Sponsorship ROI: Avg 110K Views But Median 22K — Why Brands Should Budget by Distribution, Not Expectation
Key Takeaways
- · Dhesy 협찬 영상 21,000건 표본에서 평균 조회수 109,876회, 중앙값 21,955회로 5배 격차 확인
- · 평균값 기반 ROI 예측은 실제 도달률 대비 최대 5배 과대 추정될 수 있음
- · 편당 800~2,000만 원 예산 구간에서는 분산 전략의 중앙값 합계가 집중 전략의 5배 우위
- · 고관여 업종(자동차·가전·금융)은 조회수가 아닌 회상률이 KPI일 때 집중 전략도 유효
- · 제안서에는 평균 CPV와 함께 중앙값 CPV를 병기하는 관행이 예산-실집행 괴리를 줄임
Across 21,000 sponsored videos sampled on Dhesy, average view count was 109,876 while median was 21,955. The mean is roughly 5x the median. This post explains why relying on a single "expected value" for sponsorship ROI breaks budget planning, and how distribution-based budgeting fixes it.
What a 5x Gap Between Mean and Median Actually Says
Half of all sponsored videos never cross 22K views. The "average" is lifted by a small number of outliers.
"Average sponsored view is 100K" is a dangerous sentence in a marketing budget meeting. That number is the mean, not the median, and YouTube sponsorships do not follow a normal distribution — they follow an extreme long tail. Here is what the sample shows.
| Metric | Value |
|---|---|
| Mean views | 109,876 |
| Median views | 21,955 |
| Mean likes | 1,947 |
| Mean comments | 194 |
| Mean ER (likes+comments/views) | 1.95% |
| Sample size | 21,000 |
A 5x mean-to-median gap means half of sponsored videos end below 22K. A CPV calculation that assumes a 100K average reach is off by 5x in a median scenario. That is a material forecasting error.
협찬 영상 조회수 분포 (21,000건 표본)
* Source: Dhesy
Why Sponsored Videos Follow a Long Tail
Sponsored clips sit on top of a creator's already long-tailed view distribution and add a further constraint — the brand message. Variance compounds.
협찬 영상 vs 일반 영상 평균 지표
* Source: Dhesy
A small number of top 1% videos absorbs a disproportionate share of total views. JISOO's collaboration music video with Google recorded 98.3M views. BABYMONSTER's episode with Natuur recorded 16.4M. Those two alone exceed the sum of 50,000 median-performing videos.
The implication: expecting "average-level" performance per collaboration is already a statistically biased assumption. The honest expectation sits near the median, with outliers treated as upside.
Expected-Value Budget vs Distribution Budget: Same 50M KRW, Two Scenarios
| Scenario | Concentrated (1 × 50M) | Distributed (5 × 10M) |
|---|---|---|
| Mean expectation | 100,000 views | 100,000 views (5-video sum) |
| Median scenario | 22,000 views | 110,000 views (5-video sum) |
| Worst (bottom 25%) | 5,000 views | 50,000 views (5-video sum) |
| Chance of an outlier | One attempt | 5 attempts |
Recalculated at the median, the distributed plan delivers 5x more aggregate reach than concentration. The distributed plan also retains multiple chances of catching an upside video.
5,000만 원 예산: 집중 vs 분산 시나리오별 조회수
* Source: Dhesy
But Distribution Is Not Always the Right Answer
Per-creator production has a price floor. Below ~10M KRW per clip, brands fall out of the tier of top creators and mid-tier averages drift back down the distribution. The distribution strategy is strongest in the 8–20M KRW per-clip band where 5–8 creators are realistic.
Awareness-driven campaigns add another wrinkle. High-involvement verticals (auto, appliances, finance) often see higher unaided recall from a single long-form immersive clip than from five short mentions. Before applying distribution logic, decide whether ROI is measured in views or in recall.
Takeaways for Marketers
Advertisers should reject single-number ROI justifications. Present three lanes in the proposal: bottom-25%, median, and top-10%. Post-campaign arguments about "why the numbers missed" mostly disappear once the worst case is priced in upfront.
Creators benefit from leading with distribution data during rate negotiations. "My median is 280K, my top 25% is 800K" signals professionalism and rate discipline more than "my average is 500K."
Agencies can normalize median CPV alongside mean CPV on proposals. Median CPV runs 1.5–3x higher than mean CPV but maps far more closely to what campaigns actually deliver.
FAQ
Q. Should budgets use mean or median view counts?
A. Use the median — or the 25th percentile — as the baseline. Treat the mean as a ceiling, not a target. Because the mean is 5x the median in this dataset, planning to the mean implies 5x overstatement.
Q. When does a distributed strategy win?
A. Distribution wins in the 8–20M KRW per-clip band with 5+ creators, where the brand doesn't drop below top-tier thresholds and still buys multiple shots at the upside.
Q. How is median CPV calculated?
A. Take the creator's 6–12 month median view count, not the mean, as expected reach. Divide the quote by that to get median CPV. It will look higher than mean CPV but predicts delivery more reliably.
Q. Is distribution still better for awareness goals?
A. Not always. Auto, appliances, and finance often see stronger unaided recall from one immersive clip than from five short mentions. If the KPI is recall or brand search volume, concentration can be justified.
Q. Can we count top 1% hits in budget planning?
A. Treat outliers as upside, not as the plan. In 21,000 sampled videos, the top 1% (about 210 clips) absorbs a large share of total views. Budgets premised on outlier repeats have poor reproducibility.
Methodology: Based on 2,336 brands and 6,206 collaborations on Dhesy, with a 21,000-video sample of sponsored uploads (some overlap in counts). Analysis window: 2025-04-01 to 2026-04-15. Source: Dhesy External sources: Nielsen 2024 Influencer Marketing ROI Report, HubSpot 2025 Media Mix Modeling Guide Updated: 2026-04-16