41. What is Feature Engineering?
Estimated Reading Time: 1 minute
Data, data everywhere, but no insight to glean.
This modified Rime of the Ancient Mariner is the lament of every marketing analyst.
One of the main reasons is data silos - data residing in the “walls” of the application where its generated.
A problem that can be solved by feature engineering.
Feature engineering is a fancy data science term for transforming raw data into more meaningful features. It can be as simple as adding a new column derived from existing data.
These new ‘features’ help you create a more in-depth analysis. Some typical examples are:
Calculating recency and frequency of transactions
Extracting day and time from timestamps
Deriving sentiment scores from customer reviews
Most data science models and analyses involve some form of feature engineering to add more meaning and improve model performance.
Joining data from different sources to create new features helps you break down data. For example, making a “customer lifetime value” feature allows marketing, sales, and finance to improve decision-making.
One of the most potent applications of feature engineering in marketing is segmentation. It can help us go beyond simple demographics or behavior. One can create segments based on customized, advanced features such as:
Engagement score based on community membership and app interactions
Cross-sell potential based on type of purchase and browsing behavior
Repeat Purchase Likelihood based on average purchase interval and product reviews
Brand sentiment index based on customer reviews, social media posts, and survey feedback
I am sure you can imagine more - the potential is limitless!
That’s it for this week. See you next week.