Bidding tune for 30 days before adjusting.

Value-based bidding is the second-largest unlock after margin-tier asset groups in our MCC. Stores that switch from revenue to margin bidding consistently see ROAS appear to drop - the target unit changed - while profit absolute increases.

The Performance Max leaks that bleed dropshipping accounts

Specific patterns we see on accounts that come to us with PMax bleeding money.

Leak 1: out-of-stock SKUs in the auction. Smart Bidding spends on products it cannot ship. ROAS falls without explanation. Fix: real-time inventory sync from Shopify to Merchant Center. Auto-pause out-of-stock SKUs before they show in PMax.

Leak 2: branded traffic charged to PMax. PMax bids on your brand name and charges you for traffic Google would have routed to organic anyway. Fix: tight Search backstop campaign on branded queries. Set PMax negative match for the brand name where possible.

Leak 3: pixel-only tracking. Smart Bidding sees 60-70% of conversions because iOS Safari and ad blockers eat the rest. Bidding against incomplete data. CPA inflates without explanation. Fix: server-side tracking via Shopify webhooks (the ZenoX app ships this in one click).

Leak 4: catalogue-wide single asset group. Already covered above. Single biggest leak on dropshipping accounts.

Leak 5: creative drift on Champions group. Old assets that were good when uploaded but have rotated past their useful life. Fix: creative refresh quarterly. New product photography for top-spend SKUs annually.

Leak 6: tROAS set too aggressively for the margin. tROAS at 4.0x on a 30% margin product means Smart Bidding stops bidding because the math will not clear. Fix: realistic tROAS at 90% of margin breakeven, not aspirational.

Leak 7: campaign budget caps cycling. Daily budget capped, PMax exits the auction at 4pm, missing evening conversion volume. Fix: monthly budget allocation, not daily caps. Or move to Maximize conversion value with no target if conversion volume is high enough.

If your Performance Max account is bleeding money, it is almost certainly one of these seven. Read why Google Ads dropshipping fails for the broader failure modes.

What "good" Performance Max looks like on a dropshipping account

Specific benchmarks from our MCC across stable dropshipping accounts.

Champion asset group ROAS: 4.5-7.0x revenue ROAS. Margin ROAS 2.8-4.2x.

Waster / Sleeper asset group ROAS: 1.4-2.4x revenue ROAS at the tight floor. Margin ROAS 0.8-1.4x. Profitable but barely - this group exists to soak up impressions on cheaper inventory.

Blended ROAS at the campaign level: 3.2-4.8x revenue, 2.0-2.8x margin. This is the steady-state target.

Conversion volume per asset group: Champions hit 30+ conversions/week to keep Smart Bidding tuning cleanly. Wasters / Sleepers can hit 10+ conversions/week without breaking learning.

Spend split: 60-70% of campaign spend on Champions, 30-40% on Wasters / Sleepers. Smart Bidding allocates this naturally given the tROAS differential.

Audience signal performance: Customer Match audiences should outperform unsignalled traffic by 15-25% on Champions. If they do not, the customer list is too small or too stale.

The Performance Max plus Standard Shopping stack

Performance Max alone is not the answer for dropshipping. The full stack is PMax plus Standard Shopping plus Search plus Demand Gen, layered for distinct roles.

• PMax: broad demand capture across Shopping, Search, Display, YouTube, Discover, Gmail.
• Standard Shopping: branded queries and high-margin SKUs you want manual control on.
• Search: category backstops and brand defence.
• Demand Gen: top-of-funnel discovery on YouTube and Discover.

The full stack beats single-PMax setups by 17% ROAS at the same spend. Read Google Shopping Ads for dropshipping for the Shopping layer.

What to do if your Performance Max is broken

Three specific moves in order.

1. Audit the asset group structure. If you are running one asset group, split by margin tier. This alone is the single biggest unlock.
2. Audit tracking. If you are pixel-only, switch to server-side. Smart Bidding bids on real conversions instead of pixel ghosts.
3. Audit value passing. If you are bidding on revenue, switch to margin. tROAS targets recalibrate, profit absolute lifts.

If all three are clean and PMax is still bleeding, the leak is one of the seven listed above. Drop the store URL on WhatsApp and we will pull the Google Ads dropshipping account up live on a thirty-minute call. The full playbook is in the Google Ads dropshipping playbook from 200+ accounts. If you want to work through Performance Max structure questions with other operators while you figure it out, the free Google Ads community has 740+ operators who have run this exact split.