Data economics in iGaming

1) Why iGaming'y "data economy"

1. Where is the value? (deposit/rate growth, retention, fraud/chargeback reduction, CAC↓)

2. How much does it cost? (collection, storage, calculations, licenses, labor, compliance)

3. How to prove the effect? (uplift/increment, causal A/B, guardrails)

2) Units of value and basic formulas

GGR = 'bets - wins' (by segment/game/channel).
ARPPU/ARPU - average revenue per paying/user.
LTV = 'Σ (marginal cash flow _ t/( 1 + r) ^ t)' including deductions and bonuses.
CAC - cost of attraction (including affiliates and media billing).
Net Gaming Revenue (NGR) - GGR minus provider bonuses/taxes/fees.
Uplift (Δ) - metric increment from action/model vs control.

The goal of analytics is to maximize 'NGR - (Cost_data + Cost_marketing + Cost_risk)' under compliance and responsible gambling restrictions.

3) Data → solutions → money chain

1. Collection: events (sessions, rates, deposit/withdrawal), payments, KYC/AML, support, content, technical metrics.
2. Preparation: contracts, DQ, features, showcases (batch/stream).
3. Models/Rules: recommendations, risk limits, anti-fraud, NBA/pricing, lobby personalization.
4. Delivery: CRM/CDP, push/email/chat bots, on-site widgets, limits/cool-offers.
5. Measurement: A/B/bandits, causality, GGR and retention increment, cost-to-serve.

4) Cost Map (TCO) and FinOps for Data

• Collection: SDK/streaming, brokers, CDC.
• Storage: lake/OLAP, backups, versions, cold layers.
• Processing: ETL/ELT, streaming, feature platform, ML/LLM calculations.
• Licenses and tools: catalogs, DQ, observability.
• Command: DS/DE/DA, SRE data, annotation.
• Compliance/security: KYC/AML, RG (Responsible Gaming), encryption, audit, legal advice.
• Egress/partners: data exchange, reports, integrations.

• Chargeback/Showback costs to teams/products.
• Budget guardrails on clusters and showcases (p95, bytes scanned, GPU-hours).
• Quotas/limits (scan caps, concurrency, off-peak backfill).
• Cost-aware planning: hot real-time for Gold cases only.

5) Data Investment Prioritization Matrix

Evaluate initiatives along two axes: Increment to NGR/risk savings × Payback period/SarEx.

• Anti-fraud/chargeback rates, deposit/responsible play limits.
• Lobby/banner personalization, NBA for re-deposit.
• Real-time SLO alerts for payments/gaming sessions.
• Silver: dynamic promotional targeting, bonus pricing, look-alike.
• Bronze: long-term R&D models, low-frequency back-office reports.

6) Economy real-time vs batch

Real-time = latency-premium: we pay more for compute/engineering, we pay back if the deadline for solutions ≤ 1-60 seconds and the Δ to GGR/risk loss is significant.
Near-real-time (1-5 min): cheap compromise for marketing/operations.
Batch (hour/day): training, reporting, long tail analytics.
Rule: protect every real-time showcase with a business case and SLA→SLO→$ effect.

7) Data monetization

B2C (indirect): personalization of content/promotions → LTV↑, ottok↓, pretenzii↓.

• Reports/analytics to partners (game providers, affiliates) with depersonalization and aggregates.
• Recommendation/anti-fraud API for white-label/partner operators (with hard SLAs and compliance).
• Data coop within the holding: exchange of storefronts, a common feature platform.
• Important: license compliance, anonymization/diff. privacy, prohibition of re-identification.

8) Marketing and attribution economics

Incremental attribution: geo-experiments, PSA, MTA + RTA with causal adjustments.
Uplift models: we show the campaign only to those who are expected to Δ> 0.
Creative × context: mixed effects (hour/channel/segment) - target sparingly.
Guardrails: complaints, RG triggers, frequency limits and cooling windows.

9) Risk & Compliance: Impact on P&L

KYC/AML/sanction screening: automation reduces manual labor/fines.
Responsible Gaming: limits and scoring of harmful patterns → retention "healthy," legal riski↓.
Audit/logging/DSAR: there is a cost, but this is insurance against incidents and blocking.
Data localization and RLS/CLS: Infrastructure costs are offset by market access.

10) Data economy metrics

Cost-to-Serve (CTS) on 1k events/requests/scoring.
Cost-per-Insight (CPI) and Cost-per-Decision (CPD) are the full path to action.
Δ NGR/ Δ LTV per feature/model/campaign.
Payback Period and ROI analytical initiatives.
Coverage/Adoption (what proportion of traffic/agents uses the model/showcase).
Quality Guardrails: p95 latency, freshness, DQ violations/c 1k events.

11) Bonus pricing and anti-arbitration

Individual bonus limits: risk function and CLV; we fine exploit behavior.

Fair promo pricing: optimization by uplift to NGR, and not by "response in general."

Antibot/anti-multi-account: graph features, device fingerprint, behavioral vectors.

12) Architectural solutions affecting the economy

Column formats + ZSTD/clustering: fewer scans → cheaper reports.
Feature Store (online/offline single spec): less duplication, fewer errors.
Thread prioritization and admission-control: Gold showcases do not suffer from research battles.
Caching and materializations: pre-aggregates for hot dashboards.
Spot/Preemptible resources for Bronze-rebuild.
Edge-enrichment: cheap local solutions, less egress.

13) Proof of effect (causal)

A/B with increment to NGR/deposits, stratified by country/channel/device.
Bands for real-time NBA/prices - risk limit (guardrail KPI).
Diff-in-Diff/SCM for regulatory/external shocks.
Post-hoc audit: performance regression, replace the "last click" with causal uplift.

14) Roles and Ownership Model

Product Data Owner: P&L responsibility for display cases/models.
FinOps for Data: quotas, budget alerts, TCO and CTS reports.
Risk & Compliance: RG/KYC/AML, audit, privacy policy.
Analyst/DS/DE: hypotheses, models, experiments, window supply.
Partner Lead: B2B analytics packages, SLAs and licensing.

15) Antipatterns

Zero causality. Reports instead of increment → marketing "eats budget."

"All in real-time." No deadline - no speed bonus.
No FinOps. Expensive scans and ownerless showcases.
Bonuses "for all." Arbitration and budget burning.
Lack of RG/compliance in P & L. Risks and fines "eat up" the effect of analytics.
Opaque models. It is difficult to defend on audits/disputes with payments/regulator.

16) Implementation Roadmap

1. Inventory & Baseline: Showcase/Model/Value Register (CTS/CPI), Gold/Silver/Bronze Card.
2. Objectives and effects: 3-5 cases with NGR/ Δ LTV Δ forecast and payback period.
3. FinOps: quotas, limits, chargeback, value panels; off-peak/spot rules.
4. Causal dimension: experimental framework, uplift models, guardrails.
5. Compliance in the circuit: RG/KYC/AML, privacy/DSAR, RLS/CLS - as code.
6. Monetization/partners: impersonal reports, APIs with SLAs, licenses.
7. Scale: multi-region, edge, knowledge graphs, thread prioritization automation.

17) Pre-data initiative checklist

• The business case is described: effect metric (Δ NGR/ Δ LTV) and deadline solution.
• Calculated CTS/CPI/CPD and budget, there are limits and off-peak policies.
• Compliance/privacy agreed (RG/KYC/AML, RLS/CLS, DSAR).
• Set up experiments/bands, recorded guardrail KPI.
• Owners, SLA/SLO, delivery and feedback channels are defined.
• Monetization/reporting plan to partners (if applicable), license terms.
• Observability panels: p95 latency, freshness, bytes scanned, cost per insight.

18) Mini-templates (pseudo-YAML/SQL)

18. 1 Showcase Value Profile

yaml datamart_cost_profile:
name: rt_player_lobby slo: {latency_p95_ms: 200, freshness_s: 5}
traffic_qps: 1200 cost_guardrails:
max_cts_usd_per_1k: 0.45 gpu_hours_day: 4 priority: gold backfill: offpeak

18. 2 Initiative Effect Card

yaml data_initiative:
name: nba_deposit_retry target_metric: NGR expected_uplift: +2.1% (p90)
payback_days: 28 experiment: ab_test_stratified(country, device_os)
guardrails: [complaints_rate<=0.02, rg_flags_no_increase]

18. 3 Bonus pricing policy

yaml bonus_pricing:
model: uplift_ltv_v3 min_expected_uplift_pp: 0.3 max_bonus_cost_pct_ggr: 12 cooldown_days: 7 anti_arbitrage: on

18. 4 FinOps for queries

yaml query_policy:
max_scan_mb: 2048 deny_patterns: ["SELECT "]
cost_alert:
threshold_usd: 50 notify: "data-finops@"

18. 5 Incremental valuation

sql
-- uplift по сегменту select segment,
avg(treatment_outcome - control_outcome) as uplift from causal_results group by segment order by uplift desc;

19) The bottom line

Data economics in iGaming is the discipline of how every event and every model affects money, risk, and compliance. Hard SLOs and FinOps-guardrails, causal effect measurement, real-time prioritization only where there is a deadline premium, and integration of RG/KYC/AML into P&L all turn the data platform from a cost center to an engine of NGR, LTV and business sustainability.