Another month, another Hack Day at GoSquared. This time round saw some collaboration between members of the team, as well as the ever-prevalent super-super-secret projects that we can’t talk about right yet.
Actually shipping previous Hack Day work
In a radical departure from what we usually do on Hack Day, JT decided to take some previous work and actually ship it into production. Following on from last time’s work on Clarence, we’re now using it to provision a number of our EC2 instances. The game plan is to continue gradually replacing our Chef setup with Clarence, which will allow us a much faster turn-around time for new instances and Auto Scaling to meet demand.
More secret stuff
Ben worked on ████ █████████████ for the ████ ███ ███. Wish we could say more.
Getting more Analytical with some of our data
Geoff focused on adding new API endpoints to generate analytical data about Smart Groups in People Analytics. The goal was to process statistical information about People in Smart Groups so that businesses can generate reports and understand segments of their user base in more detail.
First off Geoff worked on a new People API endpoint to provide aggregated data about People in Smart Groups. The aggregations work by processing data contained in a property on those users. For numeric properties, it can compute the avg/sum/min/max functions, and generate intervals for a histogram. For string properties, it can find the most common terms across the people in the smart group and order them by frequency. It can also estimate the number of unique terms (HyperLogLog cardinality).
Once that was working, Geoff also created another API endpoint to retrieve time series data counting how many people are in a given Smart Group at regular points in time. The granularity goes right down to 5 minute intervals and can span any given time period.
To demonstrate how this data can be used, Geoff worked in collaboration with Hugh to visualise some example aggregations on a dashboard. Hugh created a gauge chart to show the min/max/avg MRR of a smart group containing a segment of paying customers. He also added time series graphs to plot the number of users in a Smart Group over the period of a few months.
Altogether the project was an exciting chance to experiment with statistical data about Smart Groups. You might see official endpoints make their way into the API soon for you use.