The different types of data
Item data means all the data that is related to the given item (product, digital content) Pure item data can also be distinguished when the data has no involvement with any user impact. Let’s have an example: Item belongs to the “female” tag and has the price category of “7” - all the elements are created by the content provider (the e-store)
Item Data can have references to the user who interacts with the item (buys, watches, visits etc.)
Let’s have another example: Item belongs to the “female” tag and has the price category of “7” and has been viewed 4 times by user X. This is still an item data, but differs from the category of pure item data because it also bears the user’s impact.
Item data includes specific characteristics (color, size, price, technical details, length of warranty, item location etc. ) that are related to the product in order to identify the product and to sort them to similarity groups.
In the craft of personalization and product recommendation item data will be important in the methods using content based filtering algorithms.
User data includes all the data that is related to the given user.
Demographic User Data
The basic demographic data is ideal for e-commerce basic user segmentation:
Stores often collect A/S/L (age, sex, location, name, postcode etc.) data from subscribers on signing up or completing a purchase. This type of data is very static, doesn’t change very often ( gender, name, address) or the change can be calculated (age) or can be detected easily (location, browser info, etc.- if the user is willing to share)
Preference data can be captured on an active way when the users fill in forms and tick boxes. It is important to separate data based on assumptions from the data the user provides actively:
-If the customer expresses her attraction to red products that means an explicit preference data. If a customer likes visiting product pages displaying red products, it’s also a preference data, but at this stage we only have the assumption of this color preference. (which might also be the trigger of displaying recommendations to her, but these predictions are based on assumptions)
During the signup process to a VIP- , loyalty, or discount program it is essential to ask for something in turn: this is the typical occasion when we can interview the user about their preferred goods, brands and makes, colors, sizes or what kind of communication channel they prefer (e-mail, messenger or sms?)
This kind of data can get outdated very quickly: if your user is interested in premium bicycle helmets in the summer, it doesn’t mean that he is interested in the rest of the year and you have to recommend premium helmets to him for 3 years. This implies that you have to make your customers update their preferences from time to time, because they will never do it by themselves.
Recency, Frequency, Monetization - Transactional or RFM data
Transactional data includes all the user actions implying purchases. This data helps you in identifying your solvent customers, or those who hasn’t bought anything yet.
The most important categories of transactional data are:
-Average order value,
-First and last purchase date,
-Total amount spent/ total cart abandonment value,
-Number of purchases in the last year/ last month etc.,
-Products purchased most recently, products purchased with the highest frequency, products whose sales has risen/fallen recently etc.