Exploring potential users by an email marketing plan with GitHub
I have been working for a startup company in a few months. Our product is a web site where people can write and read posts with their mind flow by a mindmap-like interface. The most of my job is to visualize log, helping ourselves to understand how users use our web site. Meanwhile, I also do some temporary tasks like adding a few seed data, or writing scripts for sending email to people in the early period.
For many startup companies, the marketing is the one of critical issues at the beginning. They want to develop potential users effectively. Fortunately, there was an inspired way that have been using every day on many shopping web sites. Recommendation systems, which deliver items that people may like. Our assumption is that if people like a GitHub repositoriy that is slightly similar to our product, so they might like our product as well. To practice this thought on our marketing plan, we made some preparations with GitHub. GitHub has more than a million repositories that people can do three actions such as, fork, watch, or like on them. Our plan is to look for the repositories that are similar to our web site, and send marketing emails to the people they interact with those repositories. In a nutshell, there are three main steps:
- Finding repositories that are slightly similar to our web site
- Sending marketing emails
- Understanding user behaviors
I am going to share those experince below, including observation and tools.
Finding repositories that are slightly similar to our web site
There is no fancy way to find repositories that are slightly similar to our web site, so we just pick repositories manually by associating with key concepts of web site (product). An example of ourselves, one of key concepts of web site is markdown that help people to write post more faster, and the format is human-readable, platform-independent. So we search related repositories from GitHub, then make a list for comparing and picking the worth repositories to send.
Sending marketing emails
As we made a list and decided what repositories' people are worth to send the marketing emals, two issues appear that how email addresses can be collected from repositories and how email can send to a bounch of people with their personal information like repo name, or personal ID of GitHub. Recipients they don’t want to get feelingless email content, personalized content can make email more approachable.
Fortunately, we can access GitHub by official API. So I made a tool github-email-explorer by Python to serve these two issues. The first issue can use a command with a repository name to fix, like this:
ge-explore --repo yuecen/github-email-explorer --action_type star fork watch
It allow you to use different action types for flexible arguments, the default setting type is all. All of email addresses can be collected and printed after the command finished.
We send email via the email service provider SendGrid that even provide template engine to render content. Nevertheless, github-email-explorer also render email content and with GitHub information. The reason of making the same function is not that strong, just I have alrady known how to use template engine Jinja2 to render a content, and I can use more structured variable name to express content. However, If I could rewrite again, I will use the template engine of SendGrid. A key reason is that developers, designers, or managers can discuss the same template when visiting the management page. Otherwise, we consider that the content format should use plain text but fancy css version that make your content like a junk email.
Understanding user behaviors
Brfore we send email to users, our backend have already connected with webhooks. The email progress log will be delivered to our Elasticsearch when an email is in processing on SendGrid. There are three indicators we care about:
- How many people open email?
- How many people click web site (product) link?
- How many people become users?
We can create a funnel from total email delivery to those three indicators sequentially, and observe which repositories have higher conversion rate for the next marketing plans. That is why I recommend you to make a list first.
Image:
Gibson, Charles Dana, The Social Ladder, 1902.