Ever wanted to transform rows of raw data into actionable insights that shape your business's future? Enter the world of predictive analytics, where your transactional table is the golden key.
This isn't just any table—it's the bridge between your business's past and its unlimited potential. Dive in, and let's demystify the magic behind it!
What is a transactional table?
When building a predictive flow, the first step is to retrieve your raw activity table (typically a table of transactions). The table should represent historical activity, such as all the historical transactions of all customers, or all the purchases of all your products.
Why is this table critical for building a predictive model?
This table of unfiltered, complete data will serve as the basis for your model to identify and understand patterns in past activity. The model will use this data and those patterns when it generates predictions about the future.
What should my transactional table look like?
Your predictive model needs a “signature” of the past activity of each customer or product. That unique signature is made up of three critical elements that must be contained in your table:
A column that represents the object you want to predict on, for example, “customer ID” or “product identifier.”
A column that represents the date of the activity, like “purchase date” or “event date.”
Columns of additional data that characterize the activity and/or the object. These could include the number of products in a purchase, the cost, the type of payment, the profile of the customer (age, gender, etc.), and other kinds of information.
Let’s look at an example:
The column ‘customer_id’ is the individual, unique ID of a customer. This is an important identifier column when we want to predict customer behavior.
The column ‘registration_date’ is the date column that we also need.
Additionally, the table includes 6 columns of additional data that we can use to characterize the activity or customer.
For your transactional table, remember that there’s no need to change the names of the columns or manipulate the file in any way. You just need to ensure the data is there, and Pecan will take care of the rest.
Let’s see another example:
The column “Order_id” is the individual, unique ID of a transaction. This is an important identifier column when we need to predict a transaction outcome.
The column “Date‘’ is the date column that we also need to have.
This table also includes 7 columns of further data that characterize the activity or customer.
Having trouble uploading your CSV file? We're here to help! Check out this article to ensure your file meets Pecan's requirements.