Applying “Zapier Intelligence”
Chris Peters, (@statwonk) / Data Scientist Zapier
As the sole data scientist at a fast growing start-up, Chris Peters (@statwonk) knew agility would be crucial to keeping his team informed. Two years ago, when Chris thought of business intelligence dashboards, he envisioned a highly polished web app. Chris helped build that type of app … and nobody used it. The reason was simple: By trying to do everything for everyone, the app didn’t truly succeed for anyone.
"Finally, I get all the web designer and developer’s tools in the language I already use, R — what’s not to love?”
- Chris Peters, Data Scientist at Zapier
“A business intelligence dashboard is not your standard web app,” says Chris. “Ideally, you’re able to expose massive queries as results to employees instantly. The reality is that if you have to rely on a database alone, expect to spend big-time on your database setup. A Shiny dashboard built with R gives you a powerful in-memory tool for processing data. I often do three-quarters of a Shiny-driven query as a request sent to the database, return the data to R, and then finish minor cleaning and aggregation — all within a few seconds before showing output to the user.”
Using Shiny, Chris can expose a simple database query as a sortable table or easy-to-read graph for a team member within minutes. ‘Throwaway’ graphs become everyday views in an emergent style of development that allows him to ship quickly and iterate.
Nearly every movement through the Zapier site is analyzed to improve user experience.
“Shiny gives the data scientist the abilities of a web developer and web designer, combined, without having to know HTML or Javascript,” says Chris. Shiny also lets him use the Twitter Bootstrap CSS framework. This is especially powerful in Chris’ view because it allows the data scientist to do their own design work — negating the need for a separate designer.
Shiny and Shiny Server Pro have proven highly scalable for Zapier. Even working with high volume data, Chris does not see a bottleneck.
“The developer has full access to the R console, so it’s trivial to connect to as many databases as necessary,”says Chris. “With Hadley Wickham’s dyplr package, connections are managed automatically. Since Zapier runs Shiny Server Pro on an Amazon EC2 instance, scaling to a larger server takes just a couple of mouse clicks.”
Zapier built its own multivariate testing tool with Shiny. Nearly every movement on the site is analyzed to improve user experience and understand where users are having trouble. Those results are aggregated and displayed in a rich Shiny app running on Shiny Server Pro.
Shiny and Shiny Server Pro have allowed Zapier to build an agile set of dashboards tailored to their users. The team branded it, “Zapier Intelligence”. For Chris, being able to iterate quickly, scale to high volume data, and contribute directly to business value comes back to Shiny.
“Finally, I get all the web designer and developer tools in the language I already use, R — what’s not to love?”
Zapier lets you easily connect the web apps you use, making it easy to automate tedious tasks. Zapier acts as a hub to integrate over 350 web services like Mailchimp, Salesforce, Github and Trello.
RStudio provides open source and enterprise-ready professional software for data science teams using R and Python.
