Performance Max for DTC in 2026: The Structure That Keeps Margin Healthy

7 structural decisions that decide whether Performance Max prints money or burns it

Performance Max is the campaign type DTC founders love to hate. The reporting is thin, the targeting is a black box, and Google's default recommendations push spend toward whatever converts cheapest, which usually means your own brand searches and remarketing traffic you already owned. Every audit we run on a struggling PMax account finds the same root cause: the campaign was launched with default settings, a single asset group stuffed with every SKU, and no brand exclusions. Then the account manager spent six months tuning bids on a campaign type that does not accept bid adjustments.

The brands that get PMax right treat it like a structured system, not a magic button. They split product feeds by margin tier. They feed audience signals that actually bias the algorithm. They exclude brand traffic at the campaign level. They use conversion value rules to tell Google which customers are worth more. And they run PMax alongside Standard Shopping in a way that stops the two campaigns from bidding against each other.

This guide walks through the structure we use on DTC accounts spending between fifteen thousand and four hundred thousand a month on Google. If you are picking between campaign types in the first place, start with Google Shopping vs Performance Max, then come back here for the implementation.

TL;DR

-> Split your product feed into three PLG tiers by margin (hero, core, tail) and run one asset group per tier, not one asset group per product category. -> Feed audience signals that are small, dense, and high-intent (past 30-day purchasers, add-to-cart non-buyers, high-AOV customer match list) instead of broad interest audiences. -> Exclude brand search at the account level using a brand list, and run a separate Standard Shopping campaign for hero SKUs if you want inventory-level control. -> Use conversion value rules to upweight new customers by 25 to 40 percent and downweight repeat purchasers of low-margin SKUs, so Google's bidding reflects your actual LTV math.

When PMax wins and when it loses

Performance Max wins when you have enough conversion volume for the algorithm to learn, a product feed that is already clean, and creative assets good enough to compete on YouTube and Discover. The rough threshold we use: at least 30 conversions per campaign in the past 30 days, a Merchant Center feed with zero disapprovals, and at least 10 video assets per asset group. If any of those three is missing, PMax will underperform a well-tuned Standard Shopping campaign every time.

PMax loses when the account is small, the feed is messy, or the brand is new and has no remarketing pool to draw on. We have seen sub-10k/month accounts waste eight weeks on PMax before switching to Standard Shopping and seeing ROAS double in a fortnight. The learning phase is real, and it is expensive if your daily budget is 100 and the campaign needs 30 conversions to exit learning.

The other losing case is when the account has a strong brand search footprint. Without explicit brand exclusion, PMax will spend 40 to 60 percent of its budget on people searching your brand name, then claim credit for conversions that would have happened anyway. This is the single most common failure mode we see, and it makes PMax look like it works when it is actually just stealing credit from organic brand search.

If you are running a mid-size DTC account (50k to 300k/month) with clean feed data and a healthy remarketing pool, PMax should be 40 to 60 percent of your Google spend. Our paid ads service is built around this split.

Campaign and asset group structure

The default Google rep suggestion is to create one PMax campaign and drop every SKU into a single asset group. Do not do this. Asset groups are the only lever you have for creative relevance, and stuffing 800 SKUs into one asset group means your headlines, descriptions, and images are averaged across products that have nothing in common. The algorithm cannot learn which creative works for which product.

The structure we use for most DTC accounts:

One campaign per goal. If you are running new customer acquisition and repeat purchase retargeting, those are two campaigns. Not two asset groups inside one campaign. The conversion goals, budgets, and bid strategies are different, and PMax optimizes at the campaign level.

One asset group per product cluster. A cluster is a group of SKUs that share creative. If you sell skincare and supplements, those are two clusters because the imagery, claims, and testimonials do not transfer. If you sell five flavors of a single protein powder, those are one cluster because the creative is nearly identical.

Budget sized to the learning phase. Divide your daily budget across asset groups so each one can hit 30 conversions in 30 days. If an asset group cannot support that volume, merge it with another cluster. Below the learning threshold, PMax guesses.

We usually end up with three to six asset groups for a single-category DTC brand, and eight to twelve for a multi-category brand. More than that and you are under-feeding the algorithm. Fewer than that and you are losing creative specificity.

Audience signals that actually bias the algorithm

Google describes audience signals as "suggestions" and insists PMax will target outside them. This is technically true and practically misleading. The quality of your audience signals has a massive effect on who PMax targets, especially in the first 30 days of a campaign. Weak signals produce weak targeting.

The signals we feed, in order of effectiveness:

Customer match lists, segmented. Upload three lists: all purchasers in the last 180 days, high-AOV purchasers (top 20 percent by order value), and repeat purchasers (two or more orders). The algorithm treats these as seed audiences and finds lookalikes. A single undifferentiated "all customers" list is worth about 20 percent of what three segmented lists are worth.

Website visitor segments with intent scoring. Not "all site visitors." Specifically: add-to-cart non-buyers in the last 30 days, product detail page viewers with two or more sessions, and checkout abandoners. These segments are small and dense, which is what the algorithm needs.

Search term lists, narrow. Google accepts a list of search terms as a signal. Feed it the 15 to 25 highest-intent non-brand terms from your Search campaigns, not a generic keyword dump. If your Search campaigns are not segmented by intent yet, fix that first.

Skip the "interests and detailed demographics" options unless you have tested them and seen lift. In most DTC accounts they dilute the signal rather than strengthen it.

Product listing groups by margin tier: the 3-tier PLG split

This is the framework that does the heaviest lifting in a PMax account. Most DTC brands treat every SKU as equally important and let Google decide which products to push. This is a mistake because Google optimizes for conversion volume, not margin, and your lowest-priced SKU is almost always the easiest to convert. Without PLG segmentation, PMax will sell more of your worst product and less of your best.

The 3-tier PLG split:

Hero tier. Your three to eight best products by contribution margin. High AOV, healthy margin, strong reviews, evergreen demand. These get their own asset group with their own creative, their own audience signals, and the majority of the budget. Target ROAS set lower than the campaign average because these products justify aggressive bidding.

Core tier. Your next 20 to 50 products. Decent margin, consistent demand, not the flagship but the bread and butter. One asset group, grouped by creative cluster. Target ROAS set at the campaign average.

Tail tier. Everything else. Long-tail SKUs, seasonal items, clearance products. One asset group with a higher target ROAS so Google does not overspend on low-margin sales. If a tail SKU starts selling well, promote it to core. If a core SKU underperforms for 60 days, demote it to tail.

You implement the split inside PMax using the "listing groups" section of each asset group, which lets you include or exclude specific products. Alternatively, use custom labels in your Merchant Center feed (custom_label_0 = margin_tier) and filter at the asset group level. The custom label approach is cleaner because the feed becomes the single source of truth. The same custom label structure applies to your Standard Shopping campaigns, and our Google Shopping optimization guide walks through the feed work in detail.

Review the tier assignments monthly. Product performance drifts, new SKUs launch, seasonal items come and go. A PLG structure that was correct in January is usually wrong by April.

Asset specs: what to upload and how many

The "minimum" column is what Google requires to serve. The "recommended" column is what you need for the algorithm to A/B test creative meaningfully. Uploading only the minimum is the creative equivalent of running a one-variant landing page test and calling it optimization. If your landing pages are also a bottleneck, the CRO service work usually pays for itself inside a quarter.

Brand exclusions and cannibalization with Standard Shopping

The two biggest wasted-spend problems in PMax accounts are brand cannibalization and Standard Shopping overlap. Both are solvable. Most accounts have not solved them.

Brand exclusion. Google added account-level brand lists in 2023 and improved them in 2024. Create a brand list that includes your brand name, common misspellings, and any sub-brands. Apply it as a negative at the campaign level. This stops PMax from bidding on brand search terms, which means your brand search budget stays in your dedicated brand Search campaign where you can measure it and control CPC.

Without this exclusion, PMax will eat 30 to 60 percent of your brand search volume at CPCs 2 to 4 times higher than what your brand Search campaign would pay, and the reporting will make it look like PMax is generating incremental revenue when it is cannibalizing existing demand. Turn the exclusion on. Watch ROAS drop and incrementality rise. The drop is Google admitting it was stealing credit.

Standard Shopping overlap. If you run both PMax and Standard Shopping on the same SKUs, PMax wins the auction by default because it has more placements available. Two ways to handle this:

The first approach is running PMax on all products and using Standard Shopping only for hero SKUs with manual bid control. In this case, set Standard Shopping to a higher priority in the campaign settings (High) and PMax to Low. When both campaigns are eligible to serve, Standard Shopping takes the impression. This gives you manual control over hero SKU bids while letting PMax handle core and tail.

The second approach is excluding hero SKUs from PMax entirely and running them only in Standard Shopping, while PMax handles core and tail. This is cleaner reporting but removes PMax's ability to place hero SKUs in YouTube and Discover inventory. Most DTC brands are better off with the first approach.

The wrong approach is running both campaigns on the same products without a priority structure. That just means you are paying Google twice to serve the same ad.

Conversion value rules: telling Google what a customer is worth

Conversion value rules are underused and they should not be. They let you adjust the conversion value that PMax optimizes against, based on audience, location, or device. This matters because PMax bids on value, and if you do not tell it what different customers are worth, it assumes they are all worth the same.

The three rules we set on nearly every DTC PMax campaign:

New customer uplift. Create a rule that upweights conversions from new customers by 25 to 40 percent. You need GA4 new-vs-returning segmentation or a customer match exclusion list (existing customers) to power this rule. The uplift tells PMax that a 100-dollar order from a new customer is worth more than a 100-dollar order from an existing customer, which is correct if your LTV math includes repeat purchase value.

Geo weighting. If your margin varies by shipping zone (domestic vs international, or by state for tax/shipping-cost reasons), create rules that reflect those margin differences. A 10 percent downweight on high-shipping-cost zones is often enough to rebalance.

Repeat purchaser downweight on low-margin SKUs. If your tail tier includes consumable refills that repeat customers buy frequently, downweight those conversions by 15 to 20 percent. This stops PMax from spending aggressively to hit repeat purchasers who were going to buy anyway.

Conversion value rules layer on top of your tROAS target. They do not replace it. Set tROAS based on your blended goal, then use value rules to tell PMax which conversions inside that goal are worth more.

Weekly actions

-> Check the asset group performance report on Monday, pause asset groups with ROAS below 60 percent of target for 14 days running. -> Refresh one asset group's creative every two weeks on a rolling schedule: new headline variants, one new video, two new images. -> Review search terms insights report, add negative terms at the account level for any irrelevant queries PMax is buying. -> Audit Merchant Center for disapprovals and warnings every Friday, a single disapproved product can drag an entire asset group. -> Compare PMax ROAS vs Standard Shopping ROAS on shared SKUs monthly, rebalance priority if one campaign is consistently underperforming.

FAQ

How long is the PMax learning phase? Google says two weeks. In practice, plan for four to six weeks before the campaign's performance stabilizes, and do not make major changes (budget shifts over 20 percent, asset group restructures, tROAS changes over 15 percent) during that period. Each major change restarts learning.

Should I run PMax if I am spending less than 10k/month on Google? Usually no. Below that threshold, Standard Shopping with manual bids or a tight Search campaign structure will outperform PMax because the algorithm does not have enough conversion volume to learn. Revisit PMax when you cross 15k/month and have at least 30 conversions per week.

How does PMax compare to Meta Advantage+ Shopping? They are structurally similar (black box, algorithm-led, minimal manual control) but solve different intent problems. PMax catches demand across Google's network including Search, Shopping, YouTube, and Discover. Advantage+ is paid social, lower intent, better for cold acquisition. Most DTC brands run both, and our Meta ads for DTC guide covers the Advantage+ structure in detail.

Can I see which channel PMax spend went to? Partially. The "insights" tab shows channel breakdowns (Search vs Shopping vs YouTube vs Display vs Discover) at the campaign level. You cannot see it at the asset group or product level, which is one of PMax's real limitations. If channel-level attribution matters to you, run Standard Shopping and separate YouTube campaigns alongside PMax.

What tROAS should I start with? Start without a tROAS target for the first two to three weeks, let PMax find baseline performance, then set tROAS at 85 to 90 percent of the baseline you observed. Setting tROAS too high at launch means the campaign will not spend. Setting it too low means it will spend without pressure to be efficient. Both kill the campaign.