Performance Max for Online Stores: Setup, Asset Groups, ROAS
Performance Max is the AI-driven Google Ads format that captured nearly all Shopping traffic in 2024-2026 and displaced Smart Shopping. A properly configured Performance Max for an online store drives the bulk of paid Google revenue; a misconfigured one burns budget. Below is the UPLIFY operational playbook for e-commerce: prerequisites, asset group structure, audience signals, target ROAS optimization, troubleshooting performance drops.
Definitions
Performance Max (PMax) — Google Ads' AI-driven format that automatically distributes budget across Search, Shopping, Display, YouTube, Discover, and Gmail. One campaign layer instead of six separate ones. Launched in 2021, became the default for Shopping in 2024.
Asset group — a product grouping inside a PMax campaign. Contains a set of assets (text, images, video), audience signals, listing groups (SKU selection from the feed). One asset group = one AI optimization strategy.
Audience signals — hints to the algorithm: your customers, similar audiences, interests, custom segments. The AI uses signals for accelerated ramp-up and can later expand beyond the seeded segments.
Listing groups — product feed filter for a specific asset group. Lets you create separate groups for top categories, margin-heavy SKUs, or all products.
Target ROAS bidding — strategy where you set the target return on ad spend (e.g., 4× = 400%), and the algorithm optimizes bids automatically against it.
Prerequisites before launching PMax
Setting up Performance Max for an online store doesn't work on an empty campaign. Prerequisites:
Google Merchant Center with passed moderation. All required product attributes — id, title, price, availability, image_link, brand, gtin or identifier_exists=false — must be populated and approved by Google in Diagnostics. Disapproved SKUs won't show in PMax. Detailed checklist in our GMC setup playbook.
Linking GMC ↔ Google Ads. In Google Ads → Tools → Data manager → Google Merchant Center (or in Merchant Center → Settings → Access and services → Apps and services) the link must be active. Without it the Shopping feed is unavailable to the bidding algorithm. The old "Linked Accounts" path is deprecated — Google renamed the UI flow.
Value-based conversion tracking. Not "purchase=1". You need to pass actual order value (revenue or profit). Without this PMax optimizes for conversion count, ignores margin, and may push traffic to cheap SKUs.
Customer Match or similar audiences. Audience signals need a customer list or lookalike. Google's stated minimum is 100 records; for serving eligibility, 100 active users. Practically, for meaningful reach, upload 1,000+ hashed contacts (this is a UPLIFY heuristic, not a Google rule).
Budget that covers the learning phase. Google documents a learning period of up to ~14 days for PMax bidding strategies. For value-based bidding, Google doesn't publish a hard "N conversions" threshold, but practically ~30+ quality conversions during the learning window is a working benchmark (UPLIFY heuristic, not a Google rule). If the daily budget produces 1-2 conversions, the algorithm doesn't get enough signal and stalls.
Asset group structure
One asset group, one strategy. The most common mistake is dumping the whole feed into a single asset group and waiting for the AI to figure it out. It won't — or it will, slowly and expensively. Google officially recommends splitting asset groups when margin, budget, ROAS targets, or business objectives differ materially; otherwise consolidate so the algorithm gets more signal per group.
A working segmentation for a typical online store with $30k+/mo revenue (one approach — adapt to your margin structure):
Asset group 1: top category with the highest margin. Listing group = only SKUs from the top 3 categories that drive 60-70% of profit. Target ROAS above average (e.g., 5× if your average is 4×). Audience signals: your buyers + 90-day cart adders.
Asset group 2: new arrivals and high-intent products. Listing group = SKUs tagged "new" or with high customer satisfaction. Target ROAS — average. Audience signals: Google in-market segments for your category + similar to converters.
Asset group 3: clearance / low-margin products. Listing group = SKUs on discount or with current overstock. Target ROAS — lower (e.g., 3×) so the algorithm is willing to show more aggressively. Audience signals: broader interests.
For catalogs above 10K SKUs, prefer campaign-level structure: separate PMax campaigns per key category instead of asset groups. The size of your top-margin categories dictates which approach makes more sense.
Audience signals — hinting to the algorithm
Audience signals are not targeting. They are a signal to the AI about who to consider first. PMax expands beyond signals when it finds more effective segments, but the initial scope is shaped by them.
What to add:
Customer Match — list of your customers (hashed emails or phone numbers). Strongest signal because these are people who already purchased. Feed them for lookalike modeling.
Website visitors — 30/90/180-day remarketing audience. People who visited but didn't buy. PMax will remind them first, then expand.
Add-to-cart abandoners — separate segment from GA4 (custom audience setup required). Very strong intent signal.
Similar audiences — Google automatically builds lookalikes from your Customer Match. Turn them on.
In-market segments — Google's pre-built audiences by category ("Electronics → Security → Surveillance cameras"). Useful for new categories without your own data.
Custom segments — built from competitor search queries or competitor category URLs. Helpful in hot niches.
Avoid:
- Demographic targeting on age/
genderonly — too broad, the AI gets no useful signal. - Geo-targeting only major cities if you ship country-wide. Use regional split only if geography materially affects conversion.
Conversion tracking — value-based, not lead
Without proper conversion tracking, PMax is blind. This is the most common root cause of underperformance in e-commerce accounts.
Purchase via GA4 + Google Ads conversion. Configure order-value transmission (revenue) to GA4, then import conversions to Google Ads. Alternatively — direct Enhanced Conversions transmission via Google Tag Manager.
Pass profit, not revenue. If your products have large margin spread, pass profit margin (revenue × margin). Otherwise PMax may favor expensive low-margin products.
Customer lifetime value — for businesses with repeat purchases. Pass LTV instead of one-off revenue. PMax then optimizes for customers with higher long-term value.
Check conversion duplication. Google Ads + GTM can double-count conversions if both paths are configured. Pick one official path — either via GTM or via GA4 import.
Optimization against target ROAS
Base strategy — Maximize Conversion Value with Target ROAS. Parameters:
- Target ROAS — UPLIFY starter heuristic: breakeven × 1.3-1.5 (if breakeven is 3×, start at 4-4.5×). This is not a Google rule — Google recommends relying on historical conversion value / cost. Our heuristic is safe for new accounts; calibrate from actual data afterward.
- Don't change target ROAS more than once every 2 weeks. PMax needs time to adapt. Frequent changes cause learning loops without converging.
- Budget 20-30% above your average daily cap. PMax dislikes hard limits; give it room to optimize.
When to raise target ROAS? When PMax stays above the current target for 4 consecutive weeks. Raise by 0.5-1× steps, no more.
When to lower? When volume drops persistently and cost-per-acquisition is below your cap. The algorithm is under-showing because of an overly high ROAS target, and you're leaving traffic to competitors.
Troubleshooting performance drops
PMax can suddenly lose efficiency. Most common causes:
Item disapprovals in GMC. Check Diagnostics. If new disapproved products appeared in the last 7 days, part of the catalog dropped out of serving and ROAS fell from composition change. Fix the feed, wait for Approve.
Product feed changes. Bulk updates to titles, prices, availability can trigger algorithm re-evaluation. PMax "forgets" historical data. Wait 2-3 weeks for stabilization.
Competitor lowered prices. Check Auction Insights — if impression share dropped and your brand isn't showing where it used to, competitor pricing/terms became more attractive. Adjust accordingly.
Search/Shopping SERP changes. Google regularly changes SERP layout. AI Overviews sometimes pushes organic out, which shifts traffic composition. If impressions drop across all channels, it's not your fault — it's Google.
Audience signal removed from GA4. Common issue — custom audiences became stale. Verify they're active and populating.
Learning phase reset. Any major change (budget +50%, target ROAS +1×, new asset group) resets learning. Wait 14 days before evaluating.
What you get after PMax launch with UPLIFY
Technical artifacts:
- Configured GMC + linking with Google Ads
- 2-4 PMax campaigns with asset groups matching your catalog structure
- Customer Match + similar audiences uploaded and activated
- Enhanced Conversions with value-based transmission
- Custom segments for your niche
- Listing groups with filters for top categories
Process artifacts:
- Google Ads dashboard with key metrics (ROAS by asset group, by category, top-performing SKUs)
- Weekly report with analysis + recommendations
- Monthly strategic call discussing budget alternatives
Our video-surveillance store case shows the real potential of this methodology: ROAS 12.25× across 10.98M impressions and 1,734 purchases. For the kids/baby goods e-shop — ROAS 16.19× against a 4× target.
How UPLIFY differs
Platform expertise: we are a Horoshop Expert Partner and know native feed integrations (no middleware). For Shopify, WooCommerce, custom CMS — we have deployment workflows in place.
Our own Wikidata entity (Q139583119) and founder Person entity (Q139583163) — not marketing. Entity consistency in the Knowledge Graph influences signal quality for Google when evaluating bidder behavior and product associations.
UPLIFY Content — our AI-SEO product for Horoshop stores with catalogs of 100K+ products. Auto-generation of title, description, google_product_category to GMC requirements + AI-citable structure — the same stack we apply for self-service feed optimization.
Real value-based attribution — we pass profit margin, not revenue, to Google Ads. This is a principled difference from agencies that ship default GA4 tracking unchanged.
UPLIFY is a Ukraine-based, remote-first AI marketing agency working with e-commerce clients globally.
Budget and engagement format
PMax setup is project work. Stages:
- Google Ads + GMC audit — diagnostic of current state, disapproval detection, conversion tracking evaluation. Separate phase before setup.
- PMax setup from scratch — campaign creation, asset groups, audience signals, bidding configuration. 2-3 weeks.
- Migration from Smart Shopping — phased transfer from Smart Shopping to PMax without traffic loss.
- Ongoing management — monthly optimization: target ROAS adjustments, listing groups refinement, audience signal updates, troubleshooting.
Exact budget estimate — after audit. Depends on:
- Catalog size (10K vs 100K SKUs)
- Number of priority categories
- Current state of GMC and tracking
- Whether you have an in-house content team for assets
Request an audit — we run the first audit within a fixed timeframe and return a concrete spec with estimate.
Common objections
"Smart Shopping already works for us — why PMax?" Google launched Performance Max globally in November 2021 and made it the default Shopping format in 2024. Smart Shopping was fully migrated to Performance Max during 2022-2023 per Google's product communication; classic Smart Shopping campaigns are no longer launched in 2026. Feature comparison:
| Dimension | Smart Shopping (deprecated 2022-2023) | Performance Max (current) |
|---|---|---|
| Status in 2026 | Cannot be created; auto-migrated | Default for retail Shopping |
| Inventory channels | Shopping + Display | Search + Shopping + Display + YouTube + Discover + Gmail |
| Bidding | Maximize Conversion Value | + Target ROAS, Maximize Conversions |
| Asset structure | Single feed group | Multiple asset groups + listing filters |
| Audience signals | Auto only | Customer Match, lookalike, custom segments, in-market |
| Negative keywords | Not supported | Campaign-level UI (since January 23, 2025) |
| Reporting | Basic | Asset group level + insights tab |
If you have an active Smart Shopping campaign, it's either a legacy state requiring migration or already an auto-converted PMax. The question isn't "why" — it's "how to migrate correctly".
"PMax is a black box — I have no control" Partly true. Google's AI controls tactical decisions (which audience to show, which creative to pull, which bid). You control strategy: target ROAS, listing groups (which SKUs to include), audience signals (who to consider first), assets (the content the AI uses). Transparency is limited, but levers exist.
"Performance Max will eat my Search campaign" Depends on setup. If you run an active Search campaign on branded queries, PMax can compete for the same audience. Solution — add campaign-level negative keywords in PMax (available directly in the Google Ads UI for Search + Shopping inventory since 2025; historically required a support ticket) or configure campaign priorities in GMC.
"We already work with an agency — no budget for a second one" Not a second agency. An audit is a standalone product. It shows what's working, what isn't, and what you're losing. Then you decide — stay with your current vendor and give them our report as a spec, change vendors, or work with us directly.
UPLIFY perspective
Performance Max in 2026 is not "a new Google Ads campaign" — it's a new operating model for e-commerce advertising. Stores still on Smart Shopping or classic Shopping are gradually losing impressions and share-of-voice — not sharply, gradiently. Without migration to PMax, a significant share of Shopping traffic drifts to competitors.
What works: configured GMC + structured asset groups + value-based tracking + audience signals built from real data + disciplined optimization against target ROAS. Not a "magic" combination — it's the standard playbook we apply to every client.
What doesn't work: one massive asset group with all SKUs, target ROAS set by gut, audience signals only on age/gender, conversion tracking on "purchase=1" without revenue. These mistakes repeat across the majority of audited accounts — it's the most common set of root causes.
Action checklist
- Minimum kit for self-serve Performance Max for online store setup:
FAQ
What is Performance Max in marketing?
Performance Max (PMax) is Google Ads' AI-driven campaign type that automatically distributes budget across every Google inventory channel — Search, Shopping, Display, YouTube, Discover, and Gmail — from a single campaign layer. Instead of running 6 separate campaigns, you set a single target ROAS, upload assets (text, images, video), define audience signals as hints, and let Google's AI decide where and when to show ads. PMax launched globally in November 2021 and became the default Shopping format in 2024, fully replacing Smart Shopping campaigns.
Can I launch Performance Max without Google Merchant Center?
Yes, but without a product feed. That means PMax will only run lead-generation or site-traffic objectives, without Shopping placements. For an e-commerce store, this loses most of the point — Shopping now drives the bulk of revenue through PMax.
How long does it take to reach target ROAS in PMax?
Depends on catalog and competition. For a typical store with $30-50K/mo — 4-6 weeks from launch to stable ROAS. First conversions — within 7-10 days. In the first weeks, the AI is learning from your signals and data, and ROAS can fluctuate during this phase.
Will Performance Max cannibalize my other Google Ads campaigns?
It can. PMax aggressively competes for traffic, especially branded. Set campaign priorities in GMC (Shopping campaigns have priority over PMax for specific SKUs), or add campaign-level negative keywords in PMax — available directly in the Google Ads UI since 2025.
What should I do if I still have an active Smart Shopping campaign?
Smart Shopping was fully migrated to Performance Max in 2022-2023. If you still see Smart Shopping in your account, it's either a historical record or stale UI. Creating a new Smart Shopping campaign isn't possible anymore. Verify in Google Ads whether the current campaign was auto-converted by Google to PMax (this ran in the background for all active Smart Shopping campaigns).
How often should I update asset groups?
Every 2-3 months — re-segment listing groups, add new arrivals into separate groups, refresh custom audiences. Every major catalog update (new category, renames) — re-evaluate listing groups right away.
Do I need a separate PMax campaign for each country?
Yes, if you sell in multiple countries. Bidding strategies, feeds, and competitive dynamics differ. Multi-country in one PMax means losing control and inefficient budget allocation.
How can I measure whether PMax is eating my Search campaign?
Auction Insights shows overlap with your own campaigns. If your Search loses impression share while PMax grows, that's cannibalization. Another sign — total impressions drop at the same budget.
Sources
- [1]Performance Max: overview — Google, tier 1
- [2]Asset groups in Performance Max — Google, tier 1
- [3]Audience signals in Performance Max — Google, tier 1
- [4]Enhanced Conversions setup — Google, tier 1
- [5]Target ROAS bidding — Google, tier 1
Need help?
UPLIFY builds the full Performance Max stack for e-commerce: from GMC + feed audit to launching PMax campaigns with value-based tracking and audience signals. We start with an audit: you get a concrete list of issues in current campaigns, a recovery plan, and a launch roadmap.
Request an audit — a UPLIFY operator responds within one business day.
Reviewed by Viacheslav Overkovskyi, founder of UPLIFY · Wikidata Q139583163