Andreas Weigend, Social Data Revolution | MS&E 237, Stanford University, Spring 2011 | Course Wiki

DF5: Skout and OkCupid

Due: Tuesday May 10th, 2011 at noon (CEO and CTO of Skout will come to class)

To get some ideas on analysis on social data, spend half an hour to an hour reading some of the entries in . This is the blog of the online dating site OkCupid (that was acquired this year for USD 50M by To get some variety in what students in class read, please focus on entries posted on the month of your birthday. Say, if you were born in March, make sure you read the March posts. When you do the reading, create some analysis ideas and hypotheses in your mind that you are curious about.

In last year's MS&E237, a group of students worked with data from the mobile people discovery startup (the gay version is called If you were to write a blogpost for the skouts, what is the ONE IDEA you would you like to explore? (If of interest, you can actually do this with instructor approval for extra credit.) Please submit the 5 aspects of your idea to the google form by Tuesday noon (that day the CEO and or CTO of Skout/BoyAhoy will come to class):
  • P - what the question is and why it is interesting to you
  • H - what your hypothesis is for this question
  • A - what actions the company could take based on the insights you could derive
  • M - how you would measure the improvement in user experience
  • E - what experiment would you setup in order to test your hypothesis and how would you evaluate it?

To understand Skout / BoyAhoy a bit better, you could (optionally) download the app relevant to you and/or resister on the corresponding website. If you use the email address course emails get sent to you, you will have 4000 points to start with for free. Playing on the platform gives you a better sense of what potential metrics are actually available to you and helps you start formulate genuinely interesting questions and hypothesis.

Deliverable: Submit one suggestion for Skout or BoyAhoy on the following Google Form:

Notable Answers

P - What is the question and why is it interesting to you?
After playing around with Skout, both on the website and on the iPad app, I came across an interesting insight: there always seems to be a few faces that show up regardless of how you filtered the search. Moreover, there seem to be a very finite amount of 'faces' that show up when you are searching for matches nearby. This got me thinking from the other side of the lens.

How often does my profile show up on someone else's search and how often do they actually click on me? And more interestingly, what kind of searches/people find me?

I think this is a very interesting area to explore because on dating websites, you are essentially selling yourself or, at least, a version of yourself. But there is often no feedback as to how you are being perceived, what kind of demo you are reaching, etc. If Skout were to provide some sort of 'average user' that looks at your profile, it would be a way to provide feedback.

H - What is your hypotheses for this question?
My guess is that most of the people that find me would be similar in age group and ethnicity. In my case, that would be around 25 and Asian.

It goes deeper than just age and ethnicity though; there are a variety of variables that can explain how someone landed on my profile. While my guess is age and ethnicity play the biggest role, there are probably countless other things, including the profile picture, that could influence someone.

A - What actions could the company take based on the insights you've derived?
As mentioned before, Skout can create a page dedicated to the average user that is looking at my profile. Currently, it has specific users (for a price), but those are usually one time deals that cost a lot of money. Instead, providing a free glimpse of the average user reveals no identities, but gives the original user an idea of which demographic he or she is appealing to. Depending on the results of the page, the user can then adjust his or her profile to try to attract a better following. This helps Skout in two ways: it gives some insight as to what the user is looking for and Skout can thus create better matches, and Skout can learn about what qualities people are looking for in the general sense. I.e. Maybe all 25 year old girls are looking for tall guys?

M - How would you measure the improvement in user experience?
Improvements would be measured as an increase in matches.. So if you take a before and after, an improvement would be indicated by an increase in the frequency of not only matching, but chatting as well.

E - What experiment would you setup in order to test your hypothesis and how would you evaluate the results of this experiment?
To really figure out which variables are important or not, we can set up a simple test: have two groups, one control group (does not have access) and one group which allows for changes in the profile after learning about the average user they reach. After the adjustments are made, compare it the rate of matching of the two groups to determine whether there is statistical significance. If there is, then look at which factors were changed.

This test could be run iteratively, and/or can be isolated so that you only allow one attribute to be change or group only those people that have changed the same attributes. This is to determine which factors have significance when trying to match.