Satya Patel on using the Scientific Method for Growth

Satya Patel (his Twitter) used to work for both Google and Twitter, and is now working as a partner at Homebrew venture capital.

Lessons learned
  • Companies tend to "throw spaghetti at the wall" when they grow their company - they try everything between heaven and earth, and it ends up in a mess. Satya Patel argues that it's better to use a scientific approach. 
  • No companies fail for lack of ambition - they fail for lack of focus. By using a scientific approach to ask the questions you should answer, it can help you to focus.
  • You should be learning at every step. Begin with small steps, and as you learn you can increase the steps. Crawl - walk - run. 
  • The scientific method:
  1. Purpose. What problem are you trying to solve? Twitter asked themselves 2 questions: "How do you increase active user growth?" "How do you increase revenue?" Twitter focused on those questions, so everyone in the company had to orient their product ideas around these 2 objectives.
  2. Research. Can help you to clarify the question you are trying to answer and find new answers. The goal is to conform the hypothesis so you can have a focused experiment. Talk to your customers, team, and understand your competitors. You can also find surprising answers. By looking at data, Twitter thought the users liked to click on a button. But when they talked to the users, they realized that the users had misunderstood the button.   
  3. Hypothesis. What do you believe to be true? Make sure it's measurable. AdSense's hypothesis was that they thought that publishers had to put ads on the top of the page, but AdSense realized that ads on the bottom of the pager were much better. People read the text and clicked on the ads. Twitter tried multiple hypothesis to improve their landing page: "Change the text," "Remove the search function." 
  4. Experiment. Focus on the variables that matter. Twitter could have tested a million variables to optimize the landing page, but they focused on a few (maybe 10?) main variables.  
  5. Analysis. Use enough data to make sure the result is statistically significant. But if you are new, you don't have enough data. Then take the data you have and let your intuition guide you. Remember that data is not the end since data can lie. Look at the data and use your own judgment to find the right approach.    
  6. Conclusion. From their landing page, among others, Twitter removed the search because users searched instead of signing up, and the pictures of signed up users because it increased loading time.   
  • Don't forget to write down your conclusions and share what you've learned. 
  • Reinforce what you've learned. Lather - rinse - repeat!
  • To grow, companies are using Google Analytics, Optimizely, KISSmetrics.
  • There were two good way to grow Twitter: address book import and translation of the site. Twitter also realized that growth increased if they customized the site depending on what country the user came from.  
  • Even small changes matter as long as they are statistically significant and not expensive to implement. A small change each month will grow into a large change over several months.