Google Ads RSAs vs Single Keyword Ad Groups: Which Structure Actually Lowers CPA for Local Service Campaigns?

Why This Debate Still Matters in 2024

If you inherited a Google Ads account built before 2021, there's a reasonable chance it's loaded with Single Keyword Ad Groups (SKAGs) — one keyword per ad group, tightly matched, with ads that mirror the keyword almost word-for-word. The logic was sound: maximum control over Quality Score, ad relevance, and spend.

Then Google pushed Responsive Search Ads (RSAs) as the default, deprecated expanded text ads, and leaned its algorithm toward broader, machine-driven matching. Now the question every local service advertiser has to answer is: do I consolidate into RSA-first broad groups, or defend my SKAG structure?

The answer is genuinely situational — and this teardown gives you the framework to decide, not a one-size-fits-all verdict.

What Each Structure Actually Optimizes For

SKAG logic (legacy):

• One keyword → one ad group → one tightly-matched RSA or set of ads
• High expected ad relevance score (one of three Quality Score pillars)
• Granular negative keyword management at the ad-group level
• Manual or tCPA bidding with predictable, narrow auction entry

RSA-first / consolidated structure (current Google-recommended):

• Multiple related keywords in one ad group feeding one RSA with 15 headlines and 4 descriptions
• Google's machine learning tests headline combinations and learns which resonate
• Broader auction eligibility feeds more conversion signals into Smart Bidding
• Fewer ad groups = simpler account, but less surgical negative keyword control

The core tension: SKAGs starve Smart Bidding of conversion data by fragmenting it across dozens of tiny ad groups. RSA-first consolidation feeds the algorithm — but risks diluting ad relevance if your keyword themes aren't genuinely tight.

Quality Score Signals: Where Each Structure Wins and Loses

Quality Score has three components: expected CTR, ad relevance, and landing page experience. Here's how structure affects each:

| Signal | SKAG advantage | RSA-first advantage | |---|---|---| | Expected CTR | Neutral — depends on ad copy, not structure | RSA rotation tests more combinations, can lift CTR over time | | Ad Relevance | Strong — keyword-to-ad match is near-literal | Moderate — relies on good headline asset coverage | | Landing Page Experience | Neither — depends on your page, not ad structure | Neither |

Practical implication: A SKAG account with mediocre copy often outscores an RSA account on ad relevance initially, then gets overtaken as RSA learns. For local service campaigns where search intent is narrow (e.g., 'emergency plumber Dallas' vs 'plumber'), the SKAG relevance edge is more durable because the intent doesn't vary much — the machine has less to learn.

For campaigns with broad intent variation (e.g., a multi-service HVAC company covering installs, repairs, and maintenance), RSA-first wins because no single SKAG ad can serve all variants well.

The Conversion Data Threshold: A Labeled Efficiency Model

This is the most important factor most local advertisers ignore.

Smart Bidding (tCPA, tROAS, Maximize Conversions) needs roughly 30–50 conversions per month per campaign to exit the learning phase and bid efficiently — this is Google's own published guidance, not an estimate. Below that threshold, the algorithm guesses.

Illustrative model — two local plumbing campaigns, same monthly budget:

> Note: All numbers below are labeled illustrative models, not measured benchmarks.

SKAG structure: 40 ad groups, 10 conversions/month total → ~0.25 conversions per ad group. Every ad group is perpetually in learning mode. Smart Bidding overshoots or undershoots bids continuously. Estimated CPA inflation: 20–40% above potential, purely from data starvation.

RSA-first (consolidated to 4 ad groups): Same 10 conversions/month → 2.5 conversions per ad group. Still thin, but 10x more signal per bidding unit. Smart Bidding exits learning faster. At a illustrative $180 CPA under SKAG, consolidation alone could plausibly reduce CPA to the $120–$150 range — not from better ads, but from smarter bid decisions on more concentrated data.

The rule of thumb: If your campaign generates fewer than 50 conversions/month total, SKAG structure is almost certainly costing you money by fragmenting signal. See our article Minimum Ad Spend for Google Smart Bidding (Local) for the budget floor this implies.

When SKAGs Still Win: The Case for Defending Granular Structure

Consolidation isn't always the answer. SKAGs (or tightly-themed small ad groups) still outperform in specific scenarios:

• High-volume, high-intent single services. A dedicated campaign for 'roof repair [city]' with 80+ conversions/month has enough data per group. Adding unrelated roofing keywords to chase volume can dilute relevance and hurt conversion rate.
• Competitor conquest campaigns. Competitor-keyword ad groups benefit from surgical negative keyword control and ad copy that speaks directly to brand switchers. Mixing these into a broad RSA group typically reduces relevance and wastes spend.
• Premium service tiers with very different CPAs. If your emergency service converts at a $90 CPA and your maintenance plan at $280 CPA, keeping them in separate ad groups lets you set different tCPA targets. Merging them forces Smart Bidding to split the difference — and you likely underspend on the high-value convert and overspend on the low-value one.

The decision tree: volume + intent homogeneity = consolidate. Low volume + intent fragmentation = defend your groups.

The Hybrid Architecture Most Local Advertisers Miss

The real answer isn't RSA vs SKAG — it's a tiered account structure that applies each logic where it fits:

1. Branded campaign: One tight ad group, RSA with brand terms, brand protection. Simple. 2. Core service campaigns (1–3 per service line): RSA-first with 5–8 tightly-themed keywords per ad group. Feed Smart Bidding, but don't dilute themes. 3. High-value micro-campaigns: SKAG or near-SKAG for competitor terms, emergency-intent keywords, or premium service tiers — only where conversion volume justifies the granularity. 4. Negative keyword architecture: Shared negative lists that prevent cannibalization across tiers. This is the work that makes consolidation safe.

This hybrid approach also matters for how paid and organic interact — if you're running both, the blended signal picture changes. Our article Paid Search + Organic SEO: Lowest Blended CPA? covers how to account for assisted conversions when measuring true CPA across channels.

And if you're adding lead channels alongside Google Ads, read Does Adding Lead Channels Lower Your Close Rate? before assuming more volume is always better — structure decisions downstream affect close rates, not just CPA.

The Takeaway: Structure Is a Lever, Not a Religion

SKAG architecture was a rational response to a pre-machine-learning ad auction. RSA-first consolidation is a rational response to the auction Google actually runs today. Neither is correct in the abstract.

The framework:

• Under 50 conversions/month total: Consolidate aggressively. Data starvation is your biggest CPA driver.
• Over 50 conversions/month per service line: Evaluate relevance dilution. Consolidating dissimilar intent kills Quality Score faster than fragmentation kills bidding efficiency.
• Always: Maintain tight negative keyword coverage. Consolidation without negatives is just wasted budget at scale.

If you're not sure where your account falls, the fastest diagnostic is to pull your ad group conversion volume report, segment by ad group, and count how many groups have fewer than 5 conversions in the last 30 days. If it's more than half your account, you have a fragmentation problem — and restructuring is likely your highest-ROI move before touching creative or landing pages.

Want a second set of eyes on your account structure? Book a free strategy call with the Nika Spark team — we'll audit your ad group architecture, conversion signal distribution, and Smart Bidding setup and tell you exactly where CPA is leaking.