ClassPass Changed the Metric Until the Product Worked

Stop polishing the wrong hypothesis

I first wrote about ClassPass in a January 4, 2021 First 1000 issue. The story stayed with me because the team had to abandon a product that looked successful from the outside.

ClassPass did not begin with the subscription people know today. It began as Classtivity, a beautiful search engine for fitness classes that had press, funding, and roughly 10 bookings a month. The product looked busy from the outside. The behavior was telling the team that the core hypothesis was wrong.

The useful founder question is not “does this look like a company?” It is “what repeated behavior would prove that the product is becoming part of someone’s life?”

Payal Kadakia came to the problem through dance. After spending an hour opening tabs to compare ballet classes, she imagined a search engine that would make classes easier to find.

The team spent 16 months building Classtivity. It received press from Inc., Mashable, Business Insider, and other outlets, and raised a few hundred thousand dollars. The site was still making only about 10 bookings per month.

Payal described the next period as the “summer of buttons.” The team changed colors, shapes, and interface details because those were problems they could control. The issue’s lesson is blunt: when a product is not working, polishing the interface can become a way to avoid revisiting the business hypothesis.

The reset started with conversations. Payal asked studios how they got people into classes and called users who had signed up to learn what would make them attend. She paused the scrapers and APIs and went back to the behavior underneath the search experience.

Separate the purchase from the habit

The interviews surfaced a simple pattern: many studios offered a first class for free. Classtivity turned that into Passport, a $49 one-time purchase for up to 10 classes in 30 days, with an average saving of 60 percent.

Passport separated two decisions. A customer could buy the pass once, then choose a class later. The purchase created a reason to use the credits instead of asking the customer to evaluate every studio and every class at the same moment.

The first version was deliberately manual. When someone requested a reservation, the team received an email and made the booking themselves. That work produced a useful observation: around half of cancellations happened within 15 minutes of booking. The cancellation policy came from the manual work.

Passport produced 20,000 reservations in its first six months. The number that mattered to Payal was not just reservations. She wanted reservations per person, because 150 people using the product every week told her more than 1,500 people attending once or twice a year.

That is the metric change I would copy. A top-of-funnel event is easy to celebrate. A repeated behavior is harder to fake.

Let unexpected behavior change the product

Passport was designed as a trial: help someone sample studios, then let them choose a favorite. Some users created new accounts after their 30 days ended so they could keep exploring different studios.

That behavior contradicted the original thesis. People did not only want help choosing one place. They valued the variety itself. The team turned that discovery into ClassPass, a $99 monthly product that worked with venues on referral rates and limited visits to the same venue to three per month.

The combined Passport and ClassPass products reached 100,000 reservations within two months of the ClassPass launch. The case describes Payal targeting a $1 million run rate, using roughly 1,000 paying subscribers as the next milestone.

The arithmetic is old. The decision is still useful: define the behavior that turns a promising product into a business, then let users’ attempts to bend the product reveal what they actually value.

The founder decision

When a product has attention but weak usage, I would ask:

1. Which metric is easy to inflate? Press, signups, and one-off bookings may hide a weak habit.
2. What repeated action would make the business work? Write that metric before the next redesign.
3. What can we do manually for the next 20 customers? Manual service is expensive, but it can reveal the friction a dashboard will miss.
4. How are users misusing the product? A workaround may be a better product thesis than the one in the roadmap.

Sources and original research

• Original First 1000 case: 🏋🏻‍♀️Classpass, January 4, 2021

The ClassPass First 1000 issue supplies the historical pivots, prices, reservation counts, cancellation observation, and metric choices above; the diagnostic checklist is my adaptation.