This case study is a composite. It is built from the anonymized performance data of three top-100 beauty-brand activations that launched on Roblox between Q3 2025 and Q1 2026, all sharing the same agency-of-record playbook and broadly comparable budgets. We composite-aggregate to protect commercially sensitive data the brands shared in confidence, while preserving the operational detail other brand teams need in order to learn from the campaign. Where a number is a median across the three, we say so. Where a number is from a single campaign, we mark it as such.
The brief
A global beauty brand (mass-market cosmetics, North America HQ, annual marketing budget low-nine-figures) approached its agency in March 2025 with three objectives for a Roblox activation: drive awareness among Gen Z and Gen Alpha, ship a wearable UGC product that lived on the avatar long after the campaign ended, and produce a defensible measurement framework that finance would accept. Critically, the brand did not require a direct-sales tie. Brand leadership was clear: the goal was reach and resonance with a younger cohort the brand was structurally losing to TikTok-native challenger labels.
The constraints were also explicit. Total all-in budget was capped at $1.1M. Time-to-launch from kickoff was 14 weeks. The experience had to be live for the full back-to-school window (August through early October). And the brand's product placement could not be intrusive in a way that would generate negative creator backlash.
The build
The agency proposed a Tier 1 tentpole rather than an embedded placement. The reasoning: the brand needed creator content depth that an embedded placement could not produce, and the 14-week runway was just enough to ship a small purpose-built world rather than retrofit a partnership. The experience was a social hub built around a styling-and-photography loop. Players entered a stylized "city block" environment, customized an avatar look from a rotating wardrobe (including the brand's UGC items), took photos at Instagram-style backdrops, and shared looks with friends through a gallery mechanic with daily voting.
Build-side numbers from the composite:
| Line item | Budget | % of total | Notes |
|---|---|---|---|
| World build (production studio) | $520K | 47% | 14 weeks, ~8 FTE-equivalents |
| UGC items (12 SKUs) | $45K | 4% | 4 hats, 4 hair accessories, 4 face accessories |
| Paid media (in-platform) | $180K | 16% | Discovery placements, sponsored sort, ad impressions |
| Paid media (off-platform) | $95K | 9% | TikTok, YouTube pre-roll, paid creator amplification |
| Creator partnerships (organic) | $165K | 15% | 34 creators, average $4.8K per deliverable |
| Measurement & survey panel | $48K | 4% | Pre/post brand-lift study via third-party panel |
| Contingency & operations | $47K | 4% | Held against post-launch live-ops |
| Total all-in | $1.10M | 100% | On-budget at launch |
The creator seeding plan (the part most brands skip)
The single highest-leverage decision in the campaign was the creator seeding strategy, and it is the part most brand teams under-resource when they brief their first Roblox activation. The agency identified 34 creators across YouTube and TikTok with audience composition matching the brand's target demographic and pre-existing Roblox-content track records. The creators were briefed and given sandbox access two weeks before public launch.
Creator tier breakdown:
- 6 hero creators (1M+ subs): paid $15K-$30K each for a long-form video deliverable. Briefed 14 days pre-launch.
- 14 mid-tier creators (200K-1M subs): paid $3K-$8K each. Briefed 10 days pre-launch.
- 14 niche creators (under 200K, specialized beauty or roleplay audiences): paid $500-$2K each, sometimes in UGC items only. Briefed 7 days pre-launch.
Every creator received the same brief: play the experience, react honestly, do not pitch product. The brand and agency held disclosure standards higher than required, with explicit "sponsored" callouts in every video despite no legal requirement on creators whose compensation was UGC-only. The honesty mandate is the part the brand later credited as the single biggest factor in the campaign's earned-media outperformance.
Launch and the first 72 hours
The experience went live at 9 AM Pacific on a Tuesday in mid-August. The first hero creator video dropped 4 hours later; five more in the first 24 hours; the rest within 96 hours. Per-hour metrics from the first 72 hours, composite-averaged:
| Time window | Cumulative visits | Peak CCU | Notable trigger |
|---|---|---|---|
| Hour 0-6 | 22K | 1.4K | Paid-discovery placement live |
| Hour 6-24 | 140K | 5.8K | First hero creator video published |
| Hour 24-48 | 510K | 14.2K | 5 hero + 8 mid-tier videos live |
| Hour 48-72 | 1.18M | 27.5K | Reached Roblox front-page Rising sort |
The 72-hour curve illustrates a pattern we now see consistently across successful brand launches. Paid media bought the first 2-3 days of trial; creator content extended the curve into week 2; front-page algorithmic placement (the Rising sort or genre-specific Trending) extended it into months 2-3. Activations that fail to seed creator content in the first 96 hours rarely reach the algorithmic-amplification stage, because the platform's discovery algorithm needs sustained engagement velocity to surface a new experience.
30-day results
Headline outcomes from the composite, at the end of the first 30 days:
| Metric | Result | vs Tier 1 median (from benchmarks report) |
|---|---|---|
| Cumulative visits | 12.4M | 2.95x median (median 4.2M) |
| Peak concurrent users | 38.7K | Top decile (median 8-45K range) |
| Median dwell per visit | 19 minutes | 1.73x median (median 11 min) |
| Total minutes played | 235M | ~5x implied median |
| UGC items minted (paid) | 1.8M unique buyers | n/a — not benchmarked at tier level |
| Cost per visit (CPV) | $0.089 | Top quartile (median $0.40) |
| Cost per engaged minute (CPEM) | $0.0047 | ~8x better than median |
| Earned-to-paid media multiple (EMM) | 2.9x | Vertical median 2.4x |
| Brand recall lift (aided, vs control) | +7.2 points | Above the 4-6 point industry threshold |
The numbers above do not exist in isolation. The single most important read is the joint distribution: high visits AND high dwell time AND high recall lift AND a positive EMM. Any single metric can be cherry-picked. Brand teams that pre-commit to all four before launch and report against all four after launch tend to produce campaigns that finance trusts the next year.
What worked
- The creator seeding ratio. 15% of total budget to organic creator partnerships is roughly double the median activation. The agency's argument to the brand was that paid media alone could not produce the dwell numbers the brand needed for a defensible recall study. Post-campaign analysis confirmed: creator-driven traffic dwelled 2.3x longer per visit than paid-discovery traffic.
- The UGC items priced for ubiquity, not yield. The 12 UGC SKUs were priced at 5-25 Robux ($0.05-$0.30 retail equivalent). The brand did not optimize for UGC revenue; it optimized for the number of avatars wearing brand-marked items outside the experience. The 1.8M unique buyers represent approximately 1.8M earned-media impressions per day on Roblox for every day those items remain on those avatars.
- Honest creative briefing. The brand instructed its agency not to ship a "playable ad". The experience had to stand on its own merits as something a Roblox player would choose to play if no brand were attached. The brand presence was environmental (signage, branded products in-world, the UGC wardrobe) rather than scripted into the loop. Post-campaign sentiment analysis of creator commentary found 78% positive vs 22% mixed/negative; comparable brand-activation campaigns average a 40/60 split.
- Launch-window discipline. The agency moved the launch date twice during pre-production to avoid colliding with two scheduled platform tentpoles. Both moves cost build-team overtime; both were defensible against the alternative of launching into a 30-50% traffic suppression window.
What did not work
- The photo-sharing mechanic underdelivered. The intended virality loop, where players' looks would be shared out of the experience to social platforms, accounted for less than 4% of attributable off-platform impressions. Roblox players screenshot via the native client and share to Discord, not Instagram. Future activations from the same brand have de-prioritized this mechanic.
- The off-platform paid media underperformed in-platform paid. The $95K off-platform spend (TikTok, YouTube pre-roll) returned a measured CPV roughly 2.8x worse than the $180K of in-platform paid placements. The brand reallocated the mix on a follow-up campaign to 85/15 in-platform.
- The 90-day persistence was Tier-2-typical, not Tier-1. CCU at day 90 was 19% of peak — within the 22% vertical median but below the top-quartile (35%+) the campaign aspired to. The agency post-mortem attributed this to a thin post-launch live-ops budget and the lack of a scheduled content drop in weeks 5-7.
The framework finance accepted
The brand's CFO had one specific concern before approving the campaign: how would a $1.1M Roblox spend be defended in the year-end marketing-effectiveness review against alternative spend in paid social or out-of-home? The agency proposed and the brand adopted a four-line scorecard:
- CPM-equivalent against TikTok benchmark:calculated against the brand's own historical TikTok CPM. The campaign's effective CPM of $89 was 4-5x above the brand's TikTok baseline, but adjusted for dwell (TikTok averages ~7-second view times; the campaign averaged 19 minutes), the cost-per-engaged- second was approximately 1/120th that of TikTok.
- Earned-media valuation: third-party-priced at $3.2M in equivalent paid value. Conservative methodology; the brand chose not to count UGC-item daily wears as earned impressions to maintain credibility with finance.
- Brand-lift study results: a +7.2 point lift on aided recall and a +4.1 point lift on purchase intent, both significant at the 95% confidence level on n=4,800 panel respondents.
- Year-end revenue attribution: measured via marketing-mix modeling at the brand's annual review, not in the campaign post-mortem. The campaign was directionally credited with a measurable lift in the brand's 14-24 demographic household penetration in the back-to-school window.
Takeaways for brand teams briefing a first activation
- Budget creator amplification at 15-25% of total, not 5%. The asymmetric returns on this line item are the single most consistent finding across our database.
- Pre-commit to a four-metric scorecard (reach, dwell, earned, recall) before launch. Defending a campaign on visits alone is a bad position to be in by Q1 of the following year.
- Treat UGC items as earned-impression vehicles, not yield drivers. Pricing them for ubiquity produces a multi-month earned tail that no other format on Roblox can match.
- Move the launch date if a platform tentpole is in the window. The cost of moving is always lower than the cost of launching into a 30-50% traffic suppression.
- Budget 8-12% of total for post-launch live-ops. The difference between Tier-1-median and Tier-1-top-quartile 90-day persistence is almost entirely a function of whether the brand showed up again in weeks 5-7.
Brand teams running activations on RoLearn use the Activation Archive for competitive context and the Brand Workspace's Mission Control to track all four scorecard metrics from launch through 90-day post. See the ROI Benchmarks 2026 report for tier-level baselines.
Methodology and disclosure
This case study is a composite of three anonymized beauty-brand activations launched between Q3 2025 and Q1 2026, all with total budgets between $0.95M and $1.25M and all built by US-headquartered agencies. Operational details (creator counts, build timeline, budget breakdown) are composite-averaged; headline outcome metrics (visits, dwell, CPV, CPEM, EMM, recall lift) are reported as composite medians. Brand-lift figures cite third-party-administered panel studies (Disqo or Lucid). No brand or agency is identified and no individual campaign's data is reported alone. The brands and agencies represented in this composite reviewed and approved the published figures prior to publication.