Affinity Profile and How It Works
Created: 27/02/2023
Updated: 07/03/2023
Author: Polina A.
General Information
The affinity profile of each individual user is compiled based on:
Events
Affinity attributes of products
We collect all of the user’s interactions with products and calculate the user’s affinity towards selected properties of a product (e.g. colour, brand, style, etc.). The affinity score depends on the amount of interactions and the user’s intent, for example, a user viewing a product has a lower value than the user adding the product to their cart, which in turn has a lower value than the user purchasing the product. Additionally, the recency of the interaction is taken into consideration.
Affinity is collected based on the following events:
Product View
Add to Cart
Purchase
Product Feed and Attributes:
By default, the Affinity Score is calculated based on the 'categories' field of the product feed (the entire category tree). Other attributes for calculating the Affinity Score can be added through the team working on your project.
To calculate the Affinity Score for additional product properties, it is necessary to define the list of required product feed columns (no more than 10 affinity columns in the feed, optimally — 5).
Events and Evaluation
Each user's interaction with the website or application signals about their interests, preferences, and intents. These interactions include adding a product to the cart, filtering a category, viewing a product, and making a purchase. Properties of these products, such as colour, brand, style, and others, inform the system about the user's preferences. By combining the type of interaction and product attributes, as well as the recency of a specific interaction, the platform can identify user preferences both over a period of time and in real-time.
Affinity Scores:
Viewing a product = 2 points
Adding a product to the cart = 4 points
Purchase = 6 points
Filtering products = 2 points
How is the product score calculated based on the type of action?
The score is calculated as the sum of multiplications, according to the formula: "(number of product views X coefficient X action weight) + (number of the cart additions X coefficient X action weight) + (number of purchases X coefficient X action weight)". For example, in case of the colour “green”, its final score will be: (4*2*1)+(1*4*20)+(1*6*60) = 8 + 80 + 360 = 448
Viewing a product
1
1*2
1*2*4
1*1*2
Adding to cart
20
20*4
20*4*1
20*4*0
Purchase
60
60*6
60*6*1
60*6*0
Total Score:
448
2
How is the product score calculated based on the interaction timeframe?
In this case, the calculation takes into account the action weight, the number of interactions, and the coefficient for the recency of the interaction.
Coefficients are distributed as follows:
Up to 2 days = 8
Up to 30 days = 2
Up to 180 days = 1
Let's assume the user visited the site twice: 15 days ago and today, and performed a series of actions.
Calculation formula: "coefficient of age[1] X (action weight X number of actions) + coefficient of age[2] X (action weight X number of actions)"
Previous session (15 days), coefficient: 2
Viewing a product
2
4
2
Adding to cart
4
1
0
Score
2*(2*4)+2*(4*1)=24
2*(2*2) = 8
New session (up to 2 days), coefficient: 8
Purchase
6
1
0
Score
8*(1*6) = 48
0
Total Score:
72 (sum of two scores)
8
Where is the Affinity Profile Used?
As one of the conditions for creating audiences based on affinity towards a particular product features;
In the affinity recommendation algorithm;
In personalization cases that take into account the user's preferences for specific product attributes.
Clarifications and Limitations
Affinity is based on your product feed. Only attributes with discrete values (a limited list) can be used to build affinity.
Data on affinity for each user is collected from the moment the user starts taking action on your platform.
Data retention — 180 days.
How to Prepare the Feed for Affinity Work?
It is necessary to determine the product properties to be used for affinity calculations. These could include attributes like colour, size, brand of the product, and any other characteristics relevant to your business.
There are two major requirements for field content:
a. Data uniformity: Values such as "White," "white," and "01 white" are considered three completely different colours by the system. Therefore, it is essential to standardize values to a single, consistent format.
b. Information availability for products: It is best practice to use affinity attributes that have information available for a large number of products in your feed.
Last updated
Was this helpful?