CTV for Fintech Apps: The Install Is Not the Conversion

A fintech install tells you almost nothing on its own. What you are buying is a funded account, a completed identity verification, or a first transaction, and that event sits weeks downstream of the ad behind a funnel designed by your compliance team.

That gap is why fintech is the hardest vertical to measure on CTV. The economics tolerate a high cost per acquisition. The measurement does not tolerate a short attribution window.

Here is what to settle before the budget moves.

Your funnel is a compliance funnel

Most verticals lose users to boredom. Fintech loses them to paperwork.

Between install and funded account sits some combination of identity verification, document upload, bank linking, and a risk review that can decline the user outright. Each step has real drop-off, and some of it is intentional. You are not supposed to onboard everyone.

Two consequences for media buying follow, and both get missed.

First, a declined applicant is not the same thing as a disengaged one. A cohort with a low verified-account rate may be one your risk model rejected. That is a different finding from a cohort that lost interest, and it points at a different fix. If your reporting cannot separate abandonment from decline, you will optimize away from geographies or sources whose users your own underwriting rejected, and conclude the media was at fault.

The counterweight matters as much as the point. An individual decline is not a quality signal, but a source with a persistently high decline rate usually is. Device farms, recycled identities and incentivized users whose documents do not match all show up as declines. Treat a single declined user as information about your underwriting and a source-level decline rate as information about the source.

Second, verification friction is a product variable sitting inside your media test. Ship a change to the onboarding flow mid-flight and you have confounded the campaign. Fintech teams do this constantly, because onboarding is also where the product team is always working. Log every onboarding and risk-policy change against the campaign calendar before you launch.

What you can actually measure, and when

CTV creates a measurement problem that a 30-day funnel makes worse.

A television and a phone share no advertising identifier. The connection is made by matching the household IP address inside a lookback window, which in most configurations is short: typically around a day, and rarely more than three. The exact value is a setting your measurement partner exposes, not a property of the channel, so find out what yours is before you read anything. The match is to a home, not a person, and probabilistic impression matches generally rank below deterministic click and device-ID matches in an attribution waterfall. When the match loses, the install does not disappear. It gets credited elsewhere, usually to organic. Our post on why CTV installs arrive late and land in organic walks through the decay.

Now layer a fintech funnel on top. The install might be credited inside 24 hours. The funded account arrives three weeks later. Those are different measurement problems with different failure modes.

On iOS it gets harder. Neither SKAdNetwork nor AdAttributionKit provides a direct path for a television impression to produce an attribution. Even for in-app sources, the three conversion windows close on days 2, 7, and 35 measured from the user's first app launch, and not from the impression, and each postback arrives after a further random delay of 24 to 48 hours for the first and 24 to 144 hours for the later two. A verification completed on day 35 can therefore be reported around day 41. Only the first postback can carry a fine-grained conversion value, and only for apps in the higher data tiers; lower tiers receive a coarse value or, at Tier 0, no second or third postback at all. The same arithmetic applies to any long funnel, which we worked through for subscription apps.

The practical read: use your own funnel data, keyed to your own user ID, as the source of truth for everything past install. Your verification and funding events live in your systems, not in any ad platform's. What you lose on iOS is the per-campaign attribution of those events, not the events themselves. That is why the measurement plan has to be built around a geo test.

Price the event, not the install

Fintech is one of the few verticals where the deep event is valuable enough, and discrete enough, to consider buying directly.

A funded account or a completed verification fires a server-side event you control, which makes it a candidate for outcome pricing on rewarded and offerwall inventory, where you pay per completed action and the supply side carries the risk of nobody finishing. We covered the mechanics in how to buy offerwall traffic that retains.

Three cautions specific to this vertical, and the first is the one that kills most attempts.

Confirmation latency. Offerwall economics assume the reward pays at or near the moment of completion. A KYC decision can take minutes or days once it routes to manual review, and a funded account depends on bank settlement that runs one to three business days. An event you cannot confirm while the user is still engaged is awkward to carry on a cost-per-action deal. Sometimes the answer is to price an earlier verifiable step, such as a completed application or a successful bank link, and treat funding as the metric you judge the cohort on, not the one you buy.

Set the qualifying event past the point where your product starts earning. If your funnel dies at identity verification, reward completed verification and price it like the expensive step it is. Rewarding the install buys people who wanted the reward, which in fintech means people who will never clear underwriting.

The events worth buying are the events worth faking, and here the fraud is not an invoice dispute. A funded account opened with a synthetic or stolen identity is a BSA and AML problem before it is a media problem, and no clawback window fixes it. Agree on the fraud controls and the clawback terms before launch, and make sure your financial crime team knows the campaign exists.

There is also a regulatory question worth raising with counsel early: paying a third party per funded account for a regulated product can touch referral-compensation rules, and for securities products it can touch finder or broker-dealer registration. That is a conversation to have before the insertion order, not after.

What Apple and Google require before you write creative

This is the part no media plan accounts for, and it constrains the product claims your ad can make. Both stores regulate this vertical, and they do not regulate it identically.

Google Play. The financial services policy restricts apps related to the management or investment of money and cryptocurrencies. Any app with financial features must complete a Financial features declaration in Play Console, and developers must comply with the regulations of every region the app targets, with documentation Google can connect to the developer account. Personal loan apps carry heavier requirements: the store listing must disclose all loan terms including maximum APR, the minimum and maximum repayment period, and a representative example of total cost including principal and all applicable fees. Loans must not require full repayment in 60 days or less, and apps may not offer loans with an APR of 36 percent or higher in the United States. Loan apps must sit in the Finance category and are barred from requesting sensitive permissions including contacts, photos, and precise location.

Apple, and here is a divergence worth catching. The App Store Review Guidelines at 3.2.2(ix) require personal loan apps to "clearly and conspicuously disclose all loan terms, including but not limited to equivalent maximum Annual Percentage Rate (APR) and payment due date," and state that loan apps "may not charge a maximum APR higher than 36%, including costs and fees, and may not require repayment in full in 60 days or less." Note what is missing: Apple's 36 percent cap carries no geographic qualifier, while Google's is scoped to the United States. A lender operating lawfully at 42 percent APR outside the US can clear Google Play and fail App Review.

Apple also bars binary options apps outright under 3.2.2(viii) and requires apps facilitating contracts for difference or other derivatives such as FOREX to be "properly licensed in all jurisdictions where the service is available." Under 5.1.1(ix), apps providing services in highly regulated fields, with banking, financial services and crypto exchanges named explicitly, "should be submitted by a legal entity that provides the services, and not by an individual developer." Apple sets further requirements for cryptocurrency wallets, mining and exchanges under 3.1.5, which is the section to read with counsel if that describes your product.

The gate you hit first is the ad platform, not the app store

This is the omission that costs the most time, because it arrives before any of the above.

Publishers and ad platforms restrict these categories on their own terms, often more tightly than the stores do. Roku's ad policies are a useful example because they are published in full. Crypto requires prior permission: "You must get permission from Roku prior to running any ads for cryptocurrency products or services." A narrow set of products may be approved case by case and in the United States only, covering mining hardware, cold wallets, certain tax and educational services, and certain licensed platforms and ATMs. Everything else is prohibited, including initial coin offerings, initial dex offerings, single tokens and coins, crypto loans and crypto-mining software.

Beyond crypto, Roku prohibits ads for single securities, stocks including penny stocks, and bonds. It may require advertisers promoting financial products to demonstrate that they are properly licensed or registered with the relevant regulators. And there is a direct creative constraint that catches teams by surprise: ads for credit cards, loans and insurance "must not directly request the input of any personally identifiable information or financial information."

So the sequence is: find out which platforms will take your category at all, then which markets you are licensed for, then what your creative may claim. Doing it in the other order produces an approved spot nobody will run.

Three things follow for the brief:

Your ad and your store listing have to tell the same story. The listing is where disclosed terms live. If a CTV spot implies a rate or a speed the listing contradicts, the problem surfaces on a page whose review you do not control.

Licensing shapes where you can run, market by market. For a US crypto or payments business the relevant unit is often the state rather than the country, and money transmitter licensing is a state-by-state regime separate from federal registration. The list of places you may lawfully advertise is usually much shorter than the media plan wants.

Get approved claims language in writing before the edit. If you are advertising a brokerage or investing product, note that a CTV spot is a retail communication under FINRA Rule 2210, which requires principal approval before use and, for newer member firms, filing ahead of first use. Reshooting a spot because legal rejected an on-screen rate is an expensive way to learn this.

Measure it with a geo test

Everything above points the same direction: the per-campaign attributed number will be late, partial, and credited to the wrong row.

A geo holdout avoids it. Suppress the channel in a randomly assigned set of markets, run it in the others, and compare total funded accounts between the groups using your own data. It never asks which touchpoint gets credit, so the household match and the organic spillover stop mattering.

One constraint lands hard on this vertical. You cannot randomize across markets you are not licensed in, so your licensing footprint sets your randomization pool before any statistical consideration does. For a business live in eight US states, that pool may be too small to produce a readable result, and finding that out at the design stage is much cheaper than finding it out at the read.

Size the test on your own conversion volume and run it long enough to cover the funnel plus the lag, which for a three-week funnel is considerably longer than most fintech teams plan for. Set the read date before you launch and agree not to look early.

When CTV is the wrong channel for fintech

When your licensed markets are too few to randomize across. When your category is one the major CTV publishers will not accept, which for several crypto products is simply the answer. When the compliant creative cannot say anything persuasive. And when the funnel is still being rebuilt, because a media test run across an onboarding rewrite measures the rewrite.

One claim about CTV worth treating as a hypothesis, not a finding: the channel is often argued to help an unfamiliar financial brand feel established enough to trust with bank credentials. That is plausible and it is also exactly the kind of claim a vendor has an interest in. It is not something a dashboard will show you, which is the argument for testing it rather than assuming it.

We run CTV, in-app and rewarded on one platform, so we have an interest in what you conclude about any of them. That is the reason to ask us to help design a test against our own numbers instead of taking our reporting at face value. If you want to work through which event to price and how to measure it for your funnel, get in touch.

FAQ

What should a fintech app optimize CTV toward?

A funded account, a completed identity verification, or a first transaction, depending on where your product starts earning. An install on its own says little about whether the user will clear underwriting, and in a vertical with heavy verification drop-off the install-to-funded ratio varies enough by source that optimizing to installs works against your own economics.

Can you attribute a fintech conversion to CTV on iOS?

Not directly. Apple's frameworks provide no path for a television impression to originate an attribution, and even for in-app sources the final conversion window closes 35 days after first app launch, with the postback arriving up to 144 hours later and carrying only a coarse value. Use your own funnel data for the conversion and a geo holdout for the channel's contribution.

Why do fintech cohorts show low conversion from install?

Sometimes because the users were declined, not disengaged. Identity verification and risk review remove users by design, so reporting that cannot separate abandonment from decline will blame the media for the underwriting. The reverse also holds: a source with a persistently high decline rate is usually telling you something real about its traffic.

What do the app stores require from financial services apps?

Both regulate the category and they differ. Google Play requires a Financial features declaration, compliance with the regulations of every region the app targets, and for personal loan apps a store listing disclosing maximum APR, repayment period and a representative total cost, with no US loans at 36 percent APR or higher and no repayment terms of 60 days or less. Apple's 3.2.2(ix) imposes a 36 percent APR cap with no geographic qualifier, bars binary options apps entirely, and under 5.1.1(ix) expects apps in regulated fields including banking and crypto exchanges to be submitted by the legal entity providing the service.

Can you advertise a crypto product on CTV?

Often not, and the constraint usually comes from the publisher rather than the app store. Roku, for example, requires prior permission for any cryptocurrency advertising, approves only a narrow set of products case by case and in the United States only, and prohibits the rest including initial coin offerings, single tokens, and crypto loans. Confirm category acceptance with each platform before building a plan around the channel.

Is rewarded advertising appropriate for fintech?

It can be, provided you price a deep event instead of an install and can confirm that event quickly enough for the reward to pay out. Confirmation latency from manual KYC review or bank settlement is the usual obstacle. Settle fraud controls and clawback terms before launch, and involve your financial crime team, since a synthetic identity clearing KYC is a compliance problem rather than a billing one.

‍