After raising $36.4 million and spending almost two years on the daily grind of parking customers’ cars, San Francisco-based Zirx decided in February to say no more to the madness of dispatching valets around the six cities where it operated. Instead, it’s now solely focusing on providing services to other businesses that need to have cars moved, filled with gas, or washed, such as rental car companies. Zirx says it’s had other businesses as customers all along, but has been solely focused on that category for the past four months.

“We found it very difficult to run both the consumer and the B2B offering,” co-founder and CEO Sean Behr told Fortune in an interview.

To make it easy for businesses to use Zirx’s services, the company has built a new online portal that lets them order and manage services, get price quotes, and keep track of their cars and Zirx’s staffers in real time. The company also announced Monday two of its new customers: car repair marketplace OpenBay and car leasing company BAMA Commercial Leasing, which leases cars to some Uber drivers.

So why did Zirx quit the on-demand valet business?

There were a number of factors, but like the rest of the so-called “on-demand economy,” it boils down to making the math work out. Like services that deliver burritos, laundry, house cleaners, or a ride to the airport with just the tap of an app, valet parking comes with massive labor costs. There’s a sort of Catch-22: It needs enough valets on duty to make sure that customers don’t wait too long, but at the same time, overstaffing is the equivalent of throwing money out the window.

That’s because although the company makes money when a valet is picking up or dropping off a car, it’s actually losing money when that valet is idle. In a perfect world, Zirx and its competitors would build up enough demand so that its valets are never idle. But that’s a major challenge, according to Behr.

“In the consumer business, you’re on the hook for customer acquisition, you’re buying ads on Facebook, you’re giving out fliers,” he said. “When you launch a new on-demand market in the consumer side, the choice of the market and the responsibility of getting the revenue to come all rests in the on-demand company.”

Get Data Sheet, Fortune’s technology newsletter.

And that’s not to mention the industrywide battles over whether workers at these types of companies should be classified as contractors or employees with benefits. In December, Zirx became the latest company to be hit with a lawsuit over this practice. Ride-hailing companies Lyft and Uber are embroiled in similar high-profile cases.

Now, Zirx’s demand is much easier to forecast since most of the time its business customers book services in advance.

But another challenge in the parking business is the parking spots themselves. As Behr explained, Zirx had to prepay for the parking spots it wanted to use every day, regardless of whether it ended up fully using them. In the event of a slow day or even a couple of hours, that meant a loss for the company.

Other services in the parking world have taken different approaches that enable them to bypass these costs. SpotHero, a Chicago-based company that lets customers prebook a parking spot in a garage as they get into their car, doesn’t have to pay for the parking spots in advance.

“We only have to pay a parking operator when a parking space sells. We take a commission on extra revenue that we drive,” SpotHero co-founder and CEO Mark Lawrence told Fortune via email. “In our model we make money every transaction no matter what,” he added.

For more on on-demand services, watch:

As Behr pointed out, other categories of on-demand services also enjoy the advantage of paying for labor only when needed and making money from every transaction. In food delivery, for example, companies like Postmates take a cut from each food order.

And Zirx is not the only startup to find out that not all on-demand models are created equal. Home cleaning service Homejoy and virtual assistant company Zirtual, for example, both shut down last year.

Still, it’s not yet clear whether the on-demand valet parking craze will be a fad or an enduring new service. While Zirx has shifted away from parking regular people’s cars, New York City-based Valet Anywhere has moved to monthly plans, and smaller startups like Vatler and Caarbon completely shut down in the last few months, Luxe recently closed a $50 million round led by car rental company Hertz.

And according to Behr, Zirx will be profitable next month in all eight of the markets where its new enterprise-focused service operates. That’s in stark contrast to its inability to turn a profit in any of its cities when it was operating its consumer-facing service. By the end of the year, the company plans to add an extra five to 10 cities, according to Behr.