Insight/Action/Outcome: Shortly after implementing Analytics, LingQ noticed users who completed at least one lesson by the time they reached a paywall were significantly more likely to convert to paid subscribers. Now, the team is leveraging CDP to test which lesson to offer new users and then splitting that into three smaller segments to increase the user’s sense of accomplishment.
Understanding the complexities and nuances of language is a big part of what makes us human. I’ve lived in many countries around the world, and as I’m fluent in three languages (and not-so-fluent in another three), I have a lot of experience learning languages. In hindsight, it was probably only a matter of time before I started working for a language-learning company.
That company turned out to be LingQ, which father-and-son team Steve and Mark Kaufmann founded in 2007. LingQ’s difference comes from providing lots of authentic, engaging content in a target language—and LingQ has one of the best foreign language libraries around. This approach has proven successful for LingQ’s million members.
Despite our global reach, we’re a small team. We have developers on each platform, plus teams for design, QA, marketing, and customer service. My role as Head of Learner Success is a mix between product and marketing: I work with our developers and designers to create experiments and find ways to improve the customer experience. Rather than setting targets or timelines for milestones, we’re much more focused on improving our systems. We want to build a consistent flywheel of gathering user data, improving our product, and enhancing the user experience.
Improving systems requires confidence in data. But up until recently, we didn’t have the confidence in our data that we needed.
We Couldn’t Continue Our Upward Trajectory without Analytics
I had been at LingQ for a couple of years when we started considering reevaluating our tech stack. We had grown and evolved, so looking at solutions better suited to our trajectory made sense. We wanted to understand better which users were converting to our premium plan, to scale how we sync data to third-party destinations as our stack grew, and to make recommendations for courses to users to increase retention and conversion.
We had been using CleverTap for messaging and analytics, but despite doing both things, it doesn’t do either particularly well. Our desire to boost our efforts around email and push notifications led us to adopt Iterable for messaging, which meant we needed a tool that offered targeted insights and better data infrastructure. That tool turned out to be Amplitude Analytics.
We especially liked how the platform handled the properties/attributes of users, which can change over time.
Consider the simple distinction of being a free or paid user. We want to make that distinction to understand how a user performs across various tracked events, which allows us to identify when that user converts from free to paid. Many platforms will retroactively change historical data to a user’s current attributes, but Analytics is different. Whether it’s a cohort retention or funnel analysis, Analytics shows user attributes at the time of a specific event, even if an attribute changes at a later event. That lets us get more accurate insights when comparing user properties at different points in the journey.
In addition to Analytics, we adopted Amplitude CDP (customer data platform). We tracked many events and used messaging features in CleverTap, but we also sent many emails and push notifications outside the platform. We had no way to consolidate all that data—at least not in a cost-effective way. But in Amplitude CDP, all of these messages are recorded as events.
That capability unlocks options for storing data as user or event properties, which allows us to gather all customer data in one place and send it directly to Iterable. That saves time and developer resources and ensures consistency across platforms. And if we ever want to switch to a different messaging tool, all we have to do is turn off the faucet to Iterable and send the data to the new tool without having to retool our events and properties.
A Fresh Implementation Leads to Fresh Insights
The first thing we did was set up the data infrastructure. Although we’re still working to implement some of our events, we are collecting data and have already uncovered some valuable insights.
One example involves looking at conversion rates relative to lessons completed and words saved. LingQ’s methodology is about tailoring language learning to content that interests the user, and one of our most powerful tools in the app is a private vocabulary database where users can save words and phrases for later review.
Our subscription model allows free users to save 20 words before they hit a paywall (although many users will subscribe to a paid plan before that). We looked at the funnel from the 20th to the 30th saved word and analyzed how the conversion rate compares for users who hadn’t completed any lessons versus users who had completed one or more lessons by the time they hit that paywall. We found that users who completed at least one lesson by their 20th saved word were much more likely to convert and reach their 30th word—by a statistically significant amount.
It makes sense: Completing a lesson comes with a sense of achievement, and users are more likely to chase that feeling of accomplishment. Using that insight, the team is now testing which lessons to offer new users and splitting that first lesson into three shorter lessons to make completion less daunting.
Having definitive insights makes the case for obvious wins worth our time. Without that data, we’d just be making assumptions, and sometimes a seemingly obvious assumption turns out to be wrong.
There are so many variables to try when tweaking a product. Having definitive insights makes it obvious to everyone which wins are worth our time. Without that data, we’d just be making assumptions, and sometimes a seemingly obvious assumption turns out to be wrong.
Amplitude is the only tool offering this level of insight coupled with such strong data governance capabilities. In CDP, we can set up a tracking plan that alerts us if 25% of instances of a specific event don’t contain the expected event property, for example. That’s valuable, especially as we have three different platforms that we could instrument in slightly different ways. Cumulative tracking can be difficult, but CDP makes it super easy.
Finding Opportunities for Growth
Analytics has a lot of charts to analyze user behavior and composition that weren’t available to us before. We hope to uncover areas to focus our experimentation and to provide relevant lessons to users. It’s a complex task, but Amplitude’s data and product suite will help us get there more effectively.
After onboarding Analytics and CDP, we will turn to Audiences. Audiences will help us see which lessons lead to the most engagement and personalize the user experience based on that information. LingQ is complex—we have 45 languages, and there are a lot of different variables. With Audiences, creating a hyper-personalized list of lessons that will be the most engaging for any given user and ultimately help them learn quicker.
The more insights you have, the better hypotheses you can make about what will lead to growth, and the greater your confidence in making those decisions.
For LingQ, product-led growth opportunities come from improving messaging, onboarding, and our product. Using the insights we get from Amplitude products, we expect to uncover the opportunities in those areas that lead to growth. More insights means more and better hypotheses on what changes will lead to growth, and greater confidence in deciding which path to follow.
We have unlimited choices and directions we could take LingQ. Data makes those choices more transparent, so we can dedicate time to making the right changes with the most significant impact—and help people learn the nuances and complexity of any language they wish.