With more marketing tools like Facebook’s ad platform around, getting smartphone users to install your app is increasingly straightforward. A study recently found that the average American user has 25 apps on their smartphone. But figuring out how to convince users to keep returning to your app instead of letting it languish on their mobile is much harder. Y Combinator-backed Framed Data, which just launched out of closed beta and counts Twitch.TV as a client, wants to make it easy for developers to parse user behavior, even if they don’t have a data scientist on their team.
Framed’s founders say the startup is currently processing 500 million data points from its clients’ users each month, and is on track to double that number next month. The idea behind Framed is to provide simple-to-use machine learning products that help developers quickly figure out how to increase retention rates by pinpointing why users stay or leave.
CEO Thomson Nguyen and CTO Elliot Block met in 2006 while studying at U.C. Berkeley. After graduation, Nguyen went to grad school at the University of Cambridge, while Block joined Microsoft as a program manager, where he helped launch Office Web Apps. The two reunited when they both started working at online campaigning platform Causes, where Nguyen was a data scientist and Block a software engineer.
Understanding what makes users tick
“We are more opinionated. We will tell you who are the most valuable users, why they are leaving, and when they are about to leave,” Nguyen tells me.
The startup’s first target customers include consumer app and SaaS developers.
“To get this work done, you usually hire a large consultancy and do gigantic, enterprise custom work,” says Block. “We want to make it more accessible to smaller businesses by automating and opening it up to a larger market.”
Framed can use data from a wide range of sources, including Mixpanel, Segment.io, SQL, HDFS, Mongo, and CSV. Alternatively, you can also ask the team to help configure your user behavior data. Once you upload it to Framed, its machine-learning models will look at different pathways a user might take.
The platform splits users into different groups, such as time zone, operating system, or device, and looks at how likely they are to continue using the app or when they decide to abandon it. Its tipping point analysis tells developers what numbers they need to hit in order to start seeing an improvement in retention rates. Then it is up to product managers to launch A/B testing on certain features or do customer outreach, like sending promotions to top users.
Some of the things Framed have discovered seem pretty obvious. For example, it found that Apple product users who tinker with an app’s settings the first time they open it “fit a custom set of users that are tech-savvy folks,” says Nguyen.
But other insights are more surprising. One of Framed’s current clients is a photo app developer. The company discovered that users are most likely to stick around if at least seven friends from their social network join within a month, even if they have only shared a small number of photos.
Ultimately, Framed’s goal is to help developers understand users as people, instead of just numbers.
“A common error developers make, for example, is thinking that if 500 people use a chat feature but only 100 use a photo feature, then they should keep working on the chat feature,” says Block. “But it can turn out that chatting is not correlated with staying on month after month. It may turn out that the smaller group of people are actually more correlated with long-term retention.”
Bringing data science to smaller companies
Nguyen and Block were inspired to launch Framed after realizing that the data models they made to help companies improve user retention were cost-prohibitive for smaller developers.
“It’s super tough to hire a data scientist and a data engineer, and have the time to build those projects out,” says Nguyen. “We figured we could produce a platform that would find the same insights that we were finding at Causes.”
“When you have a lot of friends building stuff, you see that a lot of people don’t have an analysis platform,” adds Block. “They just don’t know about the tools that have been built up over the years to answer some of these basic questions.”
Though Framed is currently aimed at small businesses that don’t have their own data scientist, its founders say its services can also help larger companies. Framed is also building products for other verticals. It eventually plans to start selling analytic services to larger companies like Comcast, for example.
“Software has a long way to go before replacing the human data scientist. For those in the field, it can be a great tool that helps their work,” says Nguyen. “The secret of data scientists is that it’s a purely creative profession. If they are stuck coding models and adapters, engineering things, that is time taken away from creatively thinking about how to model data.”
[[Image source: PicJumbo]]