Gambling harm shows up in a bank statement months before it shows up in a doctor’s office, a transcript, or a conversation. Almost nobody is reading the statement.
Illustrative escalation pattern. The doubling, the new merchant and the peer transfer are each a documented signal. No single row is alarming.
Why this exists
It almost ruined my life. I almost got kicked out of school.Simon Capps, founder
Freshman year, every day, unable to stop. Gamblers Anonymous did not do it. Willpower did not do it. What finally worked was his father seeing every transaction and attaching a consequence to the third one.
That is the product thesis, and it is not a hypothesis. It is a case study of one, which is one more than most companies in this category start with. The founder can walk into a treatment center, a campus health office or a National Council on Problem Gambling room and be believed in thirty seconds. That is a distribution advantage no competitor can raise money to buy.
Scale of the problem
2M
US adults meeting criteria for a severe gambling problem, roughly one percent
NCPG
4 to 6M
Additional adults with mild or moderate gambling problems
NCPG
42%
Of US college students gambled within a single year
National campus survey
3.4x
Men are this much likelier than women to gamble at a problem level
Prevalence research
Adults under 35 are roughly 1.5 times likelier than middle aged adults to gamble at a problem level. The population most at risk is the population with the least financial history, the least oversight and, frequently, a card a parent is still paying for.
The shift
31,000
Calls, texts and chats to the national helpline every month
2025 annual report
49.5%
Of all contacts came from people aged 18 to 34
2025 annual report
70%
Of contacts were men
2025 annual report
23 to 31%
Growth in contacts about online and app based gambling, in one year
2024 to 2025
State helplines tell the same story. Pennsylvania logged more than double its 2020 volume and broke its all time record. Arkansas ran 22.5 percent above the prior year. This is not a stable category. It is a curve.
The core finding
73%
of people contacting the national gambling helpline report financial stress from gambling. It is the single most cited reason for reaching out, up from 66 percent a year earlier.
National Problem Gambling Helpline, 2025
15%
of US adults have ever been asked about their gambling habits by a primary care provider. The clinical system is not screening for this.
NCPG and The Harris Poll, February 2026
The new front door
$5B to $24B
Combined monthly volume on Kalshi and Polymarket, September 2025 to April 2026
Pew Research Center
$5B
Traded on Kalshi by adults under 21 in 2026, as reported
CNN Business, August 2026
18 vs 21
Prediction markets are accessible three years before sportsbooks and casinos
Platform terms
This part of the market did not exist eighteen months ago. It is regulated as a financial exchange rather than as gambling, which means it carries none of the responsible gaming infrastructure, sits outside state self exclusion registries, and reads on a bank statement as a brokerage transaction rather than a bet.
Timing
65%
Of adults 21 and older gambled in some form before turning 21
Harris Poll for NCPG
33%
Of adults aged 21 to 44 placed a sports bet before turning 21
Harris Poll for NCPG
11%
Of adults 55 and older did the same, a threefold generational shift
Harris Poll for NCPG
79%
Of Americans think gambling addiction is as serious as drug or alcohol addiction
Harris Poll for NCPG
The last number is the commercial one. Public belief has already moved. Parents do not need to be convinced this is a real problem, which removes the most expensive step in category creation and turns the pitch into a product question rather than an awareness campaign.
Competitive landscape
Blockers
Gamban and bank level merchant code blocks stop access at the point of purchase. They do not report, do not involve another person, and are trivially routed around.
Treatment
Telehealth and counseling networks engage after the crisis, at clinical price points, with a referral required.
Recovery apps
Day counters, journaling and peer support. Self reported data only, so they are blind to what actually happened.
Financial accountability
Bank connected, partner alerted. GamFin Guardian and Whistl are building here now. The category is forming, not empty.
The product
Two tier classification
Confirmed gambling is separated from monitored signals. A sportsbook charge is confirmed. A peer transfer or a cash withdrawal is surfaced without being called a bet. The product never claims more than it knows.
New merchant detection
A first ever charge at a new sportsbook is the escalation tell. It reads as a single small transaction and it is the most predictive event on the statement.
Pattern over total
Frequency, escalation and time of day carry more signal than any single dollar figure. Gambling on 29 of 31 days is the finding, not the monthly sum.
Sample classification output. Blue rows are surfaced, never asserted.
The defensible design
There are two versions of this product and only one of them is a company. In the first, a parent enrolls a kid and watches. In the second, the person at risk connects their own accounts and invites someone to watch with them. Payer and account holder are separate fields.
Parent enrolls
Fails data provider review
Every major aggregator requires the account holder’s express consent for the specific data being shared. A dossier delivered to a third party is the shape that gets rejected.
Fails technically
An unwilling nineteen year old opens a second account, pays cash, or has a friend front the deposit. The dashboard then reports no activity, which is worse than reporting nothing at all.
Fails on age
Under 18 a parent can consent, but minors cannot hold sportsbook accounts. Over 18 the market is real and the member’s consent is legally required. The entire target market sits above that line.
Member invites
Clears review
The account holder is the customer, sharing their own data by choice, with a visible revocation path. This is the standard permissioned pattern.
Works because it is wanted
Nobody routes around a tool they chose. Revocation is itself a signal, and the partner is told when a connection ends.
Still collects from the parent
Covenant Eyes proved the buyer is frequently the spouse or the church, not the person struggling. The same model covers a partner, a sponsor, or self monitoring.
The hard part, stated plainly
Gambling detection is usually pitched as solved. It is not. Merchant category code 7995 catches the obvious sportsbook charge and misses a great deal of what matters.
Processor routing
Many sportsbook deposits clear through a payment processor and code as something other than gambling.
Prediction markets
Kalshi and Polymarket settle as exchange or brokerage activity, not as bets.
Offshore and crypto
Offshore books route through crypto on ramps that read as an ordinary Coinbase purchase.
Peer to peer
Money to a bookie is a Venmo to a friend, and always will be.
A false clean read is the failure that matters
Telling a parent no gambling was detected when it was is worse than shipping nothing. It converts a worried parent into a reassured one, and the product becomes an active harm rather than a passive one.
This is why two tier classification is the product and not a UI detail. Surfacing an unexplained two hundred dollar peer transfer without calling it a bet is the honest answer to a problem with no clean technical solution.
Infrastructure
Plaid declined the use case after quoting five hundred dollars a month across twelve months. The stated reason was the gambling vertical, which is not credible given its own client list. The likely objection is the consent shape, which means a vendor switch without a design change buys another rejection.
Teller
Self serve signup, one hundred free live connections, no compliance gate before building. Narrower coverage and connections that can break. The right place to ship a pilot.
Quiltt
An orchestration layer over MX, Finicity and Akoya. One integration, three providers behind it. Removes the single vendor dependency that caused this problem.
Card issuing, later
Lithic or Stripe Issuing expose the merchant code at authorization, which turns a smoke alarm into a sprinkler. Needs a sponsor bank. Year two.
Vendor analysis
| Provider | Access model | The gate | Verdict |
|---|---|---|---|
| Teller | Direct connections, mTLS, clean REST. 100 free live connections. | Self serve signup. No compliance desk before you build. | Ship the pilot here |
| Quiltt | Orchestration layer over MX, Finicity and Akoya. One integration, three providers. | Sales call, but a small company that will engage on the use case. | The durable answer |
| MX | Best transaction enrichment and categorization in the market, which is the core technical need. | Enterprise motion, quote based, full compliance review. | Second call |
| Finicity | Mastercard owned. Excellent, but tuned for lending and income verification. | Mastercard compliance is not friendlier than Plaid’s was. | Unlikely to help |
| Yodlee | Broadest and oldest US coverage. | Enterprise contracts, dated developer experience. | Fallback only |
| Akoya | Bank owned by Fidelity plus Chase, BofA, Citi, Wells and Capital One. API only, no scraping, very reliable. | Risk tiered onboarding where the reviewers are the banks. This use case lands in the high tier. | Wrong first door |
| Stripe Financial Connections | Built for payment and ACH verification. | Not designed for ongoing transaction monitoring. | Wrong tool |
| Plaid | Broadest coverage, best documentation, the credibility that matters in a pitch. | Declined. Its policy never mentions gambling, so the live clause is consent scope. | Reapply, reframed |
Pricing
$5.99
Per user per month, as proposed. Before data costs, payment processing or support.
Current working price
$20.00
What the nearest competitor was proposing for the same product shape.
Observed at NCPG conference
$500
Per month in fixed data cost across a twelve month term, before a single customer.
Plaid, quoted then declined
At five hundred a month in fixed data cost, the first hundred subscribers carry five dollars each in cost of goods before anything else. At $5.99 that is a gross margin near zero once payment processing is taken out, and it inverts if a customer needs support. The price has to clear the data cost with room, or growth makes the loss larger.
Business model
$250M+
Life360 revenue, January 2024 to July 2025, family location and safety
Statista
$5.3M
Bark revenue over the same period, monitoring for online risk to minors
Statista
Partner led
Covenant Eyes built the accountability partner model and grew through churches and campus groups rather than paid acquisition
Category precedent
The precedent is established on both counts. Families buy subscription software to reduce a specific risk to a young person, and the accountability partner model has a working distribution playbook that does not require an ad budget.
Distribution
Problem gambling councils
NCPG and its state affiliates run helplines taking 31,000 contacts a month, where financial stress is the leading reason people call. They currently have nothing financial to refer those people to.
Campus speaking, already offered
A counselor has offered to bring the founder to speak at Texas Tech this quarter. A dated invitation with a credible third party vouching is what most founders spend a year trying to manufacture.
Treatment and counseling
Financial accountability is the aftercare step gambling treatment programs currently hand off to a spreadsheet.
The founder
A founder in recovery who has already attended the national conference is credible in every room above. No competitor can buy that and it does not appear on a cap table.
What is not yet known
Willingness to pay
Parents have said they would look into buying. Nobody has been asked for a card. A landing page with a real price and a waitlist answers this in two weeks, and answers the pricing question at the same time.
Classification accuracy
What share of real gambling spend does merchant code detection actually catch on live accounts? The number is unknown and it is the product’s ceiling.
Member acceptance rate
Under a member led model, what percentage of invited people actually connect an account and keep it connected past sixty days?
Regulatory posture
Where the line sits between spending alerts and anything readable as a health inference about a named person. Language choices matter here before scale does.
Sources
NCPG and The Harris Poll
National survey, 19 to 23 February 2026, n=2,072 adults. Gambling before 21, generational sports betting split, public seriousness perception, primary care screening rate.
National Problem Gambling Helpline
2025 annual report, released May 2026. Monthly contact volume, age and gender distribution, financial stress share, online and app gambling share.
Pew Research Center
Prediction market trading volume, September 2025 through April 2026.
CNN Business
28 August 2026. Reported under 21 trading volume on Kalshi.
NCPG prevalence estimates
Severe and moderate problem gambling population counts.
State helpline reporting
Pennsylvania and Arkansas year over year call volume, 2025 and 2026.
Statista
Parental control and family safety app revenue, January 2024 to July 2025.
Provider documentation
Plaid developer policy, Teller, Quiltt, Akoya, Lithic and Stripe Issuing product documentation, reviewed September 2026.
Founder interview
2 September 2026. Plaid quote and decision, competitor pricing observed at the NCPG conference, current working price, campus speaking invitation.
Figures are reported as published. Ledger rows throughout are illustrative examples, not real account data.