Writing High-Performing Search Ads: RSA, Pinning, Ad Strength and Testing
Responsive Search Ads (RSAs) have become the backbone of Google Ads search campaigns. Since June 2022, when Google sunsetted Expanded Text Ads, RSAs are the default format for all new search campaigns. Yet many advertisers still struggle to leverage their full potential.
The challenge is not understanding RSAs exist. It’s mastering the mechanics: how to structure assets for maximum testing, when to pin (and why not to), how Ad Strength translates to conversions, and which testing methodologies actually work. Get this right, and you unlock 12% more conversions on average. Get it wrong, and you waste budget on ad combinations that never stood a chance.
This guide walks through the complete RSA framework: from asset architecture to pinning strategy, from decoding Ad Strength to running proper performance tests. You will learn why pinning just one headline cuts testing potential by 75%, why “Excellent” Ad Strength is not always better than “Average”, and how to structure an experiment that doesn’t mislead you.
Understanding Responsive Search Ads: How They Work
As Google Ads Help documents, an RSA is fundamentally different from its predecessor, the Expanded Text Ad. Where an ETA was a fixed combination of two headlines and two descriptions, an RSA is a template that generates hundreds of unique combinations from the assets you provide.
You supply up to 15 headlines and 4 descriptions per ad. Google’s machine learning engine then tests thousands of combinations and learns which perform best for each search query. A single RSA can serve a headline alone on mobile, or pair it with a different description on desktop. The system adapts to context: brand awareness searchers see different messaging than high-intent purchasers, even within your same ad group.
This flexibility is the whole point. Except when it isn’t. According to Search Engine Journal research, most RSA underperformance stems from advertisers undermining the system through aggressive pinning, weak asset diversity, or poor copy architecture.
How Google Tests RSA Combinations
Google’s system treats each RSA as a set of mathematical combinations. With 15 headlines and 4 descriptions, you create a theoretical maximum of 60 headline-description combinations. In reality, Google also tests single-headline variants and tests placement positions, multiplying the total possibilities into the thousands.
The engine doesn’t test randomly. It learns. Early impressions show broad combinations. As data accumulates, Google identifies patterns: which headlines correlate with clicks, which descriptions drive conversions, which pairings work in tandem. The algorithm optimizes toward your business objective, whether that’s clicks, conversions, or return on ad spend.
This learning curve matters. A new RSA needs 2-4 weeks of solid traffic to reach stable performance. Changing assets too aggressively during this window introduces noise and slows learning.
Why RSAs Outperform Fixed Ads (In Theory)
Responsive Search Ads outperform Expanded Text Ads for one reason: relevance at scale. An ETA showed the same two headlines to every searcher. An RSA shows different combinations based on search intent, user device, time of day, and predicted engagement.
According to WordStream’s analysis, Google’s research shows that advertisers combining machine learning with multiple creative assets see up to 15% more clicks. The formula is simple: more asset diversity equals more testing opportunities equals higher relevance matching.
But this only works if you give the system assets worth testing. A weak asset in your pool drags down the whole set. Similarly, over-pinning breaks the system by reducing combination diversity. The architecture must support the machine learning, not fight it.
Building Your Asset Library: Headlines and Descriptions That Convert
RSA performance starts with copy. You cannot optimize your way around weak assets. The machine learning will find your best performers, but it works only with what you provide.
Google’s recommendation is straightforward: provide at least 10 headlines and aim for all 15. But quantity without quality wastes the system. Each asset must serve a distinct purpose in your testing matrix.
Headline Structure: Five Asset Types
Your 15 headlines should include five distinct types, distributed across your asset pool. This diversity ensures Google has combinations to test across different user intent profiles.
Keyword-Rich Headlines
Include your primary keyword in at least three headlines. Not forced, not stuffed, but natural. If you sell “organic coffee delivery”, use “Organic Coffee Delivered Fresh” not “Organic Coffee Organic Coffee Delivery Best”. Google’s quality score system rewards naturalness. Users respond to clarity.
Target your top 3-5 keywords across these headlines. If you bid on both “coffee subscriptions” and “bulk coffee wholesale”, create keyword-rich variants for each. This signals relevance to Google’s matching algorithm and increases exact match impressions.
Value Proposition Headlines
Ask yourself: why should a searcher click your ad over a competitor’s? Not a feature. A benefit. “Free Shipping on Orders Over $30” is a benefit. “Organic Coffee” is a feature.
Include 2-3 headlines that answer the customer’s implicit question: “Why you, not them?” Examples: “Same-Day Delivery Guaranteed”, “100% Money-Back Guarantee”, “Rated 4.9 by 8,000+ Customers”. These headlines test your unique value against the competition.
Call-to-Action Headlines
Calls-to-action in headlines drive higher engagement. Not all searchers are ready to convert at the same stage of their journey. Early-stage awareness searchers respond to “Learn How We Source Direct”. Late-stage purchasers respond to “Buy Now, Save 15%”.
Include 2-3 CTA headlines: one soft (“Explore Our Blends”), one medium (“Get Your First Bag Free”), one hard (“Subscribe and Save 20%”). Let Google test which conversion stage your audience is in.
Feature and Benefit Headlines
Round out your mix with 3-4 headlines highlighting specific features or benefits beyond your main value prop: “Ethically Sourced”, “Single-Origin Options”, “Wholesale Pricing Available”, “Roasted Fresh Weekly”. These serve as differentiators when paired with broader headlines.
Brand and Trust Headlines
If your brand carries weight, deploy it. “Coffee by LaFactory: Since 2010”, “Award-Winning Roaster”, “Trusted by 50,000 Coffee Lovers”. Trust headlines perform well in competitive categories where reputation matters. They’re also essential legal signals (brand name, awards, certifications).
Description Copy: Building the Support Case
Headlines catch attention. Descriptions convince. Each description (you have four) should reinforce a different angle, not repeat what your headlines say.
Your four descriptions might cover: (1) the logistics (“Ships within 24 hours”), (2) the product quality (“Single-origin beans from 200+ small farms”), (3) the customer experience (“Personalized tasting notes with every order”), (4) the value proposition (“Start with 25% off, then save 15% monthly”).
Descriptions perform best when specific. “We offer great service” underperforms compared to “Average delivery: 36 hours. 98% on-time rate.” Numbers, data, and specificity boost conversion rates because they signal credibility.
Testing Headlines for Uniqueness
Before launching, manually verify: do these assets convey different messages? Or are you writing the same benefit seven ways? Read your 15 headlines in sequence. If more than two feel redundant, replace one.
Google’s Ad Strength tool gives you clues about duplication. If it flags “low variety”, examine your asset pool. But don’t rely solely on the tool: read your own copy first.
Ad Strength: What It Means and Why It Doesn’t Tell The Whole Story
As detailed in Google’s Ad Strength guide, Ad Strength is Google’s built-in quality signal for RSAs. It grades your asset pool across five levels: Incomplete, Poor, Average, Good, and Excellent. Each improvement from Poor to Excellent correlates with approximately 3% more clicks.
Moving an RSA from Poor to Excellent Ad Strength yields 12% more conversions on average. This statistic gets repeated constantly. But it’s incomplete. Ad Strength measures input quality, not output performance. They are not the same thing.
How Ad Strength Is Calculated
Ad Strength evaluates three dimensions: quantity (do you have enough assets?), relevance (do your assets match your keywords?), and diversity (do your assets express different messages?).
Quantity is straightforward. Google wants 15 headlines and 4 descriptions. Anything less is flagged. Relevance means your assets reference keywords that match your keyword list. Diversity means no message duplication.
The system also evaluates pinning. Every pin you add reduces the available combinations and lowers Ad Strength. This is intentional: pinning is a constraint that limits testing. Google’s algorithm downgrades strength when you constrain it too much.
Why Excellent Is Not Always Better
This surprises most advertisers. Case studies from real accounts show that “Good” Ad Strength sometimes converts better than “Excellent”. How is this possible?
The answer lies in what Ad Strength optimizes for versus what you optimize for. Ad Strength rewards quantity and diversity of assets. But more diversity can mean less focus. A highly specific, tightly crafted set of 4-5 very strong headlines might convert better than 15 generic ones, even though the smaller set scores lower on Ad Strength.
The 12% uplift cited by Google applies across aggregate accounts. Individual accounts with different audience profiles, offers, and market conditions can deviate from the average. Your job is to test your account’s actual relationship between Ad Strength and conversion performance, not blindly chase a score.
Actionable Ad Strength Signals
Use Ad Strength diagnostically, not prescriptively. When it flags “low relevance”, check if your assets mention your top keywords. If not, add keyword-rich headlines. When it flags “low diversity”, review message duplication and prune similar assets.
But when it tells you to add assets just to hit Excellent, question whether those assets improve conversion probability. Quality over completion. A focused asset pool that converts at 5% beats a bloated pool that converts at 3%, regardless of what the score says.
Monitor your account’s empirical relationship: as you improve Ad Strength, do conversions improve? By how much? After a few months of data, you’ll know your account’s real Ad Strength curve, which may differ from Google’s published benchmarks.
Pinning Strategy: When to Lock Assets and Why Less Is More
Pinning lets you force a headline or description into a specific position. Position 1 always shows a pinned headline. Position 2 always shows another pinned headline. Pinned descriptions appear only as you specify.
Pinning exists for legitimate reasons. You might legally need your brand name in every ad. You might need a disclaimer. You might have historical data proving one asset vastly outperforms all others, and pinning protects that winner from rotation.
But pinning has an enormous cost: it cuts testing diversity. This is why understanding the trade-off is critical.
The Mathematical Cost of Pinning
With 15 headlines and no pinning, Google can test 15 different headline combinations. Pinning one headline to Position 1 reduces testing options to the remaining 14 headlines. Pinning two headlines (one to each position) reduces options further.
More precisely: when you pin two headlines, Google can still test various combinations of descriptions and other headlines, but it has fewer combinations overall. Research shows that pinning just one headline cuts total testing potential by 75%.
This explains why Google Ads explicitly recommends pinning only when essential. The flexibility of RSAs only works at scale.
When to Pin Headlines
Pin headlines in these scenarios:
Legal or Branding Requirements
If your legal team requires your company name in every ad, pin it. “Allstate Insurance” must appear in every ad for brand consistency and legal compliance. This is non-negotiable and worth the testing cost.
Proven Winner Data
If 6+ months of historical data shows one headline dramatically outperforms all others (50%+ higher CTR than second place), consider pinning it. But only if that performance gap is recent and stable, not from an old campaign that may not reflect current market conditions.
Time-Sensitive Offers
If you’re running a “Limited Time: 40% Off” promotion, pin that headline to Position 1. Time-sensitive offers need constant visibility, and the testing cost is worth the urgency signal.
Outside these cases, avoid pinning. The testing losses exceed the benefits.
Pinning Strategy for Multiple Assets
If you must pin, pin conservatively. Instead of pinning one headline to Position 1, pin 2-3 headlines that can rotate through Position 1. This preserves testing on positions 2-3 and descriptions while protecting your pinned headline set.
Google calls this “loose pinning” and it’s more sophisticated than absolute pinning. You get protection (brand name always visible) with minimal testing loss.
Never pin all assets. A fully pinned RSA shows the same combination to every searcher and defeats the entire purpose of the format. It will score Poor on Ad Strength and underperform compared to unpinned RSAs in your account.
Monitoring Pinning Impact
After implementing pinning, track these metrics over 4 weeks: total impressions, click-through rate, conversion rate. If impressions drop 30%+ or CTR drops 20%+, the pinning cost is too high. Unpin and test a less restrictive strategy.
Pinning should preserve performance, not sacrifice it. If performance metrics decline, the strategic reason for pinning (brand requirement, legal mandate) must outweigh the loss. If it doesn’t, reconsider the approach.
Testing RSAs: How to Actually Know If One Ad Beats Another
Testing RSAs sounds straightforward: create two ads, run them against each other, declare a winner. In reality, RSA testing is methodologically complex because you’re not testing two fixed ads. You’re testing two universes of possibilities.
When you create RSA #1 with 15 headlines and RSA #2 with 15 different headlines, Google tests approximately 43,680 combinations of RSA #1 against 43,680 combinations of RSA #2. The variables that influence which combination serves when are uncontrolled: search query, user device, time of day, user history, seasonality.
This means that even if you identify a “winner”, the win is noisy. You’ve identified that Universe #1 slightly outperformed Universe #2 in your limited test window, but there are dozens of variables you didn’t isolate.
Sample Size and Statistical Confidence
RSAs need significant traffic to produce reliable test results. A small account running 5,000 monthly impressions cannot confidently test RSAs because the traffic is distributed across too many combinations. By the time one combination gets meaningful impressions, the test window closes.
For reliable RSA testing, aim for 500+ conversions per ad before declaring a winner. At a 2% conversion rate, that requires 25,000 clicks. At a 2% CTR, that requires 1.25 million impressions. This is realistic for account spending $10,000+ monthly on search, but not for smaller accounts.
Small accounts should rotate new RSAs in for 2-4 weeks, monitor aggregate performance, and keep the best performer. Don’t over-optimize on noisy data.
Testing One Asset at a Time
If you want to isolate what drives performance, test one change at a time. Replace 3 headlines in an existing RSA and monitor for 2 weeks. If metrics improve, keep the change. If they decline, revert. This sequential testing is slower than running two parallel RSAs, but it yields clearer causality.
Parallel testing is faster but muddier. You sacrifice certainty for speed. Choose based on your account size and timeline constraints.
Asset-Level Performance Data
Google recently rolled out headline-level performance metrics for RSAs. In the Ads interface, you can now see clicks and conversions attributed to individual headlines. This is a game-changer for RSA optimization.
With headline-level data, you can identify underperformers and pause them, improving overall Ad Strength and testing quality. You can also identify top performers and build future RSAs around that signal.
Before this feature, you had to use indirect signals like “Good” or “Best” labels. Now you have actual numbers. Leverage this data to iterate your asset pool every 4-6 weeks.
Building Your Testing Schedule: A Cadence for Continuous Improvement
RSA optimization is not a one-time project. It’s an ongoing cycle. Successful accounts iterate on their assets every 4-6 weeks.
Week 1-2: Baseline Performance Measurement
Set your current RSA as the control. Measure its baseline metrics: impressions, clicks, CTR, conversions, conversion rate, average CPC, ROAS. Use Google Ads reporting or your analytics platform.
Obtain at least 100 clicks and 5-10 conversions before declaring this a reliable baseline. For low-volume accounts, this may take 2-3 weeks.
Week 3-4: Asset Audit and New RSA Build
Analyze your headline performance data. Identify top 3-5 performing headlines. Identify bottom 2-3 underperformers. For the new RSA, retain your top performers and replace bottom performers with fresh assets.
Also read your descriptions: are they supporting the headlines effectively? Are you repeating the same message across multiple assets? Refresh one description that feels generic.
Week 5-8: Parallel Testing
Launch your new RSA alongside the current one. Both run in the same ad group. Because they’re both RSAs, budget will rotate based on performance, and Google will weight toward the better performer.
Monitor aggregate metrics weekly. After 2 weeks, one RSA should show slight performance edge. After 4 weeks, the pattern should be clear.
Week 9: Decision
Compare headline-level data. If your new RSA’s headlines show stronger click-through or conversion rates, pause the old RSA and keep the new one. If the old RSA performs better, delete the new one and go back to the drawing board.
If performance is nearly identical (within 5%), keep the new RSA. Fresh assets often show modest improvements that become clearer after 2-3 months.
Scaling to Multiple Audiences
If you run different ad groups for different keywords or audiences, apply this cadence to your highest-volume ad group first. Once you’ve optimized that, apply the learning to the next tier down.
Top ad groups deserve more testing iterations. Bottom ad groups stabilize faster with borrowed insights.
Integration: RSA, Pinning, Ad Strength, and Testing as One System
These four elements work together. Ad Strength measures the health of your testing environment. Pinning controls which tests run. Testing reveals which assets win. Your pinning and asset strategy must support the testing, not undermine it.
Here’s the integrated workflow:
Month 1: Foundation
Build your first RSA with 15 strong headlines and 4 descriptions. Aim for Good Ad Strength (don’t chase Excellent if it means adding weak assets). Pin only what’s legally required. Collect baseline performance data.
Month 2-3: Audit and Iterate
Review headline-level performance. Pause bottom performers. Add 2-3 fresh headlines that address different angles (CTA, value prop, feature spotlight). Monitor Ad Strength as you iterate.
Month 4+: Systematic Testing
Once you’ve built a library of proven headlines, run parallel RSAs. Rotate in new combinations. Use headline performance data to guide which assets stick around and which rotate out.
Every quarter, step back and analyze: are your assets still aligned with your target audience? Have market conditions, competition, or customer intent shifted? Refresh your asset pool accordingly.
The compounding effect of quarterly refreshes, disciplined pinning, and systematic testing yields 20-40% improvement in ROAS over 12 months. But only if you stay consistent and avoid chasing noise in small data samples.
Practical Scenarios: When Theory Meets Practice
Understanding RSA mechanics is one thing. Applying them to real budgets and timelines is another.
Scenario 1: High-Volume Account ($50,000+ Monthly Spend)
You have sufficient traffic to run multiple parallel RSA tests. You can afford to run three different RSAs per ad group and let machine learning identify winners over 3-4 weeks.
Strategy: Build diverse RSAs. One focuses on price/discount. One on features/quality. One on urgency/scarcity. Test all three. Google will distribute budget toward the best performer. After 4 weeks, pause the bottom performer, build a new one based on headline insights, and start the cycle again.
Scenario 2: Medium-Volume Account ($5,000-20,000 Monthly Spend)
You have enough traffic for sequential testing but not parallel multivariate testing. You need 4+ weeks to see reliable patterns.
Strategy: Maintain one high-quality RSA. Make one small change (replace 2-3 assets). Monitor for 4 weeks. Implement the change if it wins, revert if it loses. Move to the next change. This sequential approach is slower but more statistically sound given your traffic constraints.
Scenario 3: Small-Volume Account ($500-5,000 Monthly Spend)
You lack traffic for reliable testing. Trying to test RSA variations introduces noise. Focus on quality over quantity.
Strategy: Build one RSA with your best, most confident assets (no weak filler). Skip parallel testing. Instead, rotate new assets in once every 4-6 weeks if aggregate metrics suggest opportunity. Accept that you won’t have granular testing data; use monthly trends instead.
Common RSA Mistakes and How to Avoid Them
Most RSA underperformance stems from a few recurring errors. Knowing these prevents costly missteps.
Mistake 1: Premature Pinning
New accounts pin too much, too early. They lock in headlines before accumulating enough data to know if those headlines actually perform. Pins then sit for months, constraining testing and preventing performance optimization.
Fix: Don’t pin anything in your first RSA. Collect 4 weeks of headline-level performance data. Only then pin if you have proof and a legitimate reason (legal, brand, competitive).
Mistake 2: Low-Quality Asset Filling
Advertisers chase the “15 headlines” recommendation by adding generic assets: “Learn More”, “Check Us Out”, “Browse Our Products”. These add noise, not signal. Google’s algorithm still tests them, wasting impressions on weak combinations.
Fix: Quality over quantity. 10 strong headlines beat 15 weak ones. Stop at 12 if your 13th, 14th, 15th are filler.
Mistake 3: Overstuffing Keywords
Advertisers repeat keywords across all 15 headlines: “Cheap Coffee”, “Affordable Coffee”, “Low-Cost Coffee”, “Budget Coffee”. Google’s algorithm now understands stemming. Repetition signals desperation, not relevance.
Fix: Use your primary keyword 3 times. Use related keywords 2-3 times (“coffee subscription”, “bulk coffee”, “coffee delivery”). Fill remaining slots with benefits, CTAs, and social proof.
Mistake 4: Testing Without Sufficient Data
Advertisers declare a winner after 1 week or 100 clicks. At this scale, randomness dominates. Small sample sizes produce false positives and lead to poor decisions.
Fix: Commit to 4-week testing windows minimum. Target 500+ conversions per RSA for high confidence. Use statistical confidence calculators to know your sample size requirements.
Mistake 5: Ignoring Seasonal Swings
Advertisers test RSAs in January, lock in a winner, and never revisit it. But by June, audience intent has shifted, and the January winner is stale.
Fix: Refresh assets quarterly, not annually. Mark your calendar: every 13 weeks, audit top performers and build one new RSA to test new angles.
Platform-Specific Insights: Search Ads 360 and Google Ads API
If you manage large portfolios, you need systematic RSA management beyond the native Google Ads interface.
Search Ads 360 for Multi-Account Management
Search Ads 360 provides cross-account RSA performance reporting. You can see headline-level metrics across 50+ Google Ads accounts simultaneously. This enables portfolio-wide optimization: which headlines work across accounts? Which are account-specific?
SA360’s alert system flags declining Ad Strength or deteriorating headline performance, so you catch problems before they compound. For agencies managing multiple clients, this visibility is essential.
Google Ads API for Programmatic Testing
The Google Ads API lets you programmatically create RSAs, update assets, and retrieve headline-level performance metrics. Sophisticated teams build custom systems that:
This requires engineering expertise but yields operational efficiency and faster optimization cycles.
Conclusion: Mastering RSA Performance
Responsive Search Ads are powerful when used correctly and wasteful when built carelessly. The difference comes down to three disciplines: building diverse, high-quality assets; pinning strategically and sparingly; measuring Ad Strength as one input among many; and testing rigorously with sufficient data.
Your RSA strategy should evolve with your account. Month 1 focuses on establishing a strong baseline. Month 2-3 introduces testing and iteration. Month 4+ scales to systematic optimization with quarterly refreshes.
If you’re currently underperforming, audit your existing RSAs through this framework. Do your headlines represent five distinct types (keyword-rich, value prop, CTA, features, brand)? Are you pinning excessively? Are your assets diverse or repetitive? Is your testing window long enough to detect real signals?
Address these fundamentals, and you’ll unlock the hidden potential in Responsive Search Ads. The machine learning is already powerful. Your job is to build the right testing environment and feed it good assets. Do that, and performance improvements follow naturally.