Startup metrics – the case of Toggl

The good thing about cloud applications is that you don’t have to be constantly guessing when making business development decisions. You have a lot of data about how users interact with your site. And I mean not only the standard Google Analytics kind of things, but also what happens after users sign up – what features they use, how fast pages load, etc.

There is a whole new movement called ‘Growth Hacking’ that utilizes techniques to measure user behavior inside web-apps. We firmly believe that you can’t manage what you don’t measure, so we have several key metrics that we constantly monitor. We actually want them right in our faces, so we created custom dashboards that are displayed on the office wall on three TV screens.

So what numbers are important to us?
TV 1 features the time of day in London, New York, San Francisco and Tokyo, and 6 charts:

  • Chart 1 is all about servers – it shows our database and application servers’ load for the last 24 hours.
  • Chart 2 shows the user and pro user growth curve for the last year.
  • Chart 3 displays the hourly number of new signups for the last week.
  • Chart 4 comes from Pingdom and shows the average response time of Toggl website for the last 24 hours.
  • Chart 5 gives us an overview of support queue size for the last week.
  • And Chart 6 compares the number of new pro users versus pro user churn.

TV 2 features basically the same data for our second product Planner.

TV 3 displays a real-time view from Google Analytics. It’s a really cool way to see how the day progresses from Asia to Europe to America, as it displays traffic on a world map – growing traffic in green and fading traffic in gray.

The technical solution is based on new smart TVs that feature built-in browsers. Initially we tried to use Geckoboard, but it kept crashing with the TVs. It’s probably more of an issue with the limited TV browsers than with Geckoboard itself. So we built a quick and dirty custom dashboard that shows charts on a static HTML, reloading the page every minute. Very robust and very stable. The third TV is connected to an iMac mini, because no smart TV was able to display the Google Analytics real-time view for more than a few minutes without crashing.

The charts are created with Munin. It’s in no way a fancy technology these days, but it works. The information density is very good, it loads fast, and does not require too much capability from the smart TV to display it.

There is a lot of data that did not make it to the dashboards. It’s really tricky to choose what is important and what’s not. What is your experience? What else should we measure?

By On October 29, 2012

  1. You could measure how many devices your users access Toggl from, since the portability from one device to another is a key part of Toggl’s USP. Also, ratings, links, shares, and reviews not only by device, but also by platform. For example, is there a big delta between user reviews on iOS vs Android? These numbers would change more slowly than server load, but might help the team appreciate strengths and spot gaps.

  2. If I were in mission control like you are, one of the key metrics I’d be interested in is the rate of help requests. A spike could be an early indication of a developing trouble.

    You guys are great. Keep it up!

  3. Speaking from personal “life hacking” experience I would expect quite a few people to sign on for Toggle and then forget about it. Either immediately or after a short time.
    The people from Evernote tend to measure both members and usage. The latter I think is a valuable measurement. When evaluating Toggl I would really like to know the trend in the number of hours logged in the tool. There’s no better way to perceive usage, I think.