Google Ads Smart Bidding: Target ROAS vs Target CPA vs Maximize Conversions (2026 Guide)
Smart bidding works β but only when set up correctly and with enough data. Here's when to use Target ROAS, Target CPA, or Maximize Conversions, and the common mistakes that make smart bidding underperform.

Smart bidding uses machine learning to optimize bids for each individual auction β adjusting for device, location, time of day, user behavior, and dozens of other signals simultaneously. Manual bidding can't compete with that level of granularity. But smart bidding is only as good as your data and your target setting.
The Data Requirement: What Nobody Tells You
The most common reason smart bidding underperforms: not enough conversion data. Google's recommendation is 30 conversions per campaign per month minimum. Below that threshold, the algorithm is guessing. Too little data means it may overspend chasing rare conversions or underbid on opportunities it can't recognize as valuable.
- Under 30 conversions/month β Use Maximize Conversions (no target) or Manual CPC
- 30-100 conversions/month β Start with Target CPA (conservative target: your current actual CPA)
- 100+ conversions/month β Target ROAS or Target CPA both work well
Target CPA: When and How to Use It
Target CPA tells Google: 'For every conversion you get me, try not to spend more than X.' It works by bidding higher on users that look like your past converters and lower on those who don't. Use it when you know your target cost-per-acquisition and have 30+ conversions per month.
- Set your initial target at your current actual CPA, not your ideal CPA
- Wait 2-3 weeks after switching before evaluating β the algorithm needs a 'learning period'
- Don't change the target more than 10-15% in a single adjustment β dramatic changes restart the learning period
- If conversions drop more than 20% after switching, pause and diagnose before continuing
Target ROAS: When and How to Use It
Target ROAS tells Google: 'For every dollar I spend, try to generate X dollars in revenue.' It requires conversion value tracking (not just conversion counting) β Google needs to know the value of each conversion, not just that one happened.
- Use when you have different conversion values (e.g., product prices, lead quality tiers)
- Set your initial target at your current actual ROAS, not your aspirational ROAS
- Requires 50+ conversions per month for reliable optimization
- Works best for ecommerce where purchase values are tracked precisely
Related articles
Google Ads Account Structure: How to Organize Campaigns That Actually Convert
Learn the optimal Google Ads account structure for 2026. Covers SKAG vs STAG vs modern approaches, structures for small to large accounts, common mistakes, and how to restructure without losing data.
Average CPC by Platform 2026: Google Ads vs Meta vs LinkedIn vs Indeed vs TikTok
Compare average cost per click across Google Ads, Meta Ads, LinkedIn Ads, Indeed, and TikTok Ads in 2026. Real data by industry with benchmarks to help you allocate ad budget across platforms.
Google Ads Monitoring in 2026: Why You Need an AI Watchdog for Your Campaigns
Learn why manual Google Ads monitoring fails and how AI watchdog tools detect anomalies, budget problems, and conversion breaks hours before they cost you. Includes a comparison of monitoring approaches and what to automate first.
Full topic guide
Google Ads Smart Bidding: Complete Strategy Guide βThe complete guide to Google Ads Smart Bidding β Target CPA, Target ROAS, Maximize Conversions, Maximize Conversion Value. When to use each, data requirements, learning periods, and common mistakes.
Related free tools
Stop reading. Start optimizing.
Connect Google Ads and detect wasted spend with real data
AdPredictor analyzes your campaigns with AI: zero-conversion keywords, budget alerts, and actionable recommendations. Read-only. No registration required.
Get your free auditNo registration Β· Results in 30 seconds
Weekly AI & Google Ads Newsletter
AI news, Google Ads tips, and optimization strategies every Tuesday. No spam.
Content reviewed by the AdPredictor team. Editorial policy