Rewards on a crypto card are a balance-sheet decision first

The most dangerous sentence in payments is: “We’ll make it back on volume.”

It usually shows up wearing a nicer outfit. “Rewards.” “Cashback.” “Earn.” “Boost.”

But rewards are not a feature. They are a promise to pay money out on every swipe. That promise sits inside a stack of costs and risks most product teams never write down in one place. When they do, the conversation changes fast.

The immediate trigger for this piece is Coindesk’s reporting on crypto card startups racking up losses to buy loyalty with big rewards, with providers betting that trading, lending, and staking revenue will cover those rewards over time.

I believe the framing is right, but incomplete. The real question is not “Are rewards too high?” The question is: can you translate rewards into a unit-economics and balance-sheet story that survives three ugly months in a row?

If you cannot, the program is not generous. It is fragile.

The P&L of one swipe is the only honest starting point

A card transaction looks simple to the customer. Spend, earn, repeat. Under the hood, you are running a tiny financial system per swipe.

Here is the way I force myself to model it, even before I argue about brand or growth:

• Revenue you can count on: interchange and fees that are mechanically tied to card spend.
• Revenue you hope repeats: trading, lending, subscription, FX spread, partner commissions.
• Costs you will definitely pay: scheme, issuer-processor stack, customer support, compliance overhead.
• Losses you will eventually eat: chargebacks, card fraud, account takeovers, and the edge cases you did not imagine.
• Balance-sheet exposures: pre-funding, settlement timing, safeguarding, float management, and any yield strategy sitting behind “rewards.”

Rewards sit in the middle of this. They are not “marketing spend” in the usual sense, because they scale linearly with usage and they create expectations that are hard to walk back. They also invite behavior you might not want, including manufactured spend and bonus hunting. In a crypto context, they can also bring a more subtle problem: people who want the reward, not the product.

If your only reliable revenue per swipe is interchange, you are instantly constrained by geography. The US and Europe do not play the same game.

Europe vs the US: your ceiling is set by interchange

Founders love to argue rewards as if it is a taste question. It is not. It is math bounded by regulation and market structure.

In Europe, interchange is capped. Whether you package it as debit or credit, the top line you can extract per transaction has a low ceiling. If you offer rich rewards off interchange alone, you are implicitly subsidizing every swipe from somewhere else.

In the US, interchange is structurally higher. That gives you more room to fund rewards from the swipe itself. It does not make rewards free, it just means the break-even line is not as brutal.

This is why I do not take “We’ll run the same program in Europe and the US” seriously unless the person saying it can show me the per-swipe model for each region. Geography changes the primary engine.

And that is before you add crypto-specific complexity: rewards paid in volatile assets, yield coming from staking, and a customer base that can turn risk-on and risk-off quickly.

Rewards are funded by three levers, and one of them is a trap

In the Coindesk piece, providers are described as betting that trading, lending and staking revenue can cover rewards. That is the typical trio. Let’s unpack what each one really means when you put it on a P&L and a balance sheet.

1) Interchange: the cleanest money, but capped (especially in Europe)

Interchange is the only stream that is directly proportional to card spend without needing the customer to do anything else. That makes it attractive. It is also why everyone reaches for rewards in the first place: “If I can increase spend, I can increase interchange.”

The problem is simple. In capped markets, interchange rarely funds meaningful rewards after you pay the rest of the stack. If your rewards are meaningfully above what interchange can support, you have already decided to lose money on the swipe and earn it back elsewhere.

That can be a rational acquisition investment, but only if you are honest about which “elsewhere” is real.

2) Funding spread and float: you are now a risk manager

The second lever is treasury. Settlement cycles, pre-funding requirements, safeguarding structures, and cash buffers create float. Some businesses then try to harvest yield on that float.

This is where a rewards pitch quietly becomes a balance-sheet bet. It is not “extra revenue.” It is exposure to rate moves, liquidity timing, counterparty risk, and the operational reality of moving money safely at scale.

If you need yield to pay rewards, your question is not “How do we grow?” It is “What happens when yields compress and our reward promise stays?” If you don’t have a credible answer, you are not running a rewards program. You are running a duration mismatch with a marketing wrapper.

3) Staking yield: volatile by design, and it infects your promises

Crypto rewards are often justified with staking. The logic is seductive: stake assets, earn yield, share yield with users as rewards, win loyalty.

There are two problems.

• Yield volatility: staking yield changes. It changes with protocol incentives, market cycles, participation rates, and risk appetite. Your rewards program, on the other hand, becomes a stable expectation in the customer’s head.
• Asset volatility: paying rewards in a volatile token creates a second mismatch. Your cost is denominated in customer expectations, but your funding source is denominated in a market price.

If you promise “3% back” and your real engine is a yield stream that can halve, you will eventually face a painful choice: cut rewards (and face churn and backlash) or eat losses (and weaken your runway).

Either way, the bill arrives.

The losses nobody markets: chargebacks, fraud, and the cost of being an attractive target

Rewards programs attract two categories of users: the ones you want, and the ones who are there to extract value.

In cards, that extraction shows up as:

• Chargebacks: not just the direct loss, but the labor, evidence gathering, dispute windows, and scheme rules.
• Fraud: card-not-present attacks, account takeovers, synthetic identities, and social engineering.
• Bonus abuse: manufactured spend, cycling funds, referral gaming, and any loophole you unintentionally designed.

In crypto-linked cards, the threat surface tends to widen because you are often combining multiple systems: exchange accounts, wallets, card credentials, and sometimes cross-border rails. Each seam is a place where responsibility can blur and losses can hide.

This is the part most teams undercount. They budget for “fraud” as a percentage. Then fraud arrives as a process problem. Your controls are either real or they are theatre.

I learned this building trading and broker infrastructure: if incentives and controls are not aligned, the fastest learners in the ecosystem are the ones trying to take money out of your system. Not your product team.

What a sustainable crypto-card P&L looks like (and what I would refuse to ship)

So what does “sustainable” look like?

Not a single number. A structure. One that works when volumes are flat, when markets are risk-off, and when fraud spikes.

Here is the model I would aim for, especially in Europe where interchange is not your friend:

• Base rewards funded by predictable revenue: assume the base reward must be covered by interchange plus any fixed subscription fee, after variable processing costs. If it cannot, the base reward is a loss leader.
• Boosts funded by specific, measurable behaviors: higher rewards only when the customer does something that creates margin you can attribute (for example, a paid plan, meaningful FX, or trading activity with provable contribution margin). “We hope they trade later” is not attribution.
• Hard caps and circuit breakers: monthly reward caps, velocity limits, and explicit program clauses that let you change rates. If you do not include these, you are pretending volatility does not exist.
• Fraud and dispute cost treated as COGS: not “other.” Put it next to rewards in the model, because that is where it behaves in reality.
• Treasury policy that is boring on purpose: if yield is part of the story, model it as upside, not as the base. If the program dies when yield compresses, it is not a program. It is a bet.

And here is what I would refuse to ship:

• A headline reward rate that depends on staking yield staying high. That is how you end up cutting rewards at the worst possible time, precisely when customers are most sensitive and competitors are most aggressive.
• A single global reward promise across Europe and the US. Different interchange realities demand different product economics. If you unify the message, fine, but the engine cannot be identical.
• Any P&L that assumes “attach trading later” without a proven conversion path. If trading margin is the subsidy, you need cohort evidence, not belief.

A simple decision lens for the next quarter

If you are building a crypto card, or partnering with one, you can pressure-test the entire strategy with five questions:

1. What is the contribution margin per active card user without trading, lending, or staking? If it is deeply negative, you are buying users, not serving them.
2. Which line item actually funds rewards in a down market? Name it. Model it. If it is “yield,” assume it drops.
3. What is the maximum loss you can tolerate from fraud and chargebacks before you change the program? If you cannot answer, you will change it too late.
4. Where do losses hide? Look for seams between providers: issuer, processor, exchange, wallet, KYC stack. If ownership is blurry, costs will be blurry too.
5. What can you turn off in 24 hours? Reward boosts, referral payouts, category multipliers. Anything you cannot throttle quickly is a potential runway event.

This is not about being conservative. It is about being explicit. The reward program you can explain as a cashflow story is the one you can keep when the cycle turns.

My position: in Europe, a sustainable crypto-card rewards program is mostly a subscription and risk-control business. In the US, it can be more interchange-led. In both places, if your reward promise is funded by volatile yield, you are not building loyalty. You are renting it.

Build rewards that reflect real value delivered to customers, not a temporary subsidy that trains them to leave the moment the subsidy ends. Technology only matters when it changes something real for people. A card that teaches users to chase the next teaser rate changes something, just not in the direction you want.