@andrewchen

Get the newsletter · 2018 essays (PDF) · Featured · Recent

How do you apply math to Web 2.0 websites?

Quantitative approaches to social media
Two great blogs recently on borrowing concepts from the pay-per-click optimization world and applying them to social media sites: How about landing pages for the social media visitor? and also Optimizing social media landing pages.

When I refer to these approaches, I generally refer to a broad set of optimization techniques that advertisers use to make sure the traffic they are paying translates into sales. One example of this is Google’s free Website Optimizer tool, which you can use to optimize things like layout, headlines, copy, form placement, etc.

Multivariate regression testing as an example
For example, you could imagine having an invite page for visitors to virally add their friends. Which headline is the best?

  • Don’t be lonely, add your friends!
  • Example your network! Invite friends!
  • You have 10 free invitations for friends
  • Import your addressbook in 30 seconds

It’s hard to tell. And then when you add in 5 different pictures, 5 different explanations of the site, and 5 different color schemes, all of a sudden you have hundreds of different combinations. Well, turns out through multivariate testing you can end up testing just a couple variations and then it’ll automatically predict (and test) the best combination of the hundreds.

People in the paid advertising space use these techniques because when you’re paying $1 per click (or a $1000 CPM), you try to optimize every opportunity.

Multivariate testing example
Check out this solid example of multivariate testing:

http://dkm.com/

You’ll see slight different content, with lots of different combinations of other things.

How to apply hard metrics to social media
Ultimately, it seems like what you want to do is to pick a couple key "transactional" moments that you want to optimize. That could be things like:

  • Page where the user creates media
  • Page where the user votes or comments
  • Page where the user invites their friends
  • Page where the user fills out their profile
  • Page where the user shares out their widget

That is, a bunch of pages where the users aren’t doing the core mechanic of browsing from peer item to peer item. (Like video to video, or person to person). In these moments where people are about to invite friends, you want to figure out the exact right wording to get the person to spread the site virally.

These are the pages where you are most likely to apply these page optimization techniques.

Viral growth dashboard

Also, in a previous blog I had written about 10 different strategies to acquire users. In these cases, you want to build out pipelines and track the metrics around how people are progressing down the goals you want them to accomplish. Here are some possible metrics you’d want to measure:

  • Sources of traffic
  • Landing page views
  • % of users that register
  • % of users that send out invites
  • # of invites sent out, per user on average
  • % of invites delivered successfully
  • % of invites read by users
  • # of virally added users, per user on average

Or, alternatively, for widgets:

  • # of widgets outside
  • # of widget impressions
  • % CTR of widgets
  • Signup conversion rate
  • [other viral metrics]

You might also want to track overall growth rate, the viral coefficient, etc.

If you don’t watch these numbers, they’re unlikely to grow…

PS. Get new updates/analysis on tech and startups

I write a high-quality, weekly newsletter covering what's happening in Silicon Valley, focused on startups, marketing, and mobile.