The birth of Data debt and looking at data as a product
As a PM you learn always to try to test your hypothesis. To do that you have a set of tools under your belt that require you to always test, always try to lean and if you’re happy with the result it means that you didn’t do a good job.
But in some cases, such as data, I always hear phrases such as:
- I need to track all the events as maybe I’ll want to..
- Tracking everything gives me the ability to see things if they go wrong..
- I don’t want to waste dev time on triggering events
TL;DR
You should invest in the quality of the data and not the quantity. A proper tracking plan and data governance process will ensure that as a product manager you will get all the insights that you need when you need it. Nothing more and nothing less
So why is it that we look at data as a task that should not get a priority? Why don’t we have “Data debt” like we have Tech debt that requires us to build and rebuild.
Now.. don’t get me wrong… As a PM the last thing I need is to go all in blind but there are so many mitigations to achieve that.
So many times that I went into an org as a freelance or an employee and could not build even the most basic funnel even with the tools I use on a daily basis. Cause how do you know if you need to use “Added to cart” or “Add to cart” especially when both show data and different funnel results?
How can you understand what the users did and how they behaved if you don’t have any properties attached to those “auto tracked” events? So great! I can see I have a problem with my purchase funnel. But in what category? Did the user have a previous purchase or is her first touch with our product?
To be able to add those additional layers of data I need development anyways. So where is the lean now? Why do things twice?
Want to lead change? Let’s define our data as a product and treat it the same. No more track everything. Start lean. Understand gaps and iterate accordingly.
A few best practices that I saw over the years as a data oriented PM and analysis:
- The golden rule. Tracking plan is everything! How to aggregate the data, what event properties I need to send and what user properties I want to add to enrich my data in a proper way.
- Create 2–3 generic events that will include enough data (properties) that you can understand the top funnel with some leverage with the “free” properties you get. This will ensure that you don’t “miss” those important touch points but you’re not making a mess.
- Know Amazon start from the PR? (The PMs start by writing the press release and work back from there) Do the same with your data. Start with plotting your graphs, charts and tables then I promise you’ll see what you need to add and to what event to get that answer
- Make sure you have a data governance process. Clean your unused events, Invest in your data, change your data stack. Again.. It’s a product.
One last thing just to help you understand even further on how important it is to define the events before hand.
I have two events in e-commerce that make most of my decisions the hardest.
- Category page view (SERP - Search result page)
- Product page view
Now.. Let’s see what data I could add to each one to understand my user behavior:
Number of items in the category page, Number of pages, Page load speed, Product names, Product labels, Product prices, Product images.. Right?
But now let’s try to ask.. what do I want to really understand? What is my product questions.
“What impacts the user to select the product he clicked?”
So basically having those properties on the Category page don’t help me as much as if I send them in the product page view. That way my reports will be able to show the product page based on the number of impacting properties.
Makes sense? It’s a tough one to swallow :)
I’m always on the path of helping more companies and product managers get more out of their data. I’ll create a tracking plan best practices article soon but till then feel free to ping me
Please like and subscribe on Medium this motivates me a lot!