If you've ever sponsored a YouTube creator and been disappointed by the ROI, you've already lived through what subscriber count actually measures: not engagement, not influence, not purchase intent. Just historical clicks on a follow button. Many of those followers stopped opening videos two years ago. Some are inactive accounts. Some followed for a single piece of content that has nothing to do with your brand.
This is universally true on creator platforms, but it's especially true on Bilibili — China's YouTube. With 300M+ monthly active users skewed Gen Z and millennials, Bilibili is where Chinese creator marketing happens. And Bilibili exposes three engagement signals that YouTube doesn't, which together let you cut through the noise of follower counts and identify creators whose audiences actually engage.
The single most useful one is coin-per-view ratio. This post explains what it is, why it matters, what threshold to use, and how to compute it for any Chinese creator in a few lines of code.
Why follower count is a lying signal
Three reasons follower counts mislead in creator marketing:
1. Followers are a lagging indicator. Someone followed a creator in 2023 because they liked one video. That doesn't tell you whether they still watch in 2026, whether they engage, or whether they trust the creator's recommendations enough to buy.
2. Followers are gameable. Not everyone games them, but enough creators do that you can't trust raw counts without other signals. Bot followers, follow-for-follow campaigns, paid follower services. China specifically has a robust market for these.
3. The follower-to-engagement ratio varies wildly. A creator with 100k followers and 1M average views per video has fundamentally different audience economics than another creator with 100k followers and 5k average views per video. Both have the same "follower count" — the engagement quality is the actual signal.
This is why every serious creator marketing tool talks about "engagement rate" — which on YouTube is usually computed as (likes + comments) / views. It's better than raw follower count, but on Bilibili you can do meaningfully better.
The three Bilibili-native metrics
Bilibili was designed by anime fans for anime fans, and the engagement system reflects values around quality and creator support that YouTube's flat "like" button never captured. Three metrics that come back from any Bilibili video scrape:
Danmaku (弹幕) — real-time scrolling comments overlaid on the video as users watch. Think livestream chat, but for pre-recorded video. The danmaku count tells you how many people were engaged enough mid-watch to type something. It's a leading indicator of viewing time and attention.
Favorites (收藏) — equivalent to "save for later" or YouTube's bookmark. Strong long-term value signal: high favorites relative to views means people return to this video. Tutorials, references, and definitive content score high here.
Coins (投币) — Bilibili's tipping system. This is the interesting one. Each user gets a small daily allocation of coins (typically 5 per day for active users), and they can "throw" them at videos they want to support. Because coins are scarce by design — you only have a few to spend, ever — coin counts are a strong genuine-appreciation signal.
A user gives a coin to a video they love. They give a coin to a creator they want to keep making content. They don't give a coin to a video they passively watched and forgot. The cost is real (relative to the user's daily allocation), so the signal is real.
Coin-per-view ratio: the single best signal
If I had to pick one metric to evaluate a Bilibili creator, it would be median coin-per-view ratio across their last 20-30 videos.
The math is simple:
coin_per_view = coin_count / view_count * 100 # express as percentage
What I've found from looking at hundreds of Bilibili creators across categories:
| Coin/View % | Audience quality |
|---|---|
| < 0.5% | Passive viewers. Casual scrolling traffic, not engaged. |
| 0.5% – 1% | Average. Normal Bilibili content, decent audience. |
| 1% – 2% | Strong. Genuinely engaged audience. Worth sponsoring. |
| > 2% | Exceptional. Users actively spending limited resources on this content. |
Above 2% is rare. It typically indicates either: (a) genuinely high-quality educational/tutorial content that people return to, (b) a creator with a deeply loyal niche audience, or (c) content that struck a strong emotional/cultural nerve.
For creator vetting, my heuristic is: if median coin-per-view is below 1%, the audience is more passive than the follower count suggests; sponsorship ROI will probably disappoint.
How to compute this for any creator
The data you need: a creator's recent videos with their view and coin counts. Bilibili exposes this through their public API — no auth required. You can use the open-source bilibili-api Python library, or call their /x/space/wbi/arc/search endpoint directly.
If you'd rather skip the API integration entirely, I built a hosted scraper on Apify Store: zhorex/bilibili-scraper. $5 per 1,000 results, free tier covers ~1,000 results.
from apify_client import ApifyClient
from statistics import median
client = ApifyClient("YOUR_APIFY_API_TOKEN")
# Get a creator's last 30 videos
# user_id (mid) is the number in their profile URL: space.bilibili.com/{mid}
run = client.actor("zhorex/bilibili-scraper").call(run_input={
"mode": "user_videos",
"userIds": ["546195"], # 老番茄 (a well-known Bilibili gamer)
"maxResults": 30,
})
videos = []
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
if item.get("type") == "video":
videos.append(item)
# Compute coin-per-view ratio per video
ratios = []
for v in videos:
views = v.get("viewCount", 0)
if views < 1000: # skip videos with too few views to be meaningful
continue
coins = v.get("coinCount", 0)
ratios.append(coins / views * 100)
print(f"Videos analyzed: {len(ratios)}")
print(f"Median coin-per-view: {median(ratios):.2f}%")
print(f"Best video coin-per-view: {max(ratios):.2f}%")
# Categorize
median_ratio = median(ratios)
if median_ratio >= 2.0:
label = "EXCEPTIONAL"
elif median_ratio >= 1.0:
label = "STRONG"
elif median_ratio >= 0.5:
label = "AVERAGE"
else:
label = "PASSIVE"
print(f"Audience quality: {label}")
Run this against any Bilibili creator's user ID and you have a concrete answer about audience engagement quality.
A workflow for vetting creators at scale
If you're building a creator marketing program for the Chinese market, the workflow that works for the teams I've seen using this:
- Gather candidates. From competitor sponsorship lists, from category trending, or from agency recommendations. Aim for 30-50 candidates per round.
-
Pull their recent video portfolios. Use
user_videosmode to get the last 20-30 videos per creator. - Compute aggregate metrics. For each creator: median coin-per-view, median favorite-per-view, median danmaku-per-view, view consistency (standard deviation).
- Filter on quality threshold. Drop anyone with median coin-per-view below 1%. This usually cuts the candidate list by 40-60%.
- Manual review of the survivors. Watch a sample of their videos. Check for content fit. Evaluate sponsorship history (do their sponsored posts feel native or forced?).
- Negotiate from the qualified shortlist.
Total cost using a hosted scraper: ~$5-10 in scraping for a 50-creator vetting round. Compared to agency rates for the same work ($500-2000 per round), the math is obvious once you do it.
Cross-platform creator vetting
Bilibili is not the whole story. If you're vetting creators for a comprehensive China presence:
- Bilibili for video content (gaming, tech, anime, education)
- RedNote (Xiaohongshu) for product-discovery content (beauty, fashion, lifestyle, food)
- Weibo for public discourse and broad reach campaigns
Each platform has different engagement signals. Bilibili has coins; RedNote has saves (similarly scarce intent-to-buy signal); Weibo has reposts and verified-account hierarchy. A creator strong on one isn't necessarily strong on others.
I maintain scrapers for all three on Apify Store under the zhorex profile, with consistent output schemas across the suite. Same pricing model ($5/1000 results), same Apify infrastructure. If you're doing cross-platform creator analytics, the consistency saves integration time.
When this approach fails
Two cases where coin-per-view ratio is a misleading signal:
1. Brand-new creators with very few videos. If a creator has uploaded 3 videos and one went viral with high coins, the ratio looks artificial. Wait until you have 15-20 videos to compute median.
2. Live-stream-focused creators. Bilibili lets creators upload archived live streams. Coin economics are different in livestream context (gifts replace coins). For livestream-heavy creators, you need different analysis.
For everyone else, coin-per-view ratio is the single best signal I've found for vetting Bilibili creator quality at scale.
What this won't tell you
- Whether the audience is geographically right for your campaign (need follower demographics, which require auth)
- Whether the creator has done sponsorships before that flopped (need to scrape their content for promo patterns)
- Whether their audience overlaps with your target customer profile (need cross-reference with other platforms)
Treat coin-per-view as the engagement-quality filter. Everything else still requires manual review or additional data sources.
If you're working on creator marketing for the Chinese market and want to compare notes on what works — drop a comment. I write about Chinese platform analytics (Bilibili, RedNote, Weibo) and the build-vs-buy tradeoffs around them.
Hosted Bilibili scraper: apify.com/zhorex/bilibili-scraper
Other Chinese platform scrapers in the same suite: RedNote for product-discovery content, Weibo for public discourse.
Top comments (0)