Performance Improvements

Waiting for something to work is frustrating. The more you wait, the more you get frustrated. The more frustrating the experience of using a product is the less you're going to use it.

That's why last week we focused on improving the performance of June. Last week we released a series of improvements that will make June faster and more reliable.

Up to 100x Faster audience computation

At the core of every June report, Slack notification or companies list there's an audience computation. An audience computation is the process of computing the list of people that match a certain set of filters. For example, the audience computation for the report "People that signed up in the last 30 days" is the process of computing the list of people that signed up in the last 30 days.

The previous way we were computing audiences was hitting some performance bottlenecks when the number of people in the audience was over 10.000 people.

We've now optimised the way we compute audiences and we've seen a 100x improvement in performance. The key to this improvement was spending time profiling the queries we were running and understanding where the bottlenecks were.

These improvements should also pave the way for nested audiences. We now have the data layer to support filters like:

(Pro users OR Users from Italy) AND (Users that signed up in the last 30 days)

Other improvements

• We've pushed some fixes to our activation cards on the home that sometimes showed numbers greater than 100%

• We fixed an intermittent bug that was causing some users on the power users card on the home to appear as "Unknown"

• We fixed some issues with the funnel report not reloading correctly when changing the date-range

• We fixed a bug that was causing every card on a report to re-render when changing some settings on a card

• Many other small bug fixes and improvements