Kellogg PM week16: gentle introduction to data analytics
From previous week, we solely focused on high-level experimentation and how it can help drive growth to the company. For me the growth topic is fairly complicated especially for the statistical part behind it and how to make it count, not just running it. Professor Mahney walked through a simple process to create experiments and how to ask for help from stakeholders.
The first topic is the types of analytics you will encounter as a product manager. It’s good to know what kind of work you’re asking from your Data team either data analyst, data scientist, bi analyst, or you can group them altogether as data experts. I love the clarity of the beginning of the unit. We’ve gone through descriptive analytics or the way we tell the story of the past data via many sources including the beloved dashboard. Predictive analytics on the other hand rely on the power of guessing what the future could look like by implementing machine learning or artificial intelligence. Lastly, prescriptive analytics is the downstream of analytics, to suggest the solution, path of work, and to give business impact.
Correlation is not causation is a great topic of all time in my humble opinion. Our brain is engineered to catch trends and patterns, it’s in our nature to do this so we survive in the wood thousand years ago. However, not every time it is true. The way to derive causation in the right way is either run experimentation, control experiment, or trial in which we learned from the previous chapter. Correlation is useful to see if one interesting metric moves in the same direction as another or not. It’s useful information when combined with domain knowledge of the business to invalidate your observation against expectation.
You can’t mention correlation without mentioning human bias, we all have some kind of bias. It’s great to know types of them so we can avoid them, one good note to have is this bias is human bias when doing analytics. It’s not the bias-variance tradeoff within the modeling itself, if you happen to know some kind of modeling before. In my opinion, it’s not feasible to avoid every bias we have due to the enormous number of human biases but we can actively listen to the disagreement whenever it happens. Beginner mindset should be employed, ego should be reduced. Besides, you should ask for help from a data specialist within the company. It is usual that we don’t know everything if we are product managers, asking for help from data science or analytics team could help prevent such problems.
Google Analytics
Google Analytics is a super useful tool for web analytics aka web tracking, we can track user information and user behavior within y0ur site. It does not only provide you the solution to track but a bunch of dashboards so you don’t need to find another tool to build up a dashboard for tracking. It offers the way to split data in the fashion that you want to see including user origin, session, and funnel data.
For me it looks so overwhelming since there’s a lot of tabs and dashboards we can look at, if you don’t have anything in mind. You will be flooded with data. This tool is a great tool, if you know what you want to track and see. Apparently, even if its ui is not quite good, a lot of people still suggest using it. At the very low side, you don’t need to use its ui, you only use its feature then plug google analytics to another service provider to see a more beautiful or a cleaner dashboard.
