Date
Jan 8, 2025
Category
Agency
Reading Time
8 Min
Performance Max Not Performing? Here Are 4 Fixes That Actually Work
Fix #1: Your Creative Assets Are Having an Identity Crisis
Walk into any marketing meeting and ask what makes good creative. You'll get answers like "compelling visuals" and "clear value proposition."
That's not wrong, but it's not the whole story with Performance Max.
The real problem isn't having bad creative. It's having creative that doesn't work together. Google's testing thousands of combinations of your headlines, descriptions, and images. When your assets tell different stories, the algorithm gets confused and your performance tanks.
Six months ago, I inherited a Performance Max campaign for a company selling accounting software. Beautiful creative—professionally shot videos, clean graphics, compelling copy. They were spending $40K monthly and getting a 0.8x ROAS.
The issue? Half their assets talked about "saving time on bookkeeping" while the other half focused on "powerful financial reporting." Google's algorithm couldn't figure out if they were selling a time-saving tool or an analytics platform.
The Asset Cleanup Process That Actually Works
Start with the Asset Performance Report
Login to Google Ads, navigate to your Performance Max campaign, click on Asset Groups, then "View asset details." Pull the last 30 days of data.
Look for these red flags:
• Assets stuck on "Learning" after three weeks
• Headlines with click-through rates under 1.5%
• Images getting less than 15% of impression share
• Descriptions marked as "Low" performance
But don't just trust Google's ratings. Export the data and look at actual numbers.
The Message Alignment Test
Print every single asset you're using. Spread them on a table. If someone walked by without knowing your company, could they immediately understand what you sell and why they should care?
If there's confusion, you found your problem.
Rebuild Around Story Themes
Instead of random asset collections, create groups that tell complete stories:
Theme 1: Problem/Pain Headlines highlighting customer frustrations, images showing chaos or inefficiency, descriptions that make prospects nod and think "that's exactly my situation."
Theme 2: Solution/Benefit Headlines about outcomes, images of success or ease, descriptions focusing on life after using your product.
Theme 3: Proof/Trust Headlines with numbers or testimonials, images of real customers or results, descriptions with specific achievements.
For that accounting software client, we rebuilt everything around three themes: "Bookkeeping Hell" (problem), "Automated Simplicity" (solution), and "Real Results" (proof). Within three weeks, ROAS jumped to 3.2x.
The algorithm finally had coherent stories to work with instead of random creative soup.
Fix #2: You're Asking Google to Solve an Impossible Math Problem
Performance Max campaigns fail because they're trying to optimize for people at completely different stages of the buying journey simultaneously.
Think about it: someone searching for your company name is ready to buy. Someone who's never heard of you needs education first. Asking one campaign to optimize for both is like asking your GPS to find the fastest route to "somewhere good."
Split Your Campaigns by Customer Intent
Most advertisers think about targeting in terms of demographics. Wrong approach. Think about what people are trying to accomplish.
Campaign 1: Brand Defense (30% of budget) For people who already know your name. They're searching for you, checking out competitors, or browsing your category.
Target ROAS: 5-7x (these people are hot) Goal: Don't let competitors steal your customers
Campaign 2: New Customer Acquisition (50% of budget)
For people with your problem who don't know you exist. They're researching solutions or just realizing they have a problem.
Target ROAS: 2.5-4x (you're buying awareness and trust) Goal: Efficient growth at scale
Campaign 3: Win-Back (20% of budget) For people who visited your site, watched videos, or engaged with content but didn't convert.
Target ROAS: 6-12x (these should be your best converters) Goal: Convert warm traffic that already knows you
Audience Signals That Actually Guide the Algorithm
Most people dump every possible audience into their campaigns thinking more options help. It doesn't. Precision beats volume.
Brand Defense Audiences:
Customer email lists (exclude these from other campaigns)
Branded search term audiences from last 60 days
1% lookalike audiences based on highest-value customers
New Customer Audiences:
One demographic audience matching your best customers
One interest audience related to your solution (not broader categories)
In-market audiences for your specific product type
Win-Back Audiences:
Website visitors 1-30 days (excluding recent buyers)
Video viewers who watched 25%+ of your content
Email subscribers who haven't purchased recently
Last month, I split a massive $60K/month Performance Max campaign using this framework. The brand defense campaign converted at 9x ROAS while the acquisition campaign brought in new customers at 3.5x ROAS. Same total budget, 85% better overall performance.
Each campaign could focus on its job instead of trying to be everything to everyone.
Fix #3: Your Product Feed Is Sabotaging Your Success (eCommerce Must-Read)
If you're selling products online, your Google Merchant Center feed isn't just a catalog. It's the foundation Google's algorithm uses to understand what you're selling and who might want to buy it.
Most product feeds look like they were created by someone who actively hates customers. Product titles like "Shirt Blue Large" and descriptions that read like assembly instructions written by robots.
I audited an apparel brand burning $75K monthly on Performance Max with terrible results. Their feed had 15,000 products with titles averaging 4 words each. Google had no idea what they actually sold.
Product Information That Drives Performance
Titles That Actually Help Customers
Your product titles need to answer the question: "What exactly am I looking at and why should I care?"
Formula that works: [Brand] + [What It Is] + [Key Benefit] + [Important Details] + [Unique Element]
Instead of: "Nike Shoes Black 10" Use: "Nike Air Max 270 Men's Running Shoes Max Air Cushioning Size 10 Triple Black Limited Edition"
Yes, it's longer. Google's algorithm loves descriptive titles, and customers appreciate knowing exactly what they're buying.
Images That Convert in Tiny Thumbnails and Full Screen
Your primary image needs to work when it's the size of a postage stamp and when someone zooms in.
Requirements that matter:
1200x1200 minimum (not the 800x800 most guides suggest)
Product fills 75-85% of the frame
Clean background for primary image
Additional images show context, details, and scale
Product Attributes That Boost Rankings
Most sellers ignore optional attributes. Mistake. These help Google understand what you're selling and show your products to the right people.
Must-optimize attributes:
Product Type (use Google's taxonomy exactly)
Material (detailed, searchable descriptions)
Color (specific names, not "blue" but "navy blue" or "royal blue")
Size (include size charts and fit info)
Custom labels for bidding strategy:
Margin level (High/Medium/Low)
Performance tier (Bestseller/Standard/Clearance)
Seasonality (Year-round/Seasonal/Holiday)
Stock status (In-stock/Low-stock/Backorder)
That apparel client I mentioned? After optimizing their feed with detailed titles, better images, and proper attributes, their Performance Max ROAS went from 1.4x to 5.2x in two months. Same products, same prices, just better information for Google to work with.
Fix #4: You're Teaching Google to Optimize for the Wrong Things
This is the killer. Most Performance Max failures aren't about creative or targeting or budgets. They're about asking Google to optimize for actions that don't actually help your business.
Google's algorithm is like a very smart student who does exactly what you ask. If you ask for the wrong things, you'll get them.
Conversion Setup That Reflects Reality
The Conversion Audit
Go to Tools & Settings → Conversions. Look at everything marked "Primary."
Ask: "If this happened 100 times tomorrow, would my business be meaningfully better?"
If not, it's not a primary conversion.
Primary Conversions (What Actually Matters):
Purchases with real dollar values
Qualified sales leads
Phone calls over 90 seconds from money pages
Demo bookings or consultation requests
Secondary Conversions (Nice But Not Critical):
Email signups
Content downloads
Brochure requests
Tracking Only (Don't Optimize For These):
Page views
Time on site
Video plays
Add to cart actions
Conversion Values That Make Sense
Don't use fake values or averages. Use real business impact.
eCommerce: Pass actual purchase amounts, not average order values.
Lead Generation: Calculate average customer lifetime value, multiply by lead-to-customer rate. If customers are worth $8,000 on average and 15% of leads convert, each lead is worth $1,200.
B2B: Weight leads by quality. Demo requests are worth more than whitepaper downloads. Price accordingly.
Attribution That Reflects Your Customer Journey
Most accounts still use Last Click attribution, which gives all credit to the final touchpoint. That's like crediting a touchdown entirely to whoever carried the ball the last yard.
Data-Driven Attribution (best option if you qualify): Needs 15,000+ conversions and 600+ conversion paths in 30 days Automatically adjusts based on your actual customer behavior
Time Decay (good for longer sales cycles): Gives more credit to recent interactions Works well for B2B or high-consideration purchases
Last Click (only for specific cases): Direct response campaigns Very short sales cycles (under 3 days) Brand protection campaigns
Real-World Impact Story
Three months ago, I took over a Performance Max campaign for a software company optimizing for whitepaper downloads. They were getting 300+ "conversions" monthly but zero actual sales.
We changed the primary conversion to demo requests and set the value at $3,500 (based on their close rate and average deal size).
Results after the change:
Cost per demo dropped 55%
Lead quality improved dramatically
Monthly demos went from 8 to 32
Closed $240K in new business in the first quarter
Google's algorithm started finding people who actually wanted to see the product, not just grab free content.
Your 30-Day Recovery Plan
Week 1: Diagnosis
Run asset performance audit
Fix conversion tracking setup
Start product feed cleanup (eCommerce)
Document current performance baselines
Week 2: Surgery
Create new campaign structure by intent
Build coherent asset groups around story themes
Implement proper audience signals
Upload optimized product feed
Week 3: Optimization
Launch new campaigns with conservative budgets
Monitor performance and adjust spending
Replace weak assets with stronger versions
Fine-tune conversion values
Week 4: Scale
Increase budgets on winning campaigns
Pause or fix underperforming elements
Plan next optimization cycle
Document what worked
Success Benchmarks
Week 1-2:
80%+ assets rated "Good" or "Best"
Conversion tracking 95%+ accurate
Product feed processing without errors
Week 3-4:
30-60% ROAS improvement
25-40% lower cost per conversion
Better search term relevance
Month 2+:
Sustained month-over-month growth
Higher customer quality from campaigns
Predictable, scalable performance
The Bottom Line on Performance Max
Performance Max isn't broken. Most people just don't understand what they're actually controlling. You can't control where your ads show up or exactly who sees them. But you absolutely control the inputs: your creative assets, conversion definitions, product information, and campaign structure. Master the inputs, and the algorithm will master the outputs. The four fixes I've shared aren't just tactics for Performance Max. They're a framework for thinking about any automated advertising platform. Apply this approach to whatever Google releases next, and you'll be ahead of everyone still trying to figure out why their campaigns don't work. Stop treating Performance Max like a slot machine you feed money and hope for the best. Start treating it like the sophisticated optimization engine it actually is. Your ROAS depends on it.

Karan Naidu
Performance Marketing Head