GH GambleHub

Referral Communication Architecture

1) Why does the ecosystem need referral connections

Referral connections are stable, reproducible bridges between sources of influence (affiliates, content creators, streamers, ambassador users) and conversion actions (registration, KYC, deposit, game). The architecture specifies:
  • predictability of attributions and payouts;
  • onboarding and partner scaling speed;
  • security/privacy (PII-minimization, RG-guardrails);
  • economy (fair split, Cost-to-Serve, ROI channels).

2) Referral linkage models

2. 1 By depth

Single-level (L1): Primary referrer's reward for the target event (FTD/deposit/mission activation).
Multilevel (L2 +): payout shares go up the referral tree; the level/cap/commission limit applies.
Hybrid: L1 fix + a small "thin tail" along L2.

2. 2 By influence carrier

Affiliate/media (networks, publishers, streamers).
User-to-User (referral codes/links of players).
Co-brand/B2B (partner brands/operators, cross-chain crossings).

2. 3 By exchange topology

Hub & Spoke: a single center of attribution and postbacks is an easier conformance.
Mesh (federation): direct connections between chains - requires a strict canon of events and signatures.
Layer-to-Layer: media → offer → payments/CUS → game (clear domains and gateways).


3) Ontology and identifiers

3. 1 Key entities

`referrerId`, `referralId`, `playerId` (псевдоним), `affiliateId`, `campaignId`, `offerId`, `routeId`, `operatorId`, `providerId`, `traceId`.

3. 2 Connectivity and privacy

PII is stored in safe zones, in exchange - only tokens/pseudonyms.
Link click to event: 'traceId' stretches from click/link to registration/deposit/game.
TTL/expiry for refcode tokens; protection against re-binding of devices.


4) Event model and data contracts

4. 1 Minimum set of events

`ref_click`, `deep_link_open`, `session_start`, `registration`, `kyc_status`, `deposit`, `ftd`, `bet/spin`, `mission_progress`, `reward_granted`, `ref_postback_received`, `fraud_signal`.

4. 2 Event requirements

Semantic versions of schemas (Schema Registry), required fields and allowed values.
Idempotency via 'eventId' + deduplication window (± 5 min).
Time zones - UTC/ISO-8601, all amounts and currencies with 'currency'.


5) Protocols: links, diplinks, SmartLink, postbacks

5. 1 Generation of ref-links

Parameters: 'rid' (referrer), 'cid' (campaign), 'oid' (offer), 'sub1.. sub5', 'utm _', 'traceId'.
Parameter signature (JWS/HMAC), TTL, spoofing protection ('kid', 'timestamp').

5. 2 SmartLink/Deep Link

SmartLink redirects to geo/jurisdiction/device/SLO status of payment routes.
Deep Link leads to a specific game/landing/mission (availability check by region/license).

5. 3 Postbecks (S2S)

Signed webhooks: 'signature', 'eventId', 'timestamp', 'rid/cid/oid', 'traceId'.
Retrai with jitter and backoff; register for replay; cursor history upload.


6) Attribution and collisions

6. 1 Attribution rules

Last optional touch with windows by jurisdiction and channel.
Cross-device stitching through agreed tokens/fingerprints (without raw PII).
With multi-stage routes - deposit weights (see splits below).

6. 2 Collision resolution

The priority of the user's direct referral over the media click if the L1 window is active.
Fraud/invalid: marks "disputed," payment hold before reconciliation.


7) Antifraud and quality of referral traffic

Signals: device farm, unnatural CTR/CR, quick click/registration cycles, geo/ASN mismatch, brand-bidding, stimulating traffic under the guise of content.
Actions: auto-cooling, link/offer pause, key revocation, blacklist of sites, re-verification of the referrer.
Scorecard quality: LTV/FTD, retention, chargeback/bonus-abuse, accuracy/timeliness of postbacks.


8) Calculation of payments and splits

8. 1 Basic models

CPA/CPL (fixed per event, e.g. FTD/registration).
RevShare (Net Revenue share by window).
Hybrid (fix + share).
Multilevel split (L1/L2... with mouthguards/rapids).

8. 2 Formulas (simplified)

Proportion of referrer (i) on multistage contribution:
[
share_i=\frac{W_i \cdot Q_i}{\sum_j W_j \cdot Q_j}
]

where (W_i) is the step weight (L1> L2... or according to the channel contribution model), (Q_i) - quality factor (delivery SLI, postback accuracy, absence of fraud/RG flags).

Payment:
[
payout_i = share_i \times (Base\ Reward \ \text{или} \ Net\ Revenue \times rate) \times Adjustments
]

'Adjustments': credits/penalties for SLOs, deductions for controversial cases, RG/sanction adjustments.

8. 3 Restrictions and mouth guards

Quarterly/monthly limits; minimum payout; hold on the review of controversial events; slices by jurisdiction and payment status.


9) Security, privacy and RG

Zero Trust: mTLS for S2S, short-lived tokens, egress-allow-list.
PII minimization: in ref traffic - only tokens/pseudonyms; detoxification strictly in safe zones.
RG-guardrails: intensity constraints on vulnerable segments; mandatory disclaimers; prohibition of offers in prohibited regions.
DPA/DPIA and localization: storage and processing by jurisdiction, prohibition of cross-border PD export outside the rules.


10) Dashboards and observability

10. 1 Referral/Affiliate Panel

Clicks, registrations, KYC, FTD, deposits, CR/ARPU/LTV, postback statuses, splits and expected payouts, caps/quotas.

10. 2 Ecosystem panel

Linkage map (referrer→referal), cohorts, proportion disputed, RG/sanction fouls, MTTR by incident, channel/circuit contribution.

10. 3 SLO

Postback delivery ≥ 99. 9% (p95 ≤ 1-2 s), bus lag ≤ 200-500 ms, panel render p95 ≤ 1. 5–2. 0 s, trace coverage ≥ 95% of paths.


11) Governance и RACI

Solution/ArtifactRACI
Referral Model PolicyProduct LeadEcosystem OwnerLegal/RG, FinancePartners
Event Diagrams/PostbacksData StewardProtocol CouncilSRE, SecurityIntegrants
Change splits/ratesFinance LeadEcosystem OwnerProduct, RiskPartners
Quality stop buttonRisk/Fraud LeadEcosystem OwnerAffiliate LeadPartners
Analysis of controversial casesFinance OpsEcosystem OwnerLegal, RiskParties to the dispute

12) Operational processes

12. 1 Onboarding referrer

KYP/KYB, traffic sources, jurisdictions, brand guide, RG agreement.
Issuance of keys/API/webhooks; sandbox and conformance tests of signatures/idempotency.

12. 2 Changes/Releases

Change-calendar (green/yellow/red windows), progressive inclusions (1%→5%→25%→100%), guardrails and auto-rollback.

12. 3 Incidents/war-room

P1/P2, stop buttons (traffic/offer/route), SLA for a 60-90 s trace packet, RCA "no blame."

12. 4 Reconciliation and payments

Cursor uploads, reconciliation of event hashes, acts/invoices, NET7/14/30, holds (fraud/chargeback/bonus abuse), FX courses.


13) Integration with the rest of the ecosystem

Payments/AWS: quotas and cut-over for CR/P95 degradation.
KYC/AML: pass-rate and SLA stages; fast-track for low-risk.
Content/Missions: Diplinks to Games/Tournaments with RTP/Volatility and Regional Constraint Checks.
Analytics/ML: propensity models, ranking of offers, anti-fraud signals.


14) Anti-patterns

"Postback Zoo": different formats/signatures/windows → doubles, losses, disputes.
Offset pagination of event history under load → holes/duplicates (cursors only).
Retrai without jitter/limits → storm and double accruals.
PII in ref events → leaks and violations.
Unlimited multilevel tails → toxic economy, inability to control.
Single SPOF redirect without N + 1/health-flip.
There is no traceId → attribution cannot be proved.


15) Checklists

15. 1 Engineering

  • Ontology and Schema Registry (versions, required fields).
  • Postback and Signature Contracts (JWS/HMAC), Deduplication window.
  • SmartLink/Deep Link generator, TTL and parameter protection.
  • Attribution: rule, windows, collision priorities, cross-device stitching.
  • Pay and split models, caps/limits, deductions.
  • Zero Trust, tokenization, DPIA/DPA, RG-guardrails.
  • dashboards/scorecards and SLO, alerts and war-room.

15. 2 Launch

  • Sandbox/conformance, test events and reconciliation.
  • Progressive release, auto-rollback.
  • Reconciliation and Invoice Pipeline are configured.

15. 3 Operation

  • Weekly scorecards of quality and SLO.
  • Review of controversial/hold cases.
  • Periodic RCAs and rule updates.

16) Maturity Roadmap

v1 (Foundation): a single canon of events/postbacks, SmartLink, L1 model, basic dashboards.
v2 (Integration): cross-chain refs, cursor cues, hybrid payouts, scorecards, and auto-cool.
v3 (Automation): ML-ranking of offers, predictive cut-over payments/CCP, auto-guardrails by SLI/RG.
v4 (Networked): federated attribution between networks, collective missions/tournaments, split DAO rules.


17) Success metrics

Business: uplift FTD/ARPU/LTV from ref channels, CAC/payback, share of quality sources.
Quality: accuracy/timeliness of postbacks, controversy <X%, holds/chargebacks.
Technique: p95 redirects and postbacks, lag tires, MTTR by incident.
Compliance/RG: PD incidents = 0, jurisdictional compliance, RG triggers/1k active.
Economy: Cost-to-Serve (per rps/txn/event), credits/penalties, ROI channels.


Brief summary

Referral communication architecture is ID and event standards, SmartLink/diplinks and signed postbacks, honest attribution, strict privacy and RG, plus transparent split economics and observability. Build a canon and SLO, automate guardrails and reconciliation - and the referral circuit will become a scalable, safe and profitable foundation for the entire ecosystem.

Contact

Get in Touch

Reach out with any questions or support needs.We are always ready to help!

Start Integration

Email is required. Telegram or WhatsApp — optional.

Your Name optional
Email optional
Subject optional
Message optional
Telegram optional
@
If you include Telegram — we will reply there as well, in addition to Email.
WhatsApp optional
Format: +country code and number (e.g., +380XXXXXXXXX).

By clicking this button, you agree to data processing.