2011 was the most exciting year of my life. Wow. I was blessed to raise venture capital (VC) from the most active venture capitalists in the education space of the USA which enabled me to assemble a phenomenal team. We launched as a TechCrunch Disrupt Finalist in September.
We learned more from the launch of the below five products in one year (2011) than I had learned in the first three years of the company. For the quick read, skip over everything but the pictures/videos under Solution and brief thereof.
#1: Lily - March 2011
It's difficult to find an easy, casual, way to learn Spanish on the go and to continually stay motivated in doing so. In developing markets there is a 50% smartphone penetration but in emerging markets it's typical lower than 20% (source). One thing all cell phone users have in common is that they all are capable of text messaging (SMS). Did you know that the average teen in the USA sends over 3,000 texts per month - once every 10 minutes (source)?!
See a pdf of Lily in action here
Lily was a Spanish teacher in your pocket. Through interactive text (SMS) messages, she proactively quizzed you in Spanish, translated anything you wanted from/into Spanish and called you to pronounce any Spanish word(s). As Lily was an intelligent robot (many thought she was a real person!), she employed adaptive learning (the less you knew something the more often she would ask it).
We organically brought in over 30 users a day on average with less than a 10% weekly churn rate. The business model economics were broken, however. Sending a text legitimately (via Short Code, not Long Code) cost us $.02, receiving cost us $.02 (through Twilio). Our customer development resulted in finding that users weren't willing to pay more than $2/month for the service. We'd have to charge around $5/month to even break even as the average user exchanged eight text messages per day.
p.s. we also built Lily for Facebook (video) but never launched it
as we weren't confident enough in it's prospects
#2: iPhone Apps - May 2011
It takes a lot of time to make the paper flashcards for the vocabulary one needs to learn for their weekly quizzes/tests in their foreign language classes. Even if one did make paper flashcards, they wouldn't have them at all times. Paper flashcards don't have audio either.
Watch a video of the iPhone Apps in action here
A smartphone application that has the specific vocabulary of your language learning textbook with the ability to quiz yourself in a flashcard-type fashion with both text and audio support. Your learning is optimized through adaptive learning (the less you know an item the more often it shows).
We launched applications for the respective foreign language textbook titles in our partnerships with esteemed publishers. The nine applications that we launched are still live (find them here) and are continually in the top 25 charts of various countries around the world. Students that are aware of the applications typically purchase them and find them very useful. The two main reasons why we are not seeing as much success (i.e. revenue) as we'd like are 1) for awareness, we're reliant on the publishers to market such apps in-print (...we're still waiting for that) and 2) users lose engagement in rote study through flashcards. 1) will be resolved by the publisher's in-print promotion and 2) will hopefully be resolved in our next app update that will employ game mechanics to increase (at least short-term) engagement.
#3: (De)Construct - September 2011 (launched as a TechCrunch Disrupt Finalist)
Learner's lack of sustained engagement. Current language learning solutions are effective, if you use them. None of them, however, have proven to keep a language learner engaged. Did you know that less than 5% of Rosetta Stone's users continue past the first level? The average user time spent on a Rosetta Stone product is 2.5 hours over 6 months - you don't need to be a polyglot to know that's not even close to enough to learn (or even sustain one's knowledge of) a foreign language (source).
PlaySay will sustain language learner's engagement through a language learning layer on Facebook, the most engaging platform of the world. Users will furthermore remain engaged as the language learning solution revolves around communication - what they're learning a foreign language for in the first place. This was the world's first method of learning a foreign language by communicating in that language through pictures.
Facebook became your new classroom, your friends became your new classmates, and your checkins, status updates, comments and pictures became your new daily course material.
Your friends' interests were leveraged for learning.
Albeit incredibly innovative, our initial application's user interface (UI) was too complicated. Less than 30% of users completed a 'construction' after starting one and less than 15% ultimately posted it on their/friend's walls. Even worse, our application gave a false promise to the users by telling them they'd be able to communicate in the foreign language (through pictures) however they wanted. In fact, users could effectively only compose (and post to Facebook) a few basic sentences with slight variations in the language they were learning. Our system also didn't allow viewers of these posts to respond. Although 30% of users clicked their friend's public 'constructions', hardly anyone responded with another 'construction'. This meant that all communication was one-way, which is not communication at all. Oops.
We then rolled out a section in the app that involved a series of levels with questions that a user would answer through the 'construct' methodology. This greatly improved 'construct' completion to 90% (from 30%). It also increased time in app from a few minutes to over 10 minutes, with 60% of users completing level after level (less than a 40% average churn per leve). Although 30% of users were weekly active users (WAU's), hardly any were monthly active users (MAU's) and even more of a negligible amount were daily active users (DAU's).
By far, the section of the application that had the highest usage and engagement was the third section that we rolled out that involved a series of simple True/False questions of your friends' interests. Simple is always better. Users' average time in this section surpassed their time spent in the aforementioned 'construct' and 'questions' sections.
To our dismay, our k-factor (growth factor) for all sections of the app was actually negative as we were losing users much faster than we were gaining them.
#4: Journeys - November 2011
Learner's lack of sustained engagement (see Problem in product #3 above).
See a video of Journeys in action here
We still felt that leveraging Facebook's Open Graph was a key to solve the engagement problem for language learning. Rather than an overcomplicated method of communicating in a foreign language through pictures, we attempted to revolve a language learning platform around 'journeys' through actual Facebook users' pictures throughout the world. Such journeys would serve as the story/lesson and the respective pictures and related vocabulary as the learning material.
Engagement was much higher in this product than our (De)Construct product, but not anything to be proud of. Our largest improvement was when we implemented the 'passport' dashboard that employed game mechanics such as levels, unlocks, badges, etc. This decreased churn by 10%, increased time in the app from 1.5 minutes to 5.5 minutes and doubled users weekly rate of return (i.e. weekly retention). Unfortunately, user growth remained stagnant for over a month no matter what viral tactics we attempted to integrate.
#5: Tagalang - December 2011
Yet again, learner's lack of sustained engagement (see Problem in product #3 above).
See a video of Tagalang in action here
Although this product was never publicly launched, it attempted to teach users vocabulary by leveraging Facebook's most popular/engaging feature - pictures. Similar to the ESP Game, friends (and/or strangers) were paired to tag vocabulary items of a picture. Our twist was to leverage Facebook's Open Graph to have the picture be one of interest to both users. Furthermore, we collected the exact coordinates of the discrete vocabulary items users would tag. We used translation software in the backend to translate all input in real-time to allow users to play regardless of their native language or language of learning. Triple validation algorithms were employed to validate input and to collect a valuable data asset.
Alpha users we tested this with in private loved it. A competitor heard about the product through the grapevine and even approached us to acquire the game. We believe this product has legs but have not released it for the sake of focusing on our next product that we believe has much more potential. We learned from the success of Tagalang's simplicity and are surely incorporating those learnings into our next product, "PlaySay Survival".
Sign up for beta access to PlaySay's biggest product yet! http://signup.playsay.com