The Odds Are Against Financial Superapps

Prediction markets and the hidden cost of churn

Prediction markets are having a moment. I’m a believer in the underlying idea. I’m far less convinced they belong inside financial superapps like Robinhood.

I’ve long believed that distribution is the hardest part of building a consumer financial business. If you have it, you should be able to expand from there. But that expansion only works if it strengthens, rather than dilutes, the core value proposition.

---

Distribution Compounds. Poor Products Interrupt It.

At Inversion, we spend a lot of time studying businesses with real distribution and direct connectivity to end users, and asking whether financial modules can be layered on top. The logic is straightforward: if users already trust you, the marginal cost of offering additional financial products can be attractive.

Crypto infrastructure meaningfully changes that calculus. Neobanks successfully digitized the interface, but they remained tethered to expensive legacy banking rails that predate the internet. Crypto collapses the cost of the rails themselves. It shifts finance from a rented, intermediary-heavy model to an open-protocol one. Wallets, programmable money, and global settlement are now cheap, accessible primitives.

That doesn’t mean everything is suddenly inexpensive. You have to be precise to avoid being wrong.

Where cost truly collapses:

• Settlement: Moving $10 million globally via SWIFT takes days and involves correspondent banking fees. On a blockchain, using stablecoins, it settles in seconds at negligible cost.

• Ledger maintenance: You don’t need teams reconciling proprietary databases. The blockchain becomes the shared reconciliation layer.

Where cost remains stubbornly high:

• Compliance: In many cases, compliance costs in crypto are higher than in traditional finance. The regulatory environment is fragmented and aggressive.

• On- and off-ramps: Moving fiat into and out of crypto still touches legacy rails, with interchange fees, wire fees, and bank dependencies intact.

Crypto removes middleware. It doesn’t just wrap the bank. It replaces the settlement layer. Builders move from a high-variable-cost model, paying intermediaries at every step, to a utility model, paying network costs. The difference is akin to Netflix mailing DVDs versus streaming video directly.

This structural shift is why the idea of financial superapps keeps resurfacing. Lower costs. Broader product menus. Higher lifetime value. In theory, everyone wins.

In practice, sequencing and product mix matter enormously.

---

The Silent Killer: Cost to Serve

Most superapp strategies obsess over customer acquisition cost. The real damage usually shows up elsewhere: in cost to serve.

This is the quiet tax that legacy institutions consistently underestimate. A superapp fails when the friction introduced by a new product exceeds the efficiency of the existing infrastructure.

Consider the whale-versus-minnow problem. A bank built around high-touch systems for wealthy clients cannot simply pivot to millions of users with small balances. If a $500-balance customer requires human intervention—even once or twice a year through compliance reviews or support calls—that customer is structurally unprofitable.

The true digital-native advantage isn’t a nicer interface. It’s an automated backend that keeps cost to serve close to zero. Without that, adding products doesn’t expand margins. It creates an operational sinkhole.

---

Why Prediction Markets “Make Sense”

From Robinhood’s perspective, adding prediction markets isn’t irrational.

The same users who traded GameStop options are plausibly interested in betting on Sunday football or political outcomes. Behaviorally, the overlap is real. At some level, it’s all options. Platforms like Polymarket and Kalshi have raised substantial capital pursuing this thesis, and many industries (from casinos to insurers) want exposure to the category.

So yes, on paper, it fits.

But strategy isn’t about what you can do. It’s about what you should do.

---

Where We Draw the Line

When we evaluate adding financial modules to businesses with distribution, we start with a single question: does this improve unit economics without increasing churn?

We are extremely sensitive to churn. If a new product provokes more of it, you’ve failed no matter how attractive the near-term revenue looks.

Extracting more wallet share is tempting. Doing it without destabilizing the user base is much harder than it sounds.

The strongest financial superapps prioritize long-term lifetime value over short-term monetization. They avoid products that increase volatility in user behavior, even when those products are profitable in isolation.

There’s a reason traditional banks (and most neobanks) have largely stayed away from gambling-adjacent products. Not because they don’t understand the economics, but because they understand the risk profile.

---

Churn Is the Hidden Cost of the Casino

Robinhood’s user base skews more speculative than that of a traditional bank. That’s not a moral judgment; it’s an observation.

The problem with casino-like products isn’t that users lose money. It’s that casinos accelerate churn.

The longer you exist inside a casino, the higher the probability of liquidation. And liquidation means you’re out of the game entirely. A churned user is worth zero.

Products like Robinhood succeed initially because they are simpler, more accessible, and more digitally native than incumbents. They attract users early often when traditional institutions feel hostile or unusable.

But users age. Over time, the real opportunity is to grow with them and capture more of their financial lives not to maximize extraction at the moment of peak speculation.

If durability matters, you optimize for staying power.

---

Adjacency vs. Distraction

If I were in the seat, I’d prioritize products users naturally want as they mature financially: credit cards, insurance, savings vehicles. These are boring. The data suggests that’s precisely why they work. They are adjacent to the core relationship of managing household liquidity.

The success case: Nubank.

Nubank acquired users through low-margin, no-fee credit cards and gradually migrated them into higher-margin lending and insurance products. Early cohorts took more than a year to adopt three or more products. Newer cohorts reach that density in under three months. While Nubank’s blended revenue per active customer is roughly $9, mature customers using adjacent products generate more than $24. That lift came from deepening trust, not testing it.

The failure case: Goldman Sachs and Marcus.

Goldman attempted to pivot from high-margin investment banking to mass-market consumer finance. By early 2023, the firm disclosed roughly $3 billion in cumulative losses over three years. Personal loan originations were halted, and the BNPL business was sold at a loss.

Nubank succeeded because insurance and lending are structurally aligned with banking. Goldman failed because high-volume, low-balance consumer finance is culturally and operationally hostile to an investment bank.

A financial superapp like Robinhood moving into gambling looks far closer to distraction than adjacency. Product distraction is where durable businesses go to die. Study Raising Caine’s.

---

Why I’m Still Skeptical

Some recent signals reinforce my concern.

Earlier this year, Robinhood announced tokenized exposure to private companies. The rollout was poorly communicated. Retail users were led to believe they were accessing private equity of hot startups, when in reality they were buying derivatives that merely tracked price movements. That distinction matters. Owning a derivative is not the same as owning the underlying, and misalignment here can end badly for retail.

At the same time, Robinhood’s credit card and rewards products are a bright spot. Nearly every power user I’ve spoken to loves the card. That’s the direction I’d lean into: turning a trading app into a full-fledged consumer finance platform through adjacent, trust-building products.

Instead of doubling down on gambling (options into prediction markets) I’d expand into credit, insurance, and savings. Solve more of the user’s financial life, not just the speculative slice.

---

The Counterargument, and Why I’m Not Convinced

The obvious counterargument is simple: Robinhood users are “degenerate,” so build the ultimate degen superapp.

I understand the logic. I’m not persuaded.

Many users came for an intuitive, low-friction way to participate in markets not to live inside a casino spanning stocks, sports, and random events. Yes, everything can be framed as a bet. But the human mind doesn’t process all bets the same way. Certain combinations amplify destructive behavior and accelerate exit.

Minimizing churn is the single most important thing. If I had to make just one due diligence request for any company, it would be cohort curves. I want to understand churn; how it behaves and how it evolves across cohorts. If you understand churn, you understand the quality of the business.

My bet is that businesses optimizing for short-term extraction through gambling-adjacent products will look strong early, and much fragile later. Financial superapps that treat churn as a first-class risk will end up with stronger moats and better long-term outcomes.

That tradeoff isn’t obvious. But it’s the one that matters.

Casinos serve just enough alcohol to increase the house edge, but not so much that players black out and leave the table. That balance is hard to maintain. Robinhood and other superapps will FAFO when it comes to hyper-financializing their user base.