Why is it important?
There’s a popular adage in data science that goes: “garbage in, garbage out”. It may sound strange, but nearly all data scientists have heard this at one point. So, what does it mean?
On the most basic level, it means that to make good decisions based on data, the data itself needs to be correct. With this in mind, the longer version of the expression would be “if the data itself is corrupt, no meaning can be extracted from it”. At its crux, the Advertising insights application is supposed to help users make better digital marketing decisions, which would result in increased ROI, which is itself related to purchase data.
This test is designed to check that both Facebook and Google are measuring purchase data, so that proper insights can be given for how to improve performance.
What if the test fails?
For the purposes of this test, Facebook and Google are treated somewhat differently. For Google, we check to see if only one conversion is being tracked in Google, and if that conversion is counting the number of purchases. There are many different types of conversions that can be tracked on Google, but for ecommerce shops such as yours, we find that it’s best to focus only on purchases, as this is the most important type of conversion for all shops.
On Facebook, the test is more complicated and is related to the funnel analysis feature, which analyzes multiple events tracked by the Facebook pixel:
- landing page views
- add to carts
- initiate checkout
These events (in that order) represent the typical steps taken by a website visitor that ends up purchasing something on the website.
Each step in this journey is a potential breaking-off point for the user, which causes these metrics to decline: there will be less add to carts than landing page views, and so on (hence the name funnel).
So in order to pass this test, we make sure that all of these events are being tracked, and that they are in descending order. It’s not uncommon for errors in setup to appear and mess up the data, which is why we’re checking to see if the metrics are in decreasing order.