“Fivetran is critical to us. We use it to copy across our product and content databases that are the cornerstone of all of the data analysis that we do.”
Shrikant Narasimhan, Senior Analytics Engineer, Memrise
- Fivetran has enabled deeper insights into user behaviour, driving product improvements
- Memrise is able to analyse conversion rates when people hit the paywall, making their ‘freemium’ business model possible
- Tracking campaign success data from Apple Search Ads ensures a higher return on investment
- Customer ratings and reviews are now being tracked along with complaints, alerting the team to issues that need to be fixed
- Pipeline: Fivetran
- Sources: Apple Search Ads, Aurora MySQL, Stripe, Google Play, App Store (iTunes Connect), GitHub, Jira
- Cloud Platform: AWS
- Destination: Google BigQuery
- Business Intelligence: Google Data Studio
Founded in 2010, Memrise is a language learning app used by more than 50 million people in 189 countries. Courses developed by expert linguists are accessed via desktop browsers or smartphone apps and offer a fun and effective alternative to textbooks and more traditional pedagogical techniques. Content includes thousands of video clips of people speaking in their native language.
Identify gaps in customer engagement
Memrise built its platform in the cloud, used leading-edge technology from the outset, and knew it would rely on data analytics to inform its design and go-to-market strategies. The problem in the early days was that the five-person data team was spending too much time coding, manually building data pipelines, and fixing broken APIs.
“As a platform team, our core belief is, less code is better,” said Shrikant Narasimhan, Senior Analytics Engineer. “We want to ensure that any work we do internally is focused on making the product and the customer's experience better, not on building pipelines.”
In the first rush of the launch it was all about quick wins and finding an audience, but as customer traction grew, the need to focus more on analytics became business-critical.
“We needed to understand how customers were engaging with the product in order to make it a more compelling experience for them,” said Narasimhan. “We started thinking more about how to capture data and what metrics to focus on.”
Identifying Key Performance Indicators
The Memrise data team chose Fivetran to connect MySQL RDS databases in Amazon Web Services to Google BigQuery. The Memrise team is also able to ingest more transactional data sources with Fivetran, such as Stripe and a range of online ad services for marketing campaigns. At the other end of the process, Google Data Studio now provides data analysis and visualisations.
Event-related data from the core product database was considered particularly important. This data is the record of how people interact with each stage of an online course – when they start a new lesson, how the learning progresses, and what happens when they pass a test. The data team can now combine event data with core customer, product, and content data to see when the learning journey falters and identify parts of the course with higher-than-average drop-offs. Another important data source is the content database.
“We have a team of linguists and pedagogical specialists who use a content authoring system to design the courses. A lot of that is now driven by the data team who look at how various bits of content are performing and feed it back to the designers,” explained Narasimhan.
Delivering continuous application improvements
Having identified the KPIs (Key Performance Indicators) to measure, and with Fivetran keeping the product and content databases synced and up to date, the company is much more proactively able to respond to insights surfaced via BigQuery. Everything from course content to transactions has metrics attached to it, empowering the business with insights to drive continuous improvement and growth.
Memrise runs a ‘freemium’ business model and the ability to monitor conversion rates when people hit the paywall is business-critical.
“We use Fivetran to take data from our payment provider, Stripe, that tells us how many people are subscribing and the number of renewals,” said Narasimhan.
Fivetran connectors also help marketing by tapping into sources like Google and Apple Search Ads to measure campaign effectiveness.
“It helps us focus our activity and make sure the marketing spend goes towards areas that are performing. The corollary is that it helps us identify areas where we might not be getting as much bang for our buck,” he said.
Fivetran connectors to GitHub and Jira help the platform and product teams assess their own performance, while data from Google Play and iTunes shows what customers are actually doing with the app. Customer ratings and reviews are also tracked along with complaints, alerting the team to issues that need to be fixed.
Going forward, the plan is to use Fivetran to connect more data sources. Narasimhan praised the product, not just because it is easy to set up, but because it keeps the data flowing seamlessly in the background.
“Fivetran is critical to us,” he said. “We use it to copy across our product and content databases that are the cornerstone of all of the data analysis that we do.”