In 2013, Uber introduced an option to evenly split the fare for a ride with up to five other passengers. Data isn’t readily available on how successful the fare splitting option actually is, once Uber started doing it, many were eager for Lyft follow suit—which it did the following year. At the time, these were pretty groundbreaking revolutions: Automating the process of settling up, which used to be done by handing over some cash, buying a drink in exchange, or firing up a P2P payments app.

When customers create an account with a ridesharing service, they link a credit or debit card, so that each time they take a ride, the card is automatically charged. It makes for a fairly frictionless process—and it’s a big part of the improvement over transportation alternatives.

But fare-splitting convenience doesn’t come free: For both services, splitting the fare costs each passenger an additional $0.25. Adding a fee to process separate transactions—as splitting entails—makes some sense considering the fact that businesses have to pay a fee every time they charge a different credit or debit card. But that’s only part of the story: In fact, the transaction cost is marginal compared to the added risk that fare-splitting adds. Every extra card introduces an additional opportunity for fraud, which would be charged back to the ridesharing service.

It bears mentioning here that carpooling ride-sharing options, on the other hand, evidently treat each passenger as a separate fare, leaving out the fare-split fee. The passengers are quoted a flat discounted fee when they call the ride and are charged that fee regardless of whether or not additional passengers are picked up.

The situation illustrates how, whenever payments are concerned, there are tradeoffs between making things easier for customers without crushing margins or introducing unmanageable risks.

Transaction fees

Like all businesses that accept credit cards, ride-sharing services are charged a fee to process card transactions. For the most part, these interchange fees and assessments equal a percentage of the transaction plus a small flat fee. Interchange fees generally range from 1 to 4 percent of the transaction, depending on the credit card.

According to data from SherpaShare, the average Uber ride in the United States costs $13.36, and the average Lyft ride is $12.53, processing one credit card that could cost upwards of $0.50. Debit cards are charged a flat processing fee of up to $0.21 per transaction.

“Swipe fees,” as they’re called, are generally considered to be a major problem for merchants—and for consumers in the long run. The National Retail Federation estimates that the average household spends $400 a year on swipe fees alone.

So though the added fees are small, they do add up. And when multiple credit cards come into play, as in the case of splitting a bill, merchants often balk at the prospect of having to pay multiple transaction fees. And who hasn’t had a waiter refuse to split the bill among multiple cards?

That’s why restaurants sometimes add a surcharge for splitting the bill (although this practice is currently illegal in ten states), limit the number of credit cards, or offer a discount for paying in cash.

Fraud risks

Though merchants like restaurants offer some precedent for handling transaction fees posed by splitting the bill, ridesharing companies’ innovations can make them subject to heightened risks.

Though separate from fare-splitting, as with any merchant, they’re subject to a fair amount of unauthorized card usage. In recent years, for example, there were a spate of issues referred to as “ghost rides,” wherein users were charged for rides they didn’t take due to credit card fraud. And a recent report by Trend Micro found that compromised Uber accounts fetched more than stolen credit card information as a way for criminals to siphon money.

Ridesharing services also face so-called friendly fraud, when legitimate rides happen, but payments are reversed by riders. While rider data (and regular habits) do give them better ideas about customer behavior, applications of this data can be tricky, and knowing their customers in this way can sometimes only get them so far.

That’s why, in addition to charging a modest fare-splitting fee, such services are often working on their risk models.