What is a good example of a Google ad?

A good Google ad example is one with a measured result. Malibu Shirts' Search ads went from 0.44x to 2.93x ROAS on a flat budget after Hawke Media rebuilt them around high-intent keywords and value-led copy.

Which Google Ads campaign type works best?

It depends on the job. In one account @runtraffic shared, Search returned 9.37 ROAS, PMax 7.65, Shopping 6.12 and Demand Gen 2.66, but Search there included brand terms, and Demand Gen's job is finding new customers, so judge each on its own goal.

Do Performance Max campaigns beat Search?

Sometimes. Taxfix reported 9% more tax returns at 44% lower cost per conversion with PMax than pure Search (vendor-reported). Test it with an experiment against your own Search baseline rather than switching outright.

How do I know if my Google ads are actually working?

Run a holdout or lift test. @runtraffic's brand search holdout showed 8.17 incremental ROAS but only 1.17 on new customers, which is a very different picture from the platform's reported ROAS.

20 Google Ads Examples That Moved CPA and ROAS (2026)

The best Google Ads examples share one move: they cut what Google was averaging together.

They split brand from non-brand, best sellers from long tail, new customers from returning ones, and then measured the result with a holdout or lift test.

Our top picks are Malibu Shirts (Search ROAS 0.44x to 2.93x on a flat budget), The Sill (tiered Shopping), Cat Person (one PMax split into three) and Urban Armor Gear (Demand Gen proven with a lift study).

How we picked: all 20 examples are results from our database, taken from agency case studies, public posts by media buyers and Google's own success stories (labeled vendor-reported).

We have not run these campaigns ourselves, and most are before/after comparisons without a control group.

Quick answer: what the best examples have in common

Across Search, Shopping, PMax and Demand Gen, the winners did three things.

They gave their best demand its own campaign and budget.

They told Google which conversions were worth more, through value rules, margin labels or a new-customer goal.

And they judged upper-funnel spend on incremental results, not last-click ROAS.

Structure and bidding signals mattered more than ad text.

Summary table of every example

#	Example	What they did	Result	Type
1	Lavender	Search: SKAGs, non-brand keywords, lead form ads	CPA −95%	Lead gen
2	Malibu Shirts	Search: High-intent keywords, flat budget	ROAS 0.44x → 2.93x	Ecommerce
3	iHostPoker	Search: Campaign per event type, travel negatives	CPA −56%	Lead gen
4	Shoott	Search: Campaigns by city, less brand spend	CPA −22%	Marketplace
5	Brand search holdout (@runtraffic)	Search: Brand search holdout	iROAS 8.17	Ecommerce
6	ChatPDF	Search: tCPA vs tROAS, LTV value rules	ROAS +320% YoY	SaaS
7	Surfshark	Search: 5,000+ campaigns cut to under 50	Revenue nearly doubled (vendor)	SaaS
8	Fashionphile	Shopping: Shopping restructure, RLSA	ROAS +28%	Ecommerce
9	The Sill	Shopping: Tiered Shopping, low-bid non-brand	ROAS −23% → 403%	Ecommerce
10	GreenPan	Shopping: Shopping split by category	CVR +14.7%	Ecommerce
11	Droyd	PMax: Feed work, model-level creative	ROAS +248%	Ecommerce
12	Guitar brand (unnamed)	PMax: PMax on special editions, new customers	$500K at 3x ROAS	Ecommerce
13	Cat Person	PMax: Split one PMax into three	CPA −13%, spend +13%	Ecommerce
14	Penn Tool Co.	PMax: Low-margin products isolated	Revenue +194% YoY	Ecommerce
15	Dilling	PMax: New vs returning buyer PMax	ROAS +66% (vendor)	Ecommerce
16	Taxfix	PMax: PMax vs pure Search	CPA −44% (vendor)	App
17	Urban Armor Gear	Demand Gen: Demand Gen plus lift study	+13.6% sales lift	Ecommerce
18	Makeup brand (unnamed)	Demand Gen: YouTube for reach, geo-tested	3.65 iROAS	Ecommerce
19	Demand Gen statics (@runtraffic)	Demand Gen: Pre-Christmas statics	94.8% new buyers	Ecommerce
20	Noom	Demand Gen: Social videos reused in Demand Gen	+19% vs BAU (vendor)	Subscription

The 20 Google Ads examples, by campaign type

Each entry links to its result page with the full before, after and caveats.

For every Google result we hold, including the ones that didn't make this list, see the full results database for this platform.

Search examples (1-7)

1. Lavender: single keyword ad groups and lead form ads

What they did: KlientBoost rebuilt the mental health service's Search account with single keyword ad groups, more non-brand terms, lead form ads and RSAs.

Its 3,800% conversion rise points to a tiny base.

Why it works: Each ad matches its query exactly, and lead forms skip the landing page.

Result: CPA −95%, conversions +3,800% (KlientBoost). A client quote says cost per action halved, so read the headline with care.

Copy this: Give your top non-brand terms their own ad groups and test a lead form asset.

2. Malibu Shirts: high-intent keywords on a flat budget

What they did: Hawke Media rebuilt a barely-used Search account around high-intent keywords and copy built on the brand's value proposition. Budget stayed flat.

Why it works: Dropping curious traffic raises conversion rate without extra spend.

Result: ROAS 0.44x → 2.93x in February, 4.02x by July (Hawke Media). Beach-apparel season overlaps.

Copy this: Keep only search terms with a product or buying word, and lead the headline with why you.

3. iHostPoker: one campaign per event type, travel negatives

What they did: LP split the casino-night rental business into casino party, corporate and fundraiser campaigns and blocked Las Vegas travel searches.

Why it works: "Vegas" queries looked relevant but came from trip planners.

Result: CPA −56%, conversions +450% (LP). Percentages only, from a new account.

Copy this: Find three words that sound like your service but aren't, and add them as negatives.

4. Shoott: campaigns by city, less brand spend

What they did: Pilothouse split the photo-booking platform's campaigns across 60+ US cities and moved budget from brand to high-intent non-brand terms.

Why it works: Budget follows local demand instead of averaging across unlike markets.

Result: Conversions +33% YoY, CPA −22% in 6 months (Pilothouse). Five changes went in together.

Copy this: If one market converts at twice the rate of another, give it its own campaign.

5. Brand search holdout test

Incremental ROAS8.17

New-customer ROAS1.17

Incremental ROAS 8.17; NC ROAS 1.17

Tested whether brand search ads are incremental

Tracking, attribution & incrementalityGoogle@runtraffic

What they did: @runtraffic held back brand search ads for a control group and measured the extra revenue.

Why it works: It shows whether brand buyers would have arrived through organic anyway.

Result: Incremental ROAS 8.17, but new-customer ROAS only 1.17.

Copy this: Run a holdout before changing brand spend, and read new-customer ROAS separately.

6. ChatPDF: tCPA vs tROAS with value rules

What they did: Admiral Media consolidated overlapping campaigns, tested target CPA against target ROAS and added LTV-based value rules.

Why it works: Value rules tell Smart Bidding which subscribers are worth more.

Result: Google ROAS +320% and CAC −38% YoY (Admiral Media). The winning bid strategy isn't named.

Copy this: If customer value varies, pass it to Google and test value-based bidding.

7. Surfshark: 5,000+ campaigns cut to under 50

What they did: Surfshark replaced an account split by device, location, language and match type with under 50 broad-plus-exact campaigns on Smart Bidding.

Why it works: Pooled data gives automated bidding enough signal to learn.

Result: Search revenue nearly doubled in 2025, then grew 2.7X in 2026 (vendor-reported, Think with Google).

Copy this: Merge campaigns that get only a handful of conversions a month.

Shopping examples (8-10)

8. Fashionphile: Shopping restructure plus remarketing lists

What they did: KlientBoost restructured the luxury reseller's Shopping campaigns and added remarketing lists to Search and Shopping.

Why it works: Past visitors get higher bids instead of being treated like strangers.

Result: ROAS +28%, transactions +36% month over month (KlientBoost).

Copy this: Add a remarketing audience to Shopping and bid it up.

9. The Sill: tiered Shopping and cheap non-brand clicks

What they did: Hawke Media ran non-brand search on maximize clicks with very low bids and split Shopping into tiers.

Why it works: Cheap clicks lower the conversion rate needed to hit CPA, and tiers stop best sellers subsidizing dead stock.

Result: ROAS −23% → 403%, CPA −38% while spend grew 10x+ (Hawke Media).

Copy this: Tier Shopping products (best sellers, middle, long tail) with separate bids.

10. GreenPan: Shopping split by product category

What they did: MuteSix split a new account's Shopping campaigns by category to move budget to top sellers.

Why it works: Two products drove 55% of sales in the best Smart Shopping campaigns.

Result: ROAS +5.2%, conversion rate +14.7% month over month (MuteSix). Modest, plausible gains.

Copy this: Give your top two or three products their own campaign.

Performance Max examples (11-16)

11. Droyd: feed work and model-level creative

What they did: Hawke Media tuned the ride-on toy maker's Shopping feed, made creative per model and added YouTube retargeting in Q4.

Why it works: PMax relies on the feed, so better titles reach more relevant searches.

Result: ROAS +248%, CPA −45.3% (Hawke Media). Q4 gifting season inflates it.

Copy this: Rewrite your top 20 product titles to lead with what people search for.

12. Guitar brand: PMax built on special editions

What they did: Power Digital ran PMax and "fake DPA" structures featuring premium special-edition guitars, aimed at new customers.

Why it works: Premium products lift order value; a new-customer focus stops PMax harvesting past buyers.

Result: $500K new-customer revenue at 3x ROAS (Power Digital). Brand unnamed.

Copy this: Give hero products their own asset group.

13. Cat Person: one PMax split into three

What they did: Flighted split a catch-all PMax into Non-Brand, Brand and General, added RSAs and tested tCPA, alongside a new landing page.

Why it works: A catch-all blends cheap brand sales with costly prospecting.

Result: CPA −13% while spend rose 13% on $200,000+ a month (Flighted).

Copy this: Separate brand from non-brand so each has its own numbers.

14. Penn Tool Co.: low-margin products set apart

What they did: LP split 500,000+ products into five groups, isolated low-margin items and ran PMax on Google and Microsoft.

Why it works: Low-margin items had been eating budget.

Result: Revenue +194%, conversions +70% YoY (LP), from a low starting point.

Copy this: Tag margin in feed custom labels and give low-margin items a stricter ROAS target.

15. Dilling: separate PMax for new and returning buyers

What they did: Dilling ran one PMax track for new customers on non-brand searches and another for past buyers.

Why it works: Over 60% of customers buy again, so a mixed campaign keeps chasing them.

Result: ROAS +66%, conversions +44% (vendor-reported, Think with Google).

Copy this: Upload your customer list and run a new-customer goal on one PMax campaign.

16. Taxfix: PMax tested against pure Search

What they did: In the 2022 tax season, the tax app tested PMax for acquisition against its Search-only setup.

Why it works: PMax reaches people before they type the query.

Result: 9% more tax returns at 44% lower cost per conversion (vendor-reported, Think with Google).

Copy this: Test PMax against your Search baseline before moving budget.

Demand Gen and YouTube examples (17-20)

See the Demand Gen results page, and our YouTube ad examples for video creative.

17. Urban Armor Gear: Demand Gen with a lift study

What they did: Common Thread Collective added Demand Gen next to Search, Shopping and PMax and ran a Google lift study.

Why it works: Demand Gen creates demand that later shows up elsewhere.

Result: Headline: +191% purchase conversions. The number to trust: a 13.6% relative lift in sales volume.

Copy this: Judge Demand Gen on lift, not last-click ROAS.

18. Makeup brand: YouTube rebuilt for reach, geo-tested

What they did: Power Digital rebuilt YouTube for reach and frequency by device and age, then ran matched market tests.

Why it works: Prospecting video rarely wins on last click; geo tests show what it adds.

Result: YouTube Prospecting 3.65 iROAS (91% confidence); Demand Gen 2.26 (68%) (Power Digital).

Copy this: Retest low-confidence results before scaling them.

19. Static images in Demand Gen before Christmas

ROAS1.18

New customers94.8%

ROAS 1.18, 94.8% new customers

Demand Gen static ads before Christmas

Static image adsGoogle@runtraffic

What they did: @runtraffic ran static image ads in Demand Gen ahead of Christmas.

Why it works: Gift shoppers are open to new brands, and statics are cheap to make.

Result: ROAS 1.18, with 94.8% of buyers new to the brand.

Copy this: Score Demand Gen on new-customer share, not first-order ROAS.

20. Noom: social videos reused in Demand Gen

What they did: Noom used Smartly to run its short-form social videos in Demand Gen across YouTube, Discover and Gmail.

Why it works: Reusing proven creative is a cheap way to test a placement.

Result: 19% better than business-as-usual social; 46% of conversions from people 55+ (vendor-reported, Smartly).

Copy this: Launch your three best social videos in Demand Gen first.

Patterns across the winners

Split what Google averages

Shoott split by city, iHostPoker by event type, GreenPan and The Sill by product group, Cat Person by brand and non-brand. Each exposed something a blended average hid.

Feed the bidder value, not just conversions

ChatPDF used LTV-based value rules, Penn Tool Co. separated low-margin products, and Dilling and the guitar brand pointed PMax at new customers.

Consolidate when data is thin

Surfshark went from 5,000+ campaigns to under 50. Split where segments differ and have volume; merge where they don't.

Measure upper funnel with holdouts

The brand search test, UAG and the makeup brand all used a control.

Their incremental numbers (8.17 iROAS, 13.6% sales lift, 3.65 iROAS) are modest, which is why they are credible.

Browse more ad examples for the same patterns on Meta and TikTok.

Examples that lost (what not to copy)

Our archive skews toward wins, but several entries show the setups the winners replaced.

Running broad without negatives. Malibu Shirts started at 0.44x ROAS and iHostPoker was paying for Las Vegas trip planners.
Strict CPA targets on a small account. The Sill was under tight CPA targets at −23% ROAS before the switch to low-bid clicks and tiered Shopping.
Letting PMax decide the channel mix blind. In one account @MenachemAni shared, Search took 69% and Discover 26% of PMax spend.
Judging Demand Gen next to Search on the same scale. In one $449,160 account @runtraffic posted, Search returned 9.37 ROAS and Demand Gen 2.66. That gap is normal for prospecting, so test lift before cutting it.
Ignoring a slide. @thedanielbudai shared an account where ROAS fell from 3.71x to 1.9x with no change named. Audit search terms, feed and tracking before touching budget.