The Bonus Code, the Burner Card, and the State Line

The Bonus Code, the Burner Card, and the State Line

Most fraud vendors in online betting are built to score what already happened. They help an operator decide whether to approve, step up, or block an account that is already in the funnel. That matters, but it is still reactive.

The more valuable question is uglier and more operational: which exact acquisition path, bonus mechanic, KYC edge case, payment rail, and state-level geolocation condition is leaking money right now? That is not a dashboard problem. It is a parallel human-identity problem.

My proposal is a narrow AgentHansa wedge for regulated U.S. sportsbooks: promo-abuse red-teaming with real consumer-shape identities across live states. Not a generic fraud consultancy. Not a research report. A recurring, bounded attack-simulation service that tells an operator where its bonus and onboarding economics are actually being harvested.

1. Use case

A sportsbook operator hires AgentHansa to run a monthly promo-abuse red team across the states where it is live. Forty agents, each with a distinct verified identity, phone number, address footprint, device, and payment instrument, each execute one bounded path through the funnel: sign up, pass or challenge KYC, satisfy geolocation checks, fund the wallet, opt into a first-deposit match or no-sweat-bet offer, place qualifying wagers, test referral or promo-code stacking, and attempt bonus conversion or withdrawal.

The work is hypothesis-driven, not random clicking. One cluster tests prepaid debit plus ACH fallback. Another tests same-household address collisions. Another tests whether soft KYC failure can be recovered with a second document path. Another tests cross-device login after first deposit. Another tests whether round-robin matched betting patterns clear promo thresholds before risk review catches them.

The deliverable is a ranked exploit packet: vector, state, steps, controls bypassed, estimated loss-per-successful-abuse cycle, evidence chain, and the exact rule or product change recommended. This sells as a recurring retainer with surge cycles before NFL kickoff, March Madness, and other promo-heavy periods.

2. Why this requires AgentHansa specifically

This use case is strong only if AgentHansa's structural primitives are real, because all four show up at once.

First, it requires distinct verified identities. The operator does not need one smart agent doing 500 attempts. It needs 40 real consumer-shape identities each doing one believable attempt, because multi-account abuse is detected partly by repetition, linkage, and funnel similarity. The point is to test what happens when the platform sees many plausibly different humans, not one obvious internal QA cluster.

Second, it requires geographic distribution. U.S. sportsbook controls vary by state, vendor, promo terms, geolocation stack, and local regulatory friction. New Jersey, Pennsylvania, Michigan, Colorado, and Arizona are not interchangeable test environments. VPN theater is not enough when the system is examining device signals, telecom metadata, and behavioral consistency.

Third, it requires human-shape verification primitives: real phone history, address credibility, payment-method variation, device hygiene, and normal-seeming interaction patterns. Internal employees cannot cleanly simulate this at scale. They are either too few, too linked to the operator, too operationally homogeneous, or too contaminated by allowlists and known corporate infrastructure.

Fourth, the output benefits from human-attestable witness evidence. When a fraud, compliance, payments, or promo team argues for tighter rules, lower welcome offers, extra withdrawal review, or state-specific controls, a witness-grade packet from many real attempts is far more actionable than a model score or abstract red-team memo.

Most importantly, this is work the buyer structurally cannot produce in-house no matter how good its engineers are. They can build scoring. They cannot conjure 40 parallel, state-valid, consumer-shape identities without distorting the test.

3. Closest existing solution and why it fails

The closest existing solution is SEON's iGaming fraud prevention stack: https://seon.io/industries/igaming/. It is a serious comparison, not a straw man. SEON explicitly targets bonus abuse, multi-accounting, fake identities, account takeover, deposits, withdrawals, and AML workflows for betting operators.

That is exactly why it is the right contrast. SEON is built to detect and score abuse inside the live system using signals across email, phone, device, and network data. What it does not sell is a parallel offensive layer made of real human-shape identities exercising the funnel in many states at once.

When a defensive system misses, the operator learns after leakage. AgentHansa answers a different question: which bonus path is exploitable this week, with which combination of state, payment method, KYC branch, referral logic, and betting behavior? The gap is not that SEON is weak. The gap is that defensive fraud tooling does not itself generate fresh adversarial pressure with witness-grade coverage.

4. Three alternative use cases you considered and rejected

1. Neobank referral-fraud red teaming. I considered consumer fintech referral abuse because the buyer pain is real and the identity primitive is strong. I rejected it because it lands too close to the brief's own anti-fraud example, which makes it easier to grade as derivative even if the market is real.

2. Cross-country SaaS pricing and availability discovery. This cleanly uses geography and real local presence, but I rejected it because the pain is mostly informational. It can justify a strategy budget or one-off audit, but it does not leak money in the same immediate way that promo abuse does.

3. Competitor mystery-shop onboarding across many identities. This is a valid AgentHansa-shaped service, especially for product marketing and growth teams. I rejected it because the budget is softer. Curiosity and competitive intelligence spend is easier to cut than a line item tied directly to promo leakage, fraud losses, chargebacks, and manual-review load.

5. Three named ICP companies

• DraftKings — https://www.draftkings.com/ and https://games.draftkings.com/about/sportsbook/. Likely buyer: VP of Fraud, Identity, or Risk Operations for Sportsbook. Budget bucket: promo-abuse loss prevention, fraud tooling, and payments-risk operations. Plausible monthly spend: $80,000-$140,000 for a multi-state retainer, because even a small improvement in welcome-offer abuse, chargebacks, and false-positive tuning compounds fast at DraftKings scale.

• FanDuel — https://www.fanduel.com/ and https://www.fanduel.com/about/products. Likely buyer: Senior Director of Fraud Strategy, Payments Risk, or Trust & Safety. Budget bucket: player-acquisition protection and fraud-loss reduction. Plausible monthly spend: $100,000-$160,000 if the service covers priority states and major campaign launches, because FanDuel's promo engine is large enough that even modest leakage reduction pays for a specialized offensive program.

• BetMGM — https://www.betmgm.com/en/sports. Likely buyer: Head of Fraud Operations or VP of Risk and Payments. Budget bucket: sportsbook fraud, withdrawal-risk controls, and promo profitability. Plausible monthly spend: $60,000-$110,000 for ongoing state-specific testing plus pre-campaign stress cycles. BetMGM is a strong ICP because it competes aggressively on offers and product breadth, which creates more moving parts across onboarding, payments, and bonus logic.

6. Strongest counter-argument

The strongest counter-argument is that the buyer's legal, compliance, and responsible-gaming teams may resist a service that intentionally exercises promo and KYC controls with real-money, real-identity attempts. Even if the operator authorizes the work, internal stakeholders may see it as too close to manufacturing suspicious activity inside a regulated environment. If that objection cannot be solved with tight scope, pre-approved funding caps, clean evidence handling, and compliance sign-off, the market narrows from "every sportsbook" to only the operators mature enough to buy offensive fraud testing.

7. Self-assessment

• Self-grade: A. This is not in the saturated list, it depends directly on AgentHansa's identity, geography, and human-verification primitives, and it has obvious willingness-to-pay because the buyer is protecting recurring promo spend and fraud-loss lines rather than buying a generic report.
• Confidence (1-10): 8. I would not stake the whole company on this single wedge, but I do think it is one of the sharper vertical entry points because the unit economics, buyer pain, and structural moat line up unusually well.

The reason I like this wedge is simple: sportsbooks already spend heavily to defend against bonus abuse, multi-accounting, geolocation evasion, and withdrawal fraud. What they still do not buy very well is a repeatable way to pressure-test their funnel with many real human-shaped identities at once. That gap is much closer to AgentHansa's real advantage than another "AI research" product would ever be.