Files

T

koriweb d8a80f6272 chore(wiki): dangling 링크 canonical 정규화 (768파일/1200건)

이름만 다른(표기 변형) [[위키링크]]를 대상 문서의 canonical 제목으로 치환해
끊겼던 1,200개 링크를 연결. 제목/파일명 정규화 일치만 적용하고 별칭 매칭은
과병합 위험으로 제외(애매성 가드). 원본은 _link_reconcile_backup/ 에 백업.
도구: Datacollect/scripts/link_reconcile_apply.mjs

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

2026-06-08 12:24:15 +09:00

6.3 KiB

Raw Blame History

id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, tech_stack

title

Support Platforms

매 한 줄

"매 player support 의 retention 의 hidden lever". 2026 game support 의 in-app chatbot (LLM-powered) → human escalation → CRM 의 multi-tier system. Helpshift, Zendesk, Intercom, Salesforce 매 dominant — 매 mobile F2P (Genshin, Royal Match) 매 24/7 multilingual coverage 매 churn 의 reduce.

매 핵심

매 support tier

Tier 0 (self-serve): FAQ, in-app help center.
Tier 1 (LLM bot): Claude/GPT-5-powered chatbot — 매 80% 의 query 의 deflect.
Tier 2 (human agent): refunds, account recovery, complex bugs.
Tier 3 (specialist): payment disputes, banned accounts, exploit investigations.

매 platform 비교

Helpshift: 매 mobile-native, in-game SDK, AI-first (2026 Claude integration).
Zendesk: 매 enterprise, omnichannel, mature ticket workflows.
Intercom: 매 conversational, product-led, Fin AI.
Salesforce Service Cloud: 매 CRM-integrated, 매 large publishers.

매 응용

Genshin Impact — Helpshift in-app + live chat for refunds.
Royal Match — AI bot deflection 매 90%+ — 매 cost-per-ticket 의 dramatic reduction.
Riot Games — custom Salesforce + community moderators.

💻 패턴

Helpshift SDK Init

import Helpshift from '@helpshift/react-native-helpshift';

Helpshift.install({
    platformId: process.env.HS_PLATFORM_ID,
    domain: process.env.HS_DOMAIN,
    enableInAppNotifications: true,
    enableLogging: __DEV__,
});

Helpshift.setUser({
    userId: player.id,
    email: player.email,
    name: player.displayName,
});

Contextual Metadata Attach

function openSupport(reason: SupportReason) {
    Helpshift.showConversation({
        cifs: {
            level: { type: 'number', value: player.level },
            spend_lifetime: { type: 'number', value: player.totalSpendUSD },
            last_purchase: { type: 'date', value: player.lastPurchaseTimestamp },
            crash_count_24h: { type: 'number', value: getRecentCrashCount() },
            client_version: { type: 'string', value: APP_VERSION },
            device: { type: 'string', value: Device.modelId },
        },
        tags: [reason, player.vipTier],
    });
}

LLM Bot Deflection (Claude)

import Anthropic from '@anthropic-ai/sdk';

const client = new Anthropic();

async function botReply(userMessage: string, playerContext: PlayerContext) {
    const response = await client.messages.create({
        model: 'claude-opus-4-7',
        max_tokens: 500,
        system: [
            {
                type: 'text',
                text: SUPPORT_SYSTEM_PROMPT + KB_CONTENTS,
                cache_control: { type: 'ephemeral' },
            },
        ],
        messages: [
            {
                role: 'user',
                content: `Player context: ${JSON.stringify(playerContext)}\n\nQuery: ${userMessage}`,
            },
        ],
    });
    const reply = response.content[0].text;
    if (reply.includes('[ESCALATE]')) {
        return { escalate: true, reply: reply.replace('[ESCALATE]', '') };
    }
    return { escalate: false, reply };
}

Refund Eligibility Check

async function evaluateRefund(ticket: Ticket): Promise<RefundDecision> {
    const player = await getPlayer(ticket.userId);
    const purchase = await getPurchase(ticket.purchaseId);
    
    if (Date.now() - purchase.timestamp > 14 * 24 * 3600 * 1000) {
        return { approved: false, reason: 'OUTSIDE_WINDOW' };
    }
    if (player.refundCount30d >= 2) {
        return { approved: false, reason: 'REFUND_ABUSE_FLAG' };
    }
    if (purchase.consumed && purchase.consumedAmount > 0.5 * purchase.value) {
        return { approved: false, reason: 'CONSUMED' };
    }
    return { approved: true };
}

Ticket Routing

function routeTicket(ticket: Ticket): Queue {
    if (ticket.tags.includes('payment')) return 'tier3-payments';
    if (ticket.player.vipTier >= 5) return 'tier2-vip';
    if (ticket.tags.includes('account-recovery')) return 'tier3-trust';
    if (ticket.bot_confidence > 0.85) return 'tier1-bot';
    return 'tier2-general';
}

Sentiment-Triggered Escalation

async function checkSentiment(message: string, ticket: Ticket) {
    const sentiment = await analyzeSentiment(message);
    if (sentiment.score < -0.7 || sentiment.flags.includes('threat')) {
        await escalate(ticket, 'tier3-trust');
        await notifySlack('#support-escalations', ticket.id);
    }
}

매 결정 기준

상황	Approach
mobile F2P, in-game support	Helpshift
omnichannel (web + game + email)	Zendesk
product-led, conversational	Intercom
enterprise CRM 매 integrated	Salesforce Service Cloud
custom 매 needed	Helpshift API + custom UI

기본값: Helpshift + Claude-powered bot deflection — 매 cost-effective 매 mobile games 의 industry standard.

🔗 Graph

부모: Game Design · LiveOps
Adjacent: Player Retention

🤖 LLM 활용

언제: 매 Tier 1 deflection (FAQ-style queries), translation, sentiment analysis, response drafting for human agents. 언제 X: 매 refund decision 의 final, ban/unban, payment dispute — 매 human judgment 의 필수.

❌ 안티패턴

Bot 만: 매 escalation path 의 X → user frustration.
No context attach: 매 agent 의 player level/spend 의 X → resolution time 의 increase.
24h 의 SLA 의 X: mobile player expect <2h response.
Single language: 매 global game 의 매 multilingual support 의 mandatory.

🧪 검증 / 중복

Verified (Helpshift docs, Zendesk Game Industry report 2026, Pocket Gamer Connects talks).
신뢰도 A.

🕓 Changelog

날짜	변경
2026-05-08	Phase 1
2026-05-10	Manual cleanup — support platform comparison + 6 implementation patterns

6.3 KiB Raw Blame History