This story is part of CNBC Make It’s The Moment series, where highly successful people reveal the critical moment that changed the trajectory of their lives and careers, discussing what drove them to make the leap into the unknown.

Sometimes, you need to recognize that pretty good isn’t good enough.

Just ask Spenser Skates and Curtis Liu. Today, the two MIT graduates are known as co-founders of Amplitude, an analytics software business with a market cap of $1.35 billion and more than 2,300 corporate clients. A decade ago, they were on a completely different track — running a voice recognition startup called Sonalight.

It was “a version of [Apple’s] Siri before Siri even existed,” Skates, Amplitude’s 35-year-old CEO, tells CNBC Make It. The duo created Sonalight in 2011, earned a coveted spot in Y Combinator’s startup accelerator program and even reached 500,000 downloads of their app.

Then, they made what might sound like a surprising call: They shut it down.

Internally, Skates and Liu saw that people were using the app once, but not re-engaging with it repeatedly. “Sonalight was a 95th percentile idea,” says Skates. “Most ideas are terrible. It was pretty good, but it’s not the bestest best. And it was like, we should probably go for a 99th percentile idea and go find that.”

They found it in their analytics tools, which they’d built in-house to get insights into their users’ habits. “We probably spent half of our time doing that — this, like, silly hubris mistake by engineers, seeing if they could build it,” Skates says.

Yet at Y Combinator, the tools proved more effective than anything their peers were using, he says. Skates and Liu began working on Amplitude in 2012 and officially launched the analytics platform in 2014 along with an additional co-founder, Jeffrey Wang. By 2021, Amplitude had raised $336 million from investors, and Skates decided to to take the company public.

Here, Skates discusses the risks of ditching Sonalight, how to build great ideas rather than good ones and why software engineers don’t always make the best startup founders.

CNBC Make It: Why did you decide to move on from Sonalight? Did you worry that you might be trading a decent idea for one that wouldn’t work at all?

Skates: There’s always that risk when starting something new, but it actually wasn’t that hard of a decision for us. I think the question was: How successful was Sonalight going to be?

We had this really cool, magic demo on stage [at Y Combinator] where I put my phone in my pocket, I talked to it and had a conversation back and forth with it. We got this amazing amount of press out of it, a little bit of seed investment and like, 500,000 downloads. So it was like, “OK, this is real. Someone’s using this. This is cool.”

But we’d been working on it for almost a year, and it was starting to become clear to us that the technology wasn’t good enough, in terms of creating a great user experience and getting people to engage and come back. It wasn’t useful enough as a product to be really sticky.

We could probably grind on it for the next four or five years and get to some OK success as a company, but it wouldn’t be some breakout, massive success. [Sonalight] was not the best thing we could work on. There was more impactful stuff to do.

Once you decided to change gears, how did you pick the “best” new idea to focus on?

We spent a month just talking and exploring different ideas. You really want to find a problem that fits your strengths, weaknesses and interests. Voice recognition was almost too hard technically to solve. It’s like this probabilistic problem where there’s not a clear right answer.

Analytics, to the average engineer, it’s a pretty hard problem — but to us, it was a cakewalk, because we were algorithms guys. Building a distributed data store was very straightforward for us. It’s like, “OK, that’s a solvable problem with a clear answer. If we do it, people want it. Great, let’s go work on that.” It was a million times easier.

We had built our own analytics in-house. What was interesting was a lot of the insights we were getting about [Sonalight’s] customer journey, so many other companies wanted those exact same insights. We were like, “OK, this is fantastic.”

We talked to 30 companies, and found enough that had the need. So we started building.

Why did you feel the need to chase “breakout, massive” success? Was there something wrong with good enough?

[After college] I spent a lot of time thinking: How is it that I can have such a positive impact on the world? What do I know? I know how to build software. Let me figure out the biggest way I can do that.

I worked for a year in finance and high-frequency trading. I was trying, at the time, to recruit a lot of my friends from MIT [to build a startup]. I asked classmates, peers and other folks. No one wanted to start a company.

You know, the funny thing about engineers: A lot of people would talk about starting a company and get really excited, but very few would actually take the leap. Most of them would go to Google and get sucked in and just never come back.

Engineers are a risk-averse bunch. They want to do it only if there’s a clear path to success and there’s validation along the way. But startups and entrepreneurs, it’s the exact opposite. You have to be willing to take on all that uncertainty and risk yourself. Your bosses and your teachers aren’t there. No one’s there to be like, “Hey, you’re doing good.”

Either you have something people want or you don’t. You have to be willing to see through that, and see the potential in what you’re doing.

This interview has been edited and condensed for clarity.

CNBC