Lean Startup Stories: Voxy

Giff Constable lean

Paul Gollash, the founder and CEO of Voxy, just came and spoke at our NYC Lean Startup meetup, and I wanted to share some of his stories. These notes are from my memory, but hopefully reasonably accurate.

Voxy is a language learning startup that is around 2 years old, VC-backed and has grown to about 25 people. But at the beginning, it was non-technical-founder Paul with an idea in his head about how to make language learning more compelling.

Paul didn’t know how to code, let alone make a clickable prototype, but he did know how to use powerpoint. He built some paper prototypes in Powerpoint and got out of the building. His target market was hispanics who wanted to learn English, so he would go to parks in Hispanic neighborhoods and offer to pay people $20 for 20 minutes of their time. About one out of three people said yes, but that was enough to start learning, evolving and focusing his ideas. He also would go to hotels and talk to the hispanic staff.

Eager to gather more data, Paul took out ads in a local Spanish paper, with a phone number to call. When someone called the number, Paul would invite them to meet him at a nearby Dunkin Donuts for a customer development session for $20. Cust dev tip from Paul: don’t put your personal cell number in an ad. It started ringing at 4am and literally did not stop ringing for 3 days.

However, Paul didn’t just get out of the building. He hired an out-of-work videographer to follow him around with a Flip phone and videotape the whole thing. He then took that video and edited it down into a 2 minute clip (with English subtitles) that played a big role in landing a first seed round of funding. You can see some of the videos if you go to the earliest clips in their YourTube channel.

I want to highlight this takeaway for anyone pitching a product to VCs that they won’t grok at a personal level (female entrepreneurs take note). The video put the customer and the customer’s problem right in the investor’s face, with inescapable evidence, and also showed off Paul in a great light as a scrappy entrepreneur getting out into the market.

The biggest problem Paul faced with his qualitative testing was that it was very difficult to measure willingness to subscribe and pay. That experimentation really kicked into gear once Paul had his first web and mobile apps. If memory serves, I think Paul started by working with programmer contractors and then hired a full-time team.

As Voxy made progress, they were seeing more and more interest in human language tutoring, not just human-computer. But they didn’t know how to execute on it, how to price it, or even if interest was big enough to justify working on it. The first thing they did was to put an interstitial ad into their own product, which offered a VOIP tutoring service at different price levels. Then they studied conversion levels in combination with qualitative research. He also had his customer support person start mentioning the service when appropriate in customer support calls.

He collected all his data using Wufoo forms, in order to avoid needing a developer to set things up. Why, given that he now had dev resources? First, he didn’t want to distract his dev team who were working on more critical stories. Second, he didn’t want to have the dev team’s capacity become a bottleneck for experimentation.

One thing we spoke about was doing lean testing on mobile, and Paul’s advice is to start on Android, iterate there and then bring your lessons over to iOS. You’ll just get greater speed and flexibility. Some techniques are not even possible on iOS, for example, Apple rejected their app with the interstitial ad for tutoring services that I mentioned before. They were confused by it and the fact that a price was mentioned (ed. note — as Dropbox found out, they are going to ridiculous degrees to block anything that might lead to a transaction where they are not getting their 30% cut).

One thing that became obvious to me is how data-driven Paul and Voxy are. Maybe I can convince Paul to write up their approaches to data and metrics in more detail.

Voxy is also now dedicated to regular qualitative research. They bring in 4 people a week to do lightweight user tests. He and Andres Glusman (head of product at Meetup.com) got into a great discussion about what to do with the data from these user tests. Andres’ advice (as I understand it) is to minimize documentation work wherever possible. Meetup uses GoToMeeting to let multiple people observe a session, and they disseminate key takeaways, but they do not go overboard with documentation or reporting.

Net-net, Voxy is a young company with solid success under their belt and yet a lot to prove and accomplish. My takeaway is that they have done a great job merging vision and educational theory, on the one hand, with a very practical, humble willingness to experiment and find out what really works, on the other.