We were running concurrent experiments - one focused on easy payments for small businesses (think ecommerce version of Square, with tools like invoicing & storefronts), and one focused on an API for crowdfunding sites.
Small business payments got earlier traction, but the economics of it weren't sustainable. (it cost us about $400 to acquire a customer that had a lifetime value of ~$250). We iterated for a bit to see if we could significantly lower CPA or raise LTV, but only made marginal gains.
Our API was started as more of a skunkworks project. It took a little longer to build traction, but it grew like wildfire (double digit monthly growth since launch, sustained over many years). Once platforms integrated it, they stayed, and they grew really fast. So not only were we adding new customers, but they were all growing month over month as well.
Last year, we made the decision to shut down our small business products, and invested further in our API, expanding it to marketplaces & those building SMB software as well.