Shopify Marketing · PMax Diagnostic
Back to Shopify marketing guides
Quick answer
A Shopify Performance Max campaign is working when new customer acquisition grows with spend and ROAS holds after brand search is isolated. Most reported PMax ROAS figures include branded traffic, proxy conversion events, and retargeting revenue the campaign did not create. Five checks separate real growth from demand farming: isolate brand, audit asset groups, verify the feed, audit conversion events, and read new customer rate against spend.
Key takeaways
What this article covers
A Shopify owner opens Google Ads, sees a PMax campaign reporting 6.4x ROAS, and assumes the campaign is working. The reported number is real. The interpretation is not. Across twenty years of running paid media for Shopify stores, the same pattern repeats: PMax reports a strong ROAS, revenue looks fine, the owner keeps the budget live, and six months later the store has not grown. The dashboard is telling the truth about what happened. It is not telling the truth about what caused it. This guide is the five-check sequence I use to separate a PMax campaign that is actually acquiring customers from one that is recycling demand the store already had. For the full pillar, see the Shopify marketing guides collection.
PMax is a black box campaign that serves across Search, Shopping, Display, YouTube, Gmail, and Discover with a single bid strategy. Google's machine learning decides where each impression goes, and the reported ROAS blends all of those placements into one number. The problem is not the blend. The problem is that the highest-converting placement, branded search, gets mixed with the lowest-converting, cold display, and the average looks healthy while the acquisition work is being done by someone's existing brand equity.
The reasons a Shopify PMax ROAS figure typically overstates reality:
Reading a PMax ROAS without accounting for those five adjustments is reading a report that averages three different campaigns into one number.
Every Shopify store with any organic awareness has branded search demand: people typing the store name into Google. When no dedicated brand search campaign exists, PMax captures those queries by default. The branded clicks convert at 10 to 20 times the rate of cold traffic, often above 15 percent purchase rate, and they land in the PMax performance report as if the campaign produced them.
The fix is structural, not optimizational:
In most accounts I audit, isolated PMax ROAS drops by 30 to 60 percent once brand is removed. That drop is not performance degradation. It is the first time the campaign's real cold-acquisition performance has been visible. Decisions made against the blended number were made against a fiction.
PMax uses asset groups the way Search campaigns use ad groups. Each asset group is a container of product listings, text assets, images, videos, and audience signals that the algorithm treats as a coherent unit of intent. When one asset group covers the entire catalog, the algorithm has nothing to learn from. It cannot tell which audience signal applies to which product, which creative matches which buyer, or which margin tier to prioritize.
The segmentation rules I apply when restructuring a Shopify PMax build:
A Shopify store with 400 SKUs and one asset group is running PMax at maybe 40 percent of its potential. Splitting that into four to six asset groups by margin and category, with distinct audience signals, is often the single largest fix in a PMax account.
PMax serves Shopping placements, and Shopping placements require a healthy Merchant Center feed. The Shopify-Google app creates the feed automatically, which is both useful and dangerous. Useful because the feed exists. Dangerous because Shopify's product data rarely matches Merchant Center's expectations out of the box. Missing GTINs, incorrect Google product categories, and title fields that exceed length limits all suppress impressions without generating a visible error in Google Ads.
The feed checks that surface the most suppressed reach:
A store with 15 percent of SKUs disapproved is losing 15 percent of possible PMax reach before any auction decision is made. Fixing the feed is not marketing work. It is infrastructure. PMax cannot serve what Merchant Center will not approve.
PMax optimizes toward whatever conversion action is marked Primary in Google Ads. In most Shopify accounts, the Shopify-Google app has set up multiple conversion actions and marked several of them as Primary: Purchase, Begin Checkout, Add to Cart, sometimes View Content. The algorithm counts every Primary action equally. If Add to Cart is Primary, PMax learns to find people who add items to cart. Whether they buy is a separate question the algorithm is not solving for.
The correct Shopify PMax conversion setup:
When proxy events share Primary status with Purchase, the campaign's reported conversion volume looks strong while actual revenue stalls. Fixing the conversion hierarchy often drops reported conversion count by 40 to 70 percent and increases real revenue over the following 30 days.
The final check is the one that settles the question: is PMax growing the business. ROAS can be high while the business is flat, because returning customers convert easily and PMax retargeting catches them cheaply. New customer acquisition rate is the metric that separates acquisition from retention dressed up as acquisition.
The procedure, done in Shopify not in Google Ads:
A PMax campaign that is working shows new customer count rising with spend and new customer cost trending flat or down over time. A campaign that looks strong on ROAS while new customer count is flat is a retention system running on the ad account's budget. That is a strategic decision to make consciously. Made unconsciously, it caps the store's ceiling at the size of its existing audience.
The framework
Build a dedicated brand search campaign with exact and phrase match. Add brand terms as account-level negatives in PMax. Wait 14 days. Re-read PMax ROAS against the new baseline before making any further decisions.
Open the PMax campaign. Count asset groups. If one asset group covers more than 150 SKUs or spans categories, the segmentation is the problem. Split by margin tier and category. Populate audience signals per asset group.
In Merchant Center, filter by disapproved and limited. Read the reason codes. Check GTIN coverage, Google product category accuracy, and title structure. Fix approval issues before any campaign optimization work.
Tools, Conversions. Set Purchase as the only Primary. Move Add to Cart, Begin Checkout, and View Content to Secondary. Confirm revenue value passes through. Enable enhanced conversions.
Pull new customers per month for six months from Shopify. Pull total paid spend for the same months. Plot the trend. If new customer count is flat while spend rises, PMax is farming demand, not acquiring.
No. Without a brand search campaign to capture branded queries at a controlled cost, PMax absorbs the branded traffic and reports it as campaign-driven revenue. The ROAS figure looks strong because branded queries convert at 10 to 20 times cold traffic. Running brand search separately makes PMax's real non-brand performance visible for the first time.
There is no universal number. The rule is products grouped by margin tier, category, or price band that share a conversion pattern. A hundred SKUs with similar margins and buyer intent can share an asset group. Five hundred SKUs spanning categories and price ranges cannot. If one asset group covers the whole catalog, segmentation is the first fix.
Purchase with revenue value passed through, set as the only primary conversion action. Add to Cart, Begin Checkout, and View Content should be set to Secondary if they are present at all. When proxy events share equal weight with Purchase, the algorithm optimizes toward whichever is easiest to produce, which is never the one that generates revenue.
For the feed connection and basic conversion events, yes. For running PMax correctly, no. The app does not set up asset group segmentation, audience signals, brand exclusions, or conversion action hierarchy. It gives you a running campaign. Whether the campaign is optimizing toward the right outcome is a separate setup entirely.
When the checks surface problems but not the order to fix them, when ROAS and new customer rate contradict each other, or when the campaign runs across multiple markets with different conversion setups. The Conversion Second Opinion exists for that moment. Written findings, prioritized fix list, 72 hours, $999, no retainer.
A PMax campaign is not a single lever. It is a stack: the feed underneath, the asset groups on top, the audience signals inside them, the conversion action hierarchy in Google Ads, and the brand campaign next to it. Diagnosing PMax means reading the stack, not the top-line ROAS. Most of the reported performance on a Shopify PMax account is determined by three decisions made during setup, and none of those three are in the PMax interface itself.
The single highest-leverage check is the brand isolation one. Every other problem in a PMax setup can hide behind an inflated branded ROAS, because a 9x reported number ends the conversation before any diagnostic question gets asked. Build the brand campaign, add the negatives, wait two weeks, and read the real number. Everything else becomes visible from there.
When the diagnosis surfaces more work than a store owner can execute between other responsibilities, or when the fixes interact in an order that depends on the specific account, that is where a practitioner belongs. Stan Consulting offers Shopify marketing and PPC management once the diagnostic is complete, and the Conversion Second Opinion is the entry point for anyone who wants the findings before the management.
Related: the full marketing guides collection covers Google Ads, conversion, strategy, and agency management.
The engagement format
$5,000. One engagement. Diagnosis, build, and fix. No retainer after.
See the Revenue Sprint