Running a data-driven social software site

SlideShare began with an idea. We built it on instinct, launched it. People liked it, it grew. Our active users wrote to us, blogged, sent us feedback about what they liked or did not like. We took that into account as we planned features.

Its great to listen to active users, but it can bias you towards the superuser. As our userbase is growing, we need to take into account the different user segments, the people who don’t blog and write to us.

There is only way to keep in mind all the user types, to end the endless debates within the company about what we need to do next. And that is to be driven by data. Period.


For the past few months, I have become increasingly focused on data. Every feature we build, we think of metrics we will track. Once we launch, we watch those metrics change, decide how to refine, what to add, what to take away (yes, we regularly take away stuff that is not working out).

How we track metrics
Server logs: This gives us broad about overall activity, but its often too broad to be actionable.

Google Analytics: Its easy to setup Google Analytics out of the box, but you can get more sophisticated and track your users on a more granular level. Currently, our Google analytics is very basic. In the next week or more, we will be making adjustments to become more sophisticated in our tracking and track some metrics that we care about.

SlideShadow: Its our internal shadow app, a dashboard that tracks a bunch of metrics on a hourly basis. As we add more features, the shadow app is increasingly getting more complex. This is the primary way we monitor what’s happening on SlideShare.

Feedback emails / Blog posts: While each feedback mail is just one data point, we try to pay attention to trends.

Database reports: The final, most flexible method is to simply generate the reports that you need, when you need them. Its too much to put every bit of data you need into SlideShadow. So if we need data fast, we write a custom sql query and save it in a mysql client that is pointed at a slave database. Reports written this way that turn out to be useful are migrated to SlideShadow.

With metrics, I have realized that the mindset is important. When we think of new features or refinements, I try to think with a metrics mindset. Why are we building this? What metrics are we trying to impact. There are four or five core metrics that a website like ours cares about, and pretty much every change relates to one or more of this. This is also helpful in working with developers. They are aware of what metrics they need to impact, and the success or failure of a project is very clear – did the metric move in the right direction.

As we grow (both the site and the team), I find that metrics is good way to manage growth. It helps me stay focused!

What metrics do you use? How do you use them? What helps you stay metrics-driven?

2 thoughts on “Running a data-driven social software site

  1. Hi Rashmi,

    At Dogster and Catster we have a different userbase and value proposition, but I very much agree it’s quite risky to listen to the most vocal users. I followed their demands early on assuming they represented the general users, but I was wrong, they turned out to simply be the most vocal (and needy).

    What we do now is listen to and chart everyone’s requests, and then, when we are preparing a 6 moth road map, we survey the general user base on which features they want next. We include the most requested features, the most clever requests, ideas of advisers and trusted peers, and our own best ideas. We ask them to prioritize the features on the assumption only 3 may get done in the near future. It’s usually quite different than what we get the most emails for or blog-post business advice.

    Yet even then we gut check the results to make sure they make sense.

    However, purely going by data seems to have it’s own pit falls, as numbers don’t always tell the whole picture. I reckon you’ve got to feel your customers as much as you crunch the numbers ;>

  2. Ted,

    I like the way that you are doing the prioritizing. Maybe I will try MindCanvas DivideTheDollar for doing that!

    And very true about going by a qualitative feel of the quantitative trends. I feel as if running a social site, one is almost attached to the users. They are always close to you.

Comments are closed.