Reverse engineering and the scientific method are two useful methods to achieve private and professional goals, especially when used in combination with each other. In this post, I’m going to apply this idea to the business goal of Medium.com.
The first step is to define the objective. I’m going to define the objective of Medium.com is to maximize the revenue from premium memberships. This might or might not be their objective. It’s just an assumption I make to demonstrate the idea.
Now, we are going to start from that objective and come backwards step by step. Let’s formulate our challenge as a question and look for answers to the question. Remember, asking the right questions is the first step to success.
What is needed to maximize the revenue from premium memberships?
- Maximize new subscriptions.
- Maximize the retention of existing subscribers.
At this moment, we have two branches that we have to work on. Let’s start with the first branch.
What is needed to maximize the new subscriptions?
- We have to maximize the traffic of Medium.com.
- We have to maximize the user engagement.
- We have to create curiosity about the premium content among the existing users.
- We need an optimal landing page design of the premium content for maximum conversions.
Now, we have four branches for the first branch. Again, we need to work on each branch separately. Let’s work on the second branch.
How can we maximize the user engagement?
- Feature quality content on the homepage.
- Feature quality content at the end of the posts.
- Feature content that keeps the user in Medium.com.
- Optimize the page design to keep the user in Medium.com.
Let’s continue with the first branch.
How can we feature quality content on the homepage?
- Hire editors.
- Feature the most popular posts on the homepage.
- Create a relevance score between each user and each post. Feature the most relevant posts to each user.
At this moment, I need to test which option would produce the best results. I will divide the user base into three groups and test each hypothesis.
The Metric for User Engagement
The first step to that experiment is to define how I’m going to measure the engagement. In order to do that, I’m going to measure four variables: views, reads, fans, and responses. Before composing my metric, I’ll look into the data to see the number of reads, fans, and responses for each view.
Let’s say, the average user reads the 50%, claps for the 25%, and responds to 10% of the posts they view. In that case, I’ll construct the engagement metric as (#Views + 2 * #Reads + 4 * #Fans + 10 * #Responses). With those weights, I balance each variable based on their importance.
I’ll measure the engagement metric for each group above and see if one group significantly outperforms the other.
A Relevance Score between each User and Post
Featuring the most popular posts on the homepage is a straightforward method. Hiring editors and coming up with a relevance score is rather complicated. Let’s work on the relevance score in this post.
Which variables can we use to create a relevance score?
- Authors of the posts that the user had a high engagement score before.
- High engagement scores by other users that follow a similar set of tags as the user.
- High engagement scores by other users that had high engagement scores on the posts where the user had a high engagement score.
To start with, I could use the same weight for each variable, but eventually, this could be optimized as well.
Processing the Results
Now, all I have to do is to run this experiment for a month and look at the results. Are there significant differences between the results? My hunch is that the second option, featuring the most popular posts on the homepage, would perform the worst.
I think the third option, having a relevance score, would perform the best. Once we find the best performing option, we can continue our experiments.
Once I find the best performing option, I’d serve the best performing option to 90% of the users. I’d try to come up with another option each month and serve the choices of that option to the remaining 10% of the users.
Here, you have to use your creativity to come up with other options. Remember, creativity is finding new relationships in already existing concepts. Then, all you have to do is to test those new options on the 10% of your user base.
Here’s an example. I’d come up with a relevance score between the editors and users. I’d then use that signal in the relevance score between the posts and users. If an editor and a user has a high relevance score, the posts that are selected by that editor will have a high relevance score to that user.
The Big Picture
At this moment, we have a method to create high user engagement, but don’t forget that user engagement is one of the many branches in our tree. There are many other branches we have to work on. So, we need to apply the same reasoning and experiments to the other branches of the tree as well.
We can use the combination of reverse engineering and the scientific method to achieve our private and professional goals. The first step is define a clear objective.
Once we have a clear objective, we will use the questioning process until we have several options that we can act upon immediately. Once we have those options, we can use the scientific method to find the most effective option.
Finding the most effective option is not the end of the work, because we can use our creativity to come up with other options and test them against the best option that we came up with so far.
However, we shouldn’t lose the sight of the big picture and get lost in the details. Usually a business goal can be divided in a tree that consists of multiple branches. We need to optimize each branch of the tree to reach our goal.