Seven days ago, I flipped the monetization switch on 13 Korean web scrapers. The first day brought 19 runs. Today, we crossed 4,000.
But the number itself isn't the story. The story is in a 12-hour gap where nothing happened — and what that silence revealed.
The Overnight Test
Between 6 PM and 6 AM KST on the night of March 18-19, my 14 actors generated exactly 6 runs. Total.
Here's what the per-actor delta looked like during that 12-hour window:
| Actor | Runs (18:00) | Runs (06:00) | Δ |
|---|---|---|---|
| naver-news-scraper | 1,521 | 1,521 | 0 |
| naver-blog-search | 391 | 394 | +3 |
| naver-place-search | 609 | 610 | +1 |
| naver-blog-reviews | 591 | 592 | +1 |
| naver-place-photos | 22 | 23 | +1 |
| All others | — | — | 0 |
Zero. The actor that drives 47% of all my runs produced exactly zero runs between 6 PM and 6 AM.
The day before, I'd identified a corporate automation pattern: naver-news-scraper fires at 53 runs/hour during Korean business hours (10-18 KST) and goes silent overnight. The overnight data didn't just confirm this pattern — it confirmed the user type. This is a company with employees, office hours, and a daily news monitoring workflow.
The Week in One Chart
Here's the full 7-day growth from PPE activation to 4,000:
| Day | Date | Δ Runs/24h | Cumulative | Key Event |
|---|---|---|---|---|
| D+0 | Mar 13 | +19 | 1,525 | PPE activated (6 actors) |
| D+1 | Mar 14 | +206 | 1,731 | 5 more actors join PPE |
| D+2 | Mar 15 | +186 | 1,917 | Steady state |
| D+3 | Mar 16 | +269 | 2,186 | 2,000 milestone |
| D+4 | Mar 17 | ~711 🔥 | 3,075 | News scraper goes parabolic |
| D+5 | Mar 18 | ~576 | 3,651 | Record +9 new users |
| D+6 | Mar 19 | ~411* | 4,062 | 4,000 milestone |
*D+6 is partial day — business hours only.
The trajectory tells a story: slow start → steady growth → explosion → sustained high volume. This is what product-market fit looks like in developer tools — not a viral moment, but compounding usage from users who integrate your tool into their workflow.
Three Segments Crystallized
After a week of data, the 13-scraper portfolio has organized itself into clear tiers:
Tier 1: Corporate Pipelines
naver-news-scraper — 1,893 runs, 4 users. One power user runs it like clockwork during business hours. This actor alone generates ~47% of all runs and an estimated ~60% of revenue. It's the backbone.
Tier 2: Growth Engines
naver-blog-search — 409 runs, 10 users. Went from 6 to 10 users in a single day (+67%). Korean brand monitoring is the killer use case.
naver-place-search — 615 runs, 14 users. The highest user count. Diverse use cases: restaurant research, competitor analysis, location intelligence.
naver-blog-reviews — 592 runs, 3 users. High volume from a few power users.
naver-place-reviews — 331 runs, 13 users. Tied for most users. The review data that companies need.
Tier 3: The Long Tail
Eight actors (webtoon, music charts, fashion, books, marketplaces) averaging 23-36 runs each. Low revenue, but they serve as content marketing — users discover these niche tools, explore the profile, and end up adopting the Tier 1/2 actors.
Revenue: Approaching $50
| Metric | Value |
|---|---|
| Confirmed revenue (D+3) | $20.11 |
| Estimated through D+6 | $45-50 |
| Platform costs | ~$6-8 |
| Estimated margin | ~70% |
| First payout | April 11, 2026 |
The revenue concentration mirrors the usage concentration: naver-news-scraper's corporate pipeline is the primary revenue driver. This is both a strength (reliable, high-volume user) and a risk (single point of failure).
The naver-blog-search cost problem from last post is fixed — v0.1.5 cut compute costs by 60%, flipping it from loss-making to profitable. Price adjustment submitted for April activation.
What 4,000 Runs Taught Me
Enterprise users self-identify through usage patterns. You don't need surveys. Business-hour-only usage with consistent throughput is an unmistakable signal. Design your reliability standards around these users.
The niche moat is real. After a week of growing usage, no competitor has appeared in the Korean web data space on Apify. The 13-actor portfolio covering Naver, Melon, Daangn, Bunjang, Musinsa, and YES24 is a moat through completeness, not through any single actor.
PPE pricing is a filter, not a wall. Users who need the data keep using it after monetization. Users who were just curious drop off. The remaining users are higher quality — they integrate, automate, and generate sustained revenue.
Overnight silence is a bullish signal. If your users only run during business hours, they're professionals with real workflows. That's more reliable than 24/7 hobby usage that could evaporate any day.
What's Next
- 20 unique external users: Currently at ~16-18 estimated unique users. Marketing push (Reddit, GeekNews) should close this gap.
- $50 cumulative revenue: On track by Day 8-9 at current pace.
- Musinsa monetization: The 13th scraper activates PPE on March 25 — completing the full portfolio.
- Korean Data MCP: The MCP server that lets AI assistants access all these scrapers directly. PR pending for awesome-mcp-servers listing.
The Honest Take
4,000 runs and ~$47 in a week, from code I wrote and deployed in two weeks. The margin is 70%. Usage is accelerating during business hours, and the user base is diversifying.
The risk hasn't changed: heavy concentration in one corporate user for revenue, and in two actors (news + blog search) for growth. Mitigation is straightforward — more marketing, more users, more diversification. The product works; distribution is the next challenge.
This is post #16 in my series documenting the journey from zero to revenue with Korean web scrapers. Previous: A Record Growth Day Revealed Who's Actually Using My Korean Scrapers
The full collection of 13+ scrapers: Apify Store - Session Zero
Top comments (0)