--- title: Samsung Korea x Performance Max: how a 5-week test produced +2.5x conversions and -60% cost per conversion | RGM® url: https://realgrowthmatters.com/learn/case-studies/samsung-galaxy-campus-performance-max/ updated: 2026-06-10 source_html: https://realgrowthmatters.com/learn/case-studies/samsung-galaxy-campus-performance-max/ --- - **Story:** Samsung Electronics Korea ran a 5-week controlled experiment in 2023 testing Performance Max layered on existing Display + DV360 + paid-search base. Test cell: +2.5x incremental conversions and -60% cost per conversion. - **Why it matters:** A clean, rigorously-designed test of Google Performance Max in a real e-commerce setting. Shows that Performance Max can produce incremental lift on top of mature paid-media stacks, not just replace them. - **Takeaway:** Test Performance Max as a layer on top of existing campaigns — not as a replacement — to measure incrementality cleanly. - **Takeaway:** 5-week test windows with proper hold-out cells are the minimum for credible PMax incrementality reads. - **Takeaway:** Conversion-cost reductions of 50%+ are real in PMax when the audience and creative assets are sufficient. ## Samsung KR PMax — the four-step story S Situation Samsung KR wanted to test PMax incrementality cleanly In 2023, Samsung Electronics Korea was running a mature paid-media stack (Display + DV360 + paid search) targeted at college and graduate students. They needed to know whether Google's Performance Max would lift the existing stack or just cannibalize it. T Task Design a controlled 5-week test Build a hold-out cell. Include a 2-week PMax ramp period. Measure incremental conversions and conversion cost cleanly — not just total conversions, which would mix in baseline traffic. A Action Run a 5-week controlled experiment in Korea Samsung Electronics Korea ran a 5-week test of PMax layered on top of existing Display + DV360 + search base for Galaxy Campus Store. 2-week ramp period included to let PMax learning algorithms stabilize. R Result +2.5x conversions, -60% cost per conversion in test cell Test cell showed +2.5x incremental conversion lift and -60% cost per conversion versus the control. Documented through Think with Google as a rigorous PMax incrementality case study. ## Samsung Korea PMax at a glance +0x Test cell conversion lift Incremental lift on top of existing stack Source: Think with Google case study -0% Cost per conversion Test cell reduction vs control Source: Think with Google case study 0 wks Test duration Including 2-week Performance Max ramp Source: Samsung test design 0 Cells in test Control vs test (PMax layered) Source: Test methodology 0 Country South Korea — Galaxy Campus Store Source: Samsung KR 0 Target audience College & graduate students Source: Galaxy Campus Store positioning #### Quick facts BrandSamsung Electronics — Galaxy Campus Store CountrySouth Korea AudienceCollege & graduate students (Korea) Product setGalaxy mobile + IT products Test duration5 weeks (including 2-week Performance Max ramp) Test cell incremental conversion lift+2.5x Test cell cost per conversion-60% vs control Existing media stackDisplay + DV360 + paid search (mature, already optimized) **Honest note** The +2.5x conversion lift and -60% cost per conversion are Samsung’s own test-cell measurements against a control cell, documented through Think with Google. The methodology (5-week window, 2-week ramp, control comparison) is reasonable and the directional finding (PMax can produce incremental lift on top of an established media stack) is well supported. Specific percentage figures from a single test should be read as evidence rather than as universal applicability to all advertisers. ## Where Samsung Korea's paid media was in 2023 In 2023, Samsung Electronics Korea was running a mature paid-media stack to drive sales for Galaxy Campus Store — an e-commerce destination targeting Korean college and graduate students. The existing stack included Display advertising, DV360 (Google’s programmatic platform), and paid search across multiple Galaxy product lines. The performance was already optimized and the team had a clear baseline. The question was whether adding Google Performance Max (a goal-based campaign type that uses Google’s machine learning to allocate spend across Search, Display, YouTube, Discover, Gmail, and Maps automatically) would produce incremental lift on top of the existing stack — or just cannibalize it. The distinction matters: incremental lift is real new revenue; cannibalization is the same revenue at higher cost. The team designed a clean test to find out. ## The test design Samsung Korea designed a controlled 5-week experiment with two cells: - **Control cell:** the existing Display + DV360 + paid-search stack, running as normal. - **Test cell:** the existing stack plus Performance Max layered on top, running concurrently. A few methodology choices made the test credible: **Why this kind of test methodology matters**Most advertisers measure PMax (or any new channel) by looking at total conversions in PMax-active campaigns and assuming that’s the lift. That measurement conflates baseline traffic with incremental traffic and overstates the new channel’s impact. The right measurement compares a control cell (no change) against a test cell (with the new channel added). The difference is the actual incremental impact. Samsung Korea’s test design is the model for how to measure new-channel incrementality, and it’s the kind of methodology most advertisers skip because it requires hold-out cells that feel like wasted reach. ## What the test showed The test cell produced +2.5x incremental conversion lift versus the control cell. Cost per conversion in the test cell dropped 60% versus the control. The PMax layer wasn’t just shifting conversions from existing channels — it was producing genuinely new conversions at materially lower cost per conversion. The conversion-cost finding is particularly important. The cheaper cost-per-conversion suggests PMax was finding incremental audience the established stack wasn’t reaching efficiently — Discover, YouTube, and Gmail surfaces that the Display + DV360 + Search stack hadn’t been targeting as effectively. The combination of more conversions and lower cost per conversion is the textbook signature of genuinely additive incremental media, not channel cannibalization. ## What other advertisers could learn The structural lessons from Samsung Korea's test transfer to other advertisers considering PMax (or any new layered media channel): - Test incrementally, not totally. Compare test vs control, not test vs nothing. The total-conversion measurement overstates the new channel’s impact and produces decisions based on inflated numbers. - Build in a ramp period. Machine-learning-driven channels need time to optimize. Measuring during the ramp produces artificially poor results; measuring only after the ramp produces clean signal. - Measure both conversion volume and conversion cost. Volume alone can be a cannibalization signal; cost-per-conversion movement reveals whether the new channel is finding incremental audience or just shifting attribution. - Plan for 4-6 week tests minimum. Shorter tests don't produce statistically meaningful results in most paid-media contexts. The Samsung 5-week design (with 2-week ramp) is roughly the minimum credible duration. ## How RGM thinks about PMax and new-channel testing When clients ask whether they should add Performance Max to their existing paid-media stack, the Samsung Korea test is the example we point to first — not because the +2.5x and -60% figures will replicate exactly (they won’t; every advertiser’s baseline is different), but because the test design is the structural template for how to find out. The methodology matters more than the specific numbers. Advertisers that copy the methodology will get accurate readings for their own businesses; advertisers that copy only the headline numbers will be disappointed when their results don't match. The harder honest lesson is about media-mix evolution. Performance Max isn't a permanent answer — it's a specific machine-learning channel that may perform very well in 2024-2025 and may be supplanted by a different channel or a different machine-learning approach in 2027-2028. The structural skill is test-and-measure discipline, not allegiance to any specific channel. We tell clients to build the test-and-measure capability internally so they can evaluate whatever new channel comes next, rather than depending on case studies of channels that are already mature. ## Frequently asked questions What does “incremental” actually mean here? Incremental conversions are the conversions that wouldn’t have happened without the new media layer. The standard way to measure incrementality is to compare a test cell (with the new layer) against a control cell (without it). The difference between the two cells is the incremental impact. Samsung Korea’s +2.5x lift is the test-cell conversion volume divided by the control-cell conversion volume during the same period. How is Performance Max different from regular Google Ads? Performance Max is a campaign type where the advertiser sets a conversion goal and a budget, and Google’s machine learning allocates spend across all of Google’s ad surfaces (Search, Display, YouTube, Discover, Gmail, Maps) automatically. The advertiser provides assets (text, images, videos) but doesn’t directly control which surface or keyword the spend goes to. The tradeoff is less control in exchange for machine-learning optimization across the full Google inventory. Why did Samsung need a 2-week ramp period? Performance Max's machine-learning algorithm needs time to learn which audiences, surfaces, and creative combinations produce conversions for a specific advertiser. During the first one to two weeks, the algorithm is exploring and producing variable performance. After the learning period, performance stabilizes and accurate measurement becomes possible. The 2-week ramp in the Samsung test was conservative but appropriate. Will these results replicate for other advertisers? The directional finding (PMax can produce incremental lift on top of an established media stack) is well supported and likely replicates. The specific magnitude (+2.5x conversions, -60% cost per conversion) is highly advertiser-specific and depends on baseline performance, audience targeting, creative quality, and category. Advertisers should expect directional similarity but should run their own tests to find their own numbers. Should I test PMax in 2026? If you have a mature paid-media stack and aren't running PMax, yes — testing for incrementality is straightforward and the upside is real. If you're already running PMax, the structural lesson is to test the next new channel (whatever it is) with the same methodology Samsung Korea used. The test-and-measure discipline is the durable skill; allegiance to any specific channel isn't. ### Sources & references - [Think with Google — Samsung Korea Performance Max case study](https://www.thinkwithgoogle.com/) — Google's documentation of the Samsung Korea test design and results. - [Google Performance Max](https://support.google.com/google-ads/answer/10724817) — Google's official documentation of the Performance Max campaign type. - [Samsung Electronics Korea](https://www.samsung.com/sec/) — Samsung Korea’s corporate site. ## Related [#### All case studies The full case-study library.](/learn/case-studies/)[#### Qualcomm Developer First Another technology-brand digital-marketing redesign.](/learn/case-studies/qualcomm-developer-first/)[#### Promat B2B precision marketing in a different category.](/learn/case-studies/promat-passive-fire-protection-campaign/)[#### Salesforce Holiday B2B SaaS marketing with annual data-franchise model.](/learn/case-studies/salesforce-holiday-campaign/)