Optimizing a display campaign comes down to five levers: targeting the right people, feeding the algorithm strong creative assets, choosing a bidding strategy your conversion volume can support, cutting wasteful placements, and measuring results beyond the last click. Each lever interacts with the others, so improving just one can lift overall performance. Here’s how to work through each one systematically.
Give Responsive Ads Enough Assets to Test
Google’s responsive display ads use machine learning to combine your headlines, descriptions, and images into thousands of variations, then serve whichever combination performs best for each impression. The system can only optimize what you give it, though, and most advertisers underload their campaigns.
You can upload up to 5 short headlines (30 characters each), 1 long headline (90 characters), 5 descriptions (90 characters each), and up to 15 images. Google recommends at least 5 images. You need two aspect ratios: horizontal at 1.91:1 (ideally 1200 x 628 pixels, minimum 600 x 314) and square at 1:1 (ideally 600 x 600, minimum 300 x 300). Uploading both ratios is critical because the ad platform selects whichever fits the available slot. If you only provide one ratio, you lose access to a large share of inventory.
Fill every asset slot. Five headlines beat two because the algorithm gets more combinations to test. Write headlines that each make a different appeal: one focused on price, one on a benefit, one on urgency, one on social proof. Descriptions should work independently of any headline since the system pairs them freely. Avoid repeating the same message across assets.
Check your asset performance labels regularly. Google grades each headline, description, and image as “Best,” “Good,” or “Low.” Replace low-performing assets with new variations rather than leaving them in rotation. This iterative swapping is one of the simplest ways to improve click-through rates over time without touching targeting or bids.
Match Your Bidding Strategy to Your Conversion Volume
Smart Bidding strategies like Target CPA (cost per acquisition) and Target ROAS (return on ad spend) use Google’s AI to set bids in real time for each auction. They work well, but only when they have enough conversion data to learn from. Without sufficient volume, the algorithm swings wildly and your costs become unpredictable.
Google recommends at least 30 conversions per month per ad group for Target CPA or Target ROAS to perform consistently. At fewer than 30 conversions, your actual CPA or ROAS can fluctuate by up to 100% from your target, the system takes up to two weeks to react to changes, and the initial learning period stretches to four weeks. At 50 conversions, fluctuation drops to about 50%. At 100, it falls to around 20% with a faster reaction time. Campaigns hitting 500 or more conversions per month see very low fluctuation and can adapt to changes in as little as six hours.
If your campaign generates fewer than 15 conversions in 30 days, Target ROAS may not even be eligible to run. In that situation, start with Maximize Clicks or Maximize Conversions (without a target) to build volume first, then switch to Target CPA or Target ROAS once you have enough data. Another option is to use a broader conversion action, like tracking “add to cart” instead of “purchase,” to give the algorithm more signals during the learning phase. You can tighten the conversion action later once volume supports it.
Use Optimized Targeting Strategically
Display campaigns offer two ways to expand your reach beyond manually selected audiences: optimized targeting and audience expansion. They sound similar but work differently, and choosing the wrong one can waste budget.
Optimized targeting looks at real-time conversion data to find people most likely to convert, even if they fall outside your selected audience segments. Your manual targeting choices act as a starting point that informs the model, but the system will go beyond them if it finds converting users elsewhere. This is the better option when your goal is conversions and you want to find new customers while staying close to your CPA goals.
Audience expansion finds people who resemble your existing audience segments. It’s designed to increase reach, impressions, and views rather than directly optimize for conversions. It works best for awareness and consideration campaigns, particularly video, where the goal is volume rather than a specific cost-per-conversion target.
When optimized targeting is turned on (it’s enabled by default in most campaign types), monitor your audience insights to see where the algorithm is finding converters. If it’s pulling in traffic from irrelevant segments, you can add audience exclusions to steer it. If performance is strong, consider narrowing your manual audience signals further and letting the algorithm do more of the prospecting work. The key is treating your manual segments as hints rather than hard boundaries.
Cut Low-Quality Placements
Display campaigns can serve ads across millions of websites and apps, and not all of that inventory is worth paying for. Reviewing and excluding poor placements is one of the fastest ways to reduce wasted spend and improve your cost per conversion.
Start by checking your placement report regularly. Sort by cost and look for sites or apps that have consumed budget without producing conversions. Mobile game apps are a frequent offender: they generate accidental clicks from users trying to close interstitial ads. Exclude specific apps or entire app categories that consistently underperform.
Beyond individual sites, consider excluding broader categories of content that tend to produce low-quality traffic. Pages with high ad density, including sites that run “skin” ads wrapping the entire page, are usually cluttered environments where your ad competes for attention against dozens of others. These placements may appear cheap on a CPM basis, but the engagement quality is typically poor. Parked domains, error pages, and sites in languages your customers don’t speak are other common exclusions.
Build a shared placement exclusion list at the account level so you don’t have to re-exclude the same low-quality sites across every new campaign. Update this list monthly as new poor performers surface. Some advertisers also use third-party brand safety tools or Google’s content suitability settings to preemptively block categories like sexually suggestive content, tragedy-related pages, or below-the-fold-only inventory.
Measure Beyond the Click
Display ads influence purchasing decisions differently than search ads. Someone might see your banner, not click it, but visit your site directly the next day and convert. If you only measure click-through conversions, you’ll undervalue display and potentially cut campaigns that are actually driving revenue.
View-through conversions (VTCs) track exactly this scenario: a user sees your ad, doesn’t click, then later completes a conversion on your site. A “view” in this context means at least 50% of the ad’s pixels were visible on screen for at least one second. The last display impression before the conversion gets credit. You can find this metric in the “View-through conversions” column in Google Ads, though you may need to add it to your reporting view manually.
VTCs are valuable directional data but require some skepticism. Not every person who saw your ad was influenced by it. A common approach is to weight view-through conversions at a fraction of click-through conversions, perhaps 20% to 50%, when evaluating campaign ROI. This gives display credit for its awareness-building role without treating every impression as a direct cause of the sale.
When setting up conversion tracking, also pay attention to your conversion window. The default lookback window for display view-through conversions is typically shorter than for click-through conversions. Extending or shortening this window changes how many VTCs get counted, which directly affects how your campaign’s performance appears in reports and how Smart Bidding algorithms optimize. If your product has a long consideration cycle, a longer window may better reflect reality.
Structure Campaigns for Cleaner Data
How you organize your campaigns and ad groups directly affects how well automation can optimize. A single ad group mixing prospecting audiences with remarketing audiences gives the algorithm conflicting signals, since remarketing traffic converts at higher rates and will absorb most of the budget.
Separate prospecting and remarketing into different campaigns with their own budgets and bidding targets. Your remarketing campaign will likely have a lower CPA target since those users already know your brand. Your prospecting campaign needs room to spend more per conversion because it’s reaching cold audiences.
Within each campaign, group ad groups by theme or product line so you can tailor creative assets to each audience. An ad group targeting in-market audiences for running shoes should show running shoe imagery and copy, not generic brand messaging. The tighter the alignment between audience, creative, and landing page, the higher your conversion rate.
Finally, give new campaigns or significant changes at least two to three weeks before judging results. Smart Bidding enters a learning period after any major change, including new bid strategies, large budget shifts, or new conversion actions. Performance during this period is typically volatile and doesn’t reflect long-term results. Avoid making additional changes during learning, as each one resets the clock.