GuardRail

The money tells first.

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.

Date
Merchant
Classification
Amount
Aug 04
FanDuel
Confirmed gambling
$25.00
Aug 09
FanDuel
Confirmed gambling
$50.00
Aug 16
PrizePicks
New merchant
$100.00
Aug 21
Venmo, S. Brooks
Peer transfer
$200.00
Aug 27
Caesars Sportsbook
New merchant
$400.00

Illustrative escalation pattern. The doubling, the new merchant and the peer transfer are each a documented signal. No single row is alarming.

Market research briefSeptember 2026Prepared for GuardRail

Why this exists

Built by someone it happened to

01 / OriginPrimary account
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.

The insight is not surveillance. Visibility alone changed nothing. What changed the behavior was a named person seeing the data, and a consequence attached to it, at a moment when he already wanted to stop.

Scale of the problem

Two million severe, millions more at risk

02 / PrevalenceNCPG, peer reviewed surveys

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

Help seeking is getting younger and moving online

03 / Demand signalNational Problem Gambling Helpline, 2025 report

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

Money is the presenting symptom

04 / ThesisThe slide the company is built on

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

Put those two numbers next to each other and the category defines itself. The harm arrives as a financial event, and the place people are supposed to be caught is not looking. The bank statement is the only surface where this is visible early, and it is currently unmonitored.

The new front door

Prediction markets opened the door at eighteen

05 / TailwindPew Research, CNN Business

$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.

Worth naming plainly. Plaid told GuardRail its largest clients in this space are Kalshi and Polymarket. The data rails this product needs are partly owned by companies whose revenue is the behavior it exists to interrupt. That is a structural headwind to plan around rather than discover later.

Timing

It almost always starts before twenty one

06 / WindowNCPG and The Harris Poll, n=2,072, Feb 2026

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

Four categories exist. None own the money.

07 / PositioningVerified September 2026

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 honest read. GuardRail is not first and should not claim to be. GamFin Guardian is attached to an established gambling financial counseling nonprofit, which is real clinical credibility. The open ground is the consent architecture and the classification quality, not the idea.

The product

A smoke alarm on the account

08 / What it doesWorking prototype, September 2026

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.

Date
Merchant
Classification
Amount
Aug 25
Caesars
Confirmed
$50.00
Aug 25
PayPal, S. Brooks
Peer transfer
$200.00
Aug 25
ATM withdrawal
Cash
$80.00
Aug 24
ESPN BET
Confirmed
$25.00

Sample classification output. Blue rows are surfaced, never asserted.

The defensible design

The parent pays. The member consents.

09 / Consent architectureThe decision upstream of everything else

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.

One product, one consent model. The preventative message aimed at parents is marketing, not a second mode. Building both modes is how this becomes two half products that each fail their own compliance review.

The hard part, stated plainly

Merchant codes are a leaky net

10 / Technical riskNamed, because reviewers find it anyway

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

Bank data is the dependency to solve now

11 / Build pathVendor review, September 2026

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.

If Plaid has declined the whole category rather than one applicant, the rails are the competitive moat, not the obstacle. Whoever solves data access first owns a category everyone else is locked out of. Reapply with the member led consent model and a written data use and retention policy in hand.

Vendor analysis

Every Plaid alternative, and the honest verdict

12 / Data railsReviewed 2 September 2026
ProviderAccess modelThe gateVerdict
TellerDirect connections, mTLS, clean REST. 100 free live connections.Self serve signup. No compliance desk before you build.Ship the pilot here
QuilttOrchestration 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
MXBest transaction enrichment and categorization in the market, which is the core technical need.Enterprise motion, quote based, full compliance review.Second call
FinicityMastercard owned. Excellent, but tuned for lending and income verification.Mastercard compliance is not friendlier than Plaid’s was.Unlikely to help
YodleeBroadest and oldest US coverage.Enterprise contracts, dated developer experience.Fallback only
AkoyaBank 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 ConnectionsBuilt for payment and ACH verification.Not designed for ongoing transaction monitoring.Wrong tool
PlaidBroadest 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
The trap is treating this as a shopping problem. MX, Finicity, Akoya and Yodlee all carry substantially the same end user consent language Plaid does, so switching vendors without changing the consent model buys three more rejections. Fix the model first, then the vendor list becomes easy.

Pricing

Six dollars is the wrong number

13 / Unit economicsWorking assumptions, to be tested

$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.

The deeper problem is positioning. This is a peace of mind purchase for a parent frightened about their child, sitting next to family safety subscriptions people already pay for. Pricing it below a streaming service tells the buyer it is a minor utility. Test fifteen to twenty five dollars before assuming six.

Business model

Families already pay monthly for this shape

14 / ComparablesPublic revenue data, Statista

$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.

One retention warning. A preventative product sold to a parent with no existing concern produces an empty dashboard when it works, which reads as a reason to cancel. The parent whose child has already been caught converts immediately and stays. Lead with the second.

Distribution

Channels that do not require an ad budget

15 / Go to marketRanked by cost of entry

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.

Every channel here reaches the person at risk or their clinician, not a cold parent, which matches the member led consent model exactly. Paid social is the most expensive and least differentiated of the options and should come last, not first.

What is not yet known

Open questions worth answering next

16 / DiligenceHonest gaps

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.

Nothing above is a reason not to build. Each is a two to six week experiment, and a founder who has already named them is ahead of most people pitching this category.

Sources

Every figure in this deck

17 / ReferencesCompiled 2 September 2026

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.