Live-operations content moderation for a Korean MMO gaming publisher

Client profile

The client is a Seoul-headquartered gaming publisher operating a leading massively multiplayer online role-playing game (MMORPG) with approximately 3.8 million monthly active players across South Korea, Japan, Southeast Asia, North America, and Europe. Peak concurrent users regularly exceed 480,000. The game monetizes through microtransactions and battle passes, and the health of the in-game economy and player community is directly linked to revenue.

The challenge

The client's Trust & Safety function was significantly under-resourced relative to game scale. Player reports of toxicity, harassment, cheating, and account trading were queuing for 47+ minutes before reaching a moderator during peak hours. Weekly user-generated toxicity events averaged 12,400 in the six-month baseline period. Player retention analytics showed a strong correlation between exposure to unmoderated toxicity and 30-day churn, with an estimated $4.2M in monthly revenue at risk from toxicity-driven player departure.

Cultural context added complexity. Korean gaming culture has specific patterns of coded language, in-group harassment, and gambling-adjacent account trading that require Korean-native moderator judgment. Japanese and Southeast Asian player bases have distinct patterns. Non-Korean moderators consistently missed context-dependent violations, and Korean-only moderators struggled with international player disputes.

The client had previously used a US-based content moderation vendor with a global moderator pool. Quality was inconsistent, and the vendor's per-hour pricing model did not scale efficiently as the client's live-service events drove episodic 3-4x volume spikes. The client had also experimented with in-house Korean moderators but attrition was high (68% annualized) due to the psychologically demanding nature of the work.

Why Corpshore Vietnam

Corpshore Vietnam was selected on three qualifications: our Korean-language moderator bench in the Hanoi facility (30+ Korean-fluent professionals sourced from Vietnamese universities with Korean programs and the Korean expatriate community), our documented Trust & Safety wellness framework (30% lower attrition than industry benchmark based on prior gaming engagements), and our ability to integrate AI-augmented moderation via our Corpshore AI division for the client's separate cheat-detection model training program.

Our proposal bundled human moderation with training data for the client's automated moderation model. This dual-track approach was not offered by any of the competing vendors, and directly addressed the client's stated interest in reducing manual moderation load over time as their AI models matured.

The engagement

62 content moderators (28 Korean-native, 12 Japanese, 8 Mandarin, 8 English, 6 Vietnamese for internal team coordination), 8 AI training specialists working on toxicity classification and cheat-detection model improvement, and 4 Trust & Safety leads (2 with formal T&S certifications from prior industry roles). Team operates 24/7 rotating shifts with peak coverage aligned to Korean prime-time gaming hours. All moderators complete 6-week paid training covering platform policy, decision frameworks, wellness protocols, and Korean/Japanese/Mandarin cultural context.

Approach and methodology

Culturally-native moderation. Reports about Korean player behavior are handled by Korean-native moderators. Japanese by Japanese, Mandarin by Mandarin. English handles international disputes and cross-server incidents. This routing is automated and drove a significant quality improvement over the client's previous vendor.

AI-augmented triage. Corpshore AI's toxicity classification model pre-scores every report with confidence and severity. High-confidence high-severity reports route to moderators immediately. Low-confidence reports get sample-audited to continue training the model. This drove a 65% reduction in moderator time on routine cases within 6 months.

Wellness framework. Every moderator has mandatory 15-minute breaks every 90 minutes, on-site psychological support, peer support groups, and rotation off high-intensity content (child safety, extreme graphic content) after 3-hour continuous exposure. Attrition on this account is 24% annualized vs 68% for the client's previous in-house model.

Results

Weekly toxicity reports dropped from 12,400 to 2,800 by week 24 (77% reduction), attributed primarily to the deterrent effect of consistent enforcement. Average moderator response time dropped from 87 minutes end-to-end to 13 minutes. Player retention at 30 days improved by 6.4 percentage points against a control cohort. Estimated revenue protection from improved retention: $2.4M monthly by month 6, $4.1M monthly by month 12.

The engagement extended to include a specialized AI red-team program for the client's next-generation game project, launching in month 15.

Enduring value

The toxicity classification model trained during the engagement is now an owned asset used across the client's expanding game portfolio. The wellness framework methodology is being applied to the client's own internal moderation team. Corpshore is engaged for two additional games launching in 2026-2027.