Jinshun Casino × ATG Slots — LINE OA Full-Stack Campaign
From Meta Ads to LIFF binding to LINE OA full-automation reward delivery — public architecture, CAPI tuning, retention mechanics, and ops gotchas.
RedClaw built an end-to-end stack for Jinshun Casino (gold88.uk) — a legally registered platform under Taiwan's gamerating.org.tw, powered by the ks7 game engine. The system spans Meta Ads → LIFF binding → LINE OA automation with: 5 always-on player campaigns, an 8-tier revival bracket (10K–2M points), CAPI EMQ score lifted from 4 to 9 (lookalike precision +20-30%), a transaction-safe unlimited serial pool, a full fbclid attribution pipeline, and a TG bot for real-time ops. This page publishes the architecture, the gotchas we hit, and verifiable metrics.
What is Jinshun Casino (gold88.uk) and how does it relate to RedClaw?
Jinshun Casino is a RedClaw client — a legally registered points-based gaming platform under Taiwan's gamerating.org.tw (NOT a cash or credit casino), powered by the ks7 game engine. RedClaw owns the full Meta Ads campaign architecture, LIFF player binding, LINE OA automation, and CAPI tracking pipeline. The brand operates on gold88.uk. This page publishes the actual stack, CAPI EMQ optimization methods, and revival bracket design, with verifiable deployment evidence.
The Client
System Metrics (as of 2026-05-24)
Architecture and technical metrics are public. Player-level data is NDA-protected and shared only with qualified leads under NDA.
RedClaw deployed a Meta CAPI tracking pipeline for Jinshun Casino (gold88.uk) that improved Event Match Quality score from 4 to 9 by enriching 6 user_data fields (IP, User-Agent, fbp, fbc, em, ph), driving a 20-30% lift in lookalike audience precision. Architecture is public on this page; technical specs are mirrored in RedClaw's cross-project knowledge base. (Source: RedClaw, 2026-05-24)
Why This Stack Works: 3 Pillars
Jinshun is a legally registered points game under gamerating.org.tw. All ad copy, LIFF flows, and AI chatbot prompts are stripped of "withdrawal" / "cash out" terminology. Pre-deploy grep must return 0 hits. This protects the legal status and keeps Meta ad approval rates stable.
Four LIFF entry flows (bind/me/checkin/deposit) + webhook keyword triggers + auto-payout via transactional serial pool cover the player lifecycle from first-touch to repeat top-up. AI chatbot fields common questions; TG bot pushes ops alerts in real time. Once a player binds, full ROAS attribution lights up.
Front-end fbp/fbc/em/ph + server-side IP/UA — 6 user_data fields fully wired, EMQ score from 4 to 9. fbclid is carried through LP → LIFF → OA end-to-end, fixing the classic "83 Leads/24h but 0 ad-attributed" gap. Lookalike precision +20-30%, scaling no longer stalls.
What This Proves for Future Clients
The Jinshun (gold88.uk) deployment is a live, verifiable case — visit the site to see the work in action.
- 1Can scale Meta campaigns under strict compliance constraints
Taiwan's legally registered points games are a Meta-policy gray zone — one wrong word and the account dies. From day-one copy to the AI chatbot prompt, "withdrawal" / "cash-out" terminology is fully banned. Pre-deploy grep returning 0 is a deployment gate. Jinshun running uninterrupted is the proof.
- 2Can ship an end-to-end LINE OA + LIFF automation stack
We don't just run Meta — we own the entire post-click journey: 4 LIFF flows, AI chatbot, transactional serial pool, revival bracket, TG bot for ops alerts, with fbclid carried end-to-end from LP through OA. A Jinshun player can go from joining the OA to claiming rewards without any human in the loop.
- 3Can fix the "lots of Leads, 0 attribution" gap
Jinshun once hit the classic iGaming scenario: "83 Leads in 24h but 0 ad-attributed in the report." We rebuilt the fbclid pipeline: LP captures → LIFF webview state carries → OA binding persists to Firestore → CAPI Lead event ships fbc back to Meta. Post-fix, the ROAS report finally reflects reality, giving scaling decisions a real signal.
FAQ
Is Jinshun Casino (gold88.uk) legal?▾
Yes. Jinshun is a legally registered points-based gaming platform under Taiwan's gamerating.org.tw — a distinct legal category from "cash-version" or "credit-version" gambling sites. Our campaign work strictly enforces compliance: all copy and even the AI chatbot prompts are scrubbed of "withdrawal/cash-out" terminology; pre-deploy grep must return 0.
How did you lift CAPI EMQ from 4 to 9?▾
All 6 user_data fields wired: front-end ships fbp (Pixel cookie), fbc (fbclid match), em (email SHA-256), ph (phone SHA-256). Back-end Cloud Function fills client_ip_address + client_user_agent. Meta scores EMQ on field count × quality — we went from 2-3 fields to a full 6, score climbs naturally to 9. This baseline is mirrored across all our iGaming clients via the RedClaw cross-project knowledge base.
What is the 8-tier revival system and how does it work?▾
An automatic top-up bracket for dormant or losing players. 8 face-value tiers (10K → 2M points), capped at 1x per week. A 20% same-day-loss top-up triggers above a loss threshold; +5% if loss ≥ NT$100K. Tier selection queries the ks7 API (queryPromoteReport(playerID)) for the player's P/L before issuing — fully unattended.
Why use LIFF instead of plain LINE OA webhooks?▾
Two reasons: (1) LIFF exposes userId + idToken, so we can lock down a stable "LINE userId ↔ Jinshun playerID" binding; (2) LIFF webviews accept fbp / fbc / fbclid, closing the Meta Ad → LP → LIFF → OA attribution loop. Pure webhooks can only do keyword reply — no attribution, no trusted identity. Our 4 LIFF flows handle bind / me / check-in / deposit.
How do I verify the Jinshun case is real?▾
Four ways: (1) visit gold88.uk to see the brand site; (2) add LINE OA @984dbqie to experience the binding flow; (3) check the registration at gamerating.org.tw; (4) once an NDA is in place, we share Meta Events Manager EMQ screenshots, Firestore schema, and sample CAPI payloads with the technical lead.
Can this architecture be replicated for other iGaming clients?▾
Yes — this is the design goal. The LIFF flows, the 6-field CAPI baseline, the transactional serial pool, and the revival bracket are modularised in our internal library. New-client deltas are: (1) game engine API (ks7 / godeebxp / other ATG variants); (2) legal definition (points / credit / offshore casino); (3) brand voice and mascot. End-to-end deployment for a new client is 1-2 weeks.
Why Firestore instead of traditional SQL?▾
Three reasons: (1) we need transaction-safe serial delivery (preventing two webhooks from issuing the same code), and Firestore transactions are managed — no DB cluster to babysit; (2) Cloud Functions integrate natively, no middleware between LINE webhook and reward decision; (3) multi-client isolation via flat clientId collections and security rules — lighter than SQL schema migrations. LIFF, TG bot, and dashboard all read the same Firestore — zero sync issues.
Want this stack for your iGaming / LINE OA brand?
Same team, same stack, compliance-first, verifiable.