How are algorithms used in marketing?

[hmonower921]

Customer behavior analysis allows us to notice certain patterns that occur. Given that we have a huge amount of data on traffic sources, time, behavior, pages visited, and many others, we are able to define the patterns that are desirable to us and those that we do not want.

The algorithms used in MarTech tools are very good at detecting these patterns and, based on the acquired knowledge, they influence certain behavioral scenarios so that those that are positive for us are realized.

Powered by this knowledge, the algorithm can personalize our website’s communication based on each visitor’s most likely motivation. Detecting intent allows us to serve information and products much better.

Somewhat similar, but more complicated, are the predictive algorithms used to make recommendations on sites like Amazon and Netflix. Analysts estimate that 35% of what people buy on Amazon and 75% of what they watch on Netflix is ​​driven by these algorithms.

These algorithms also work by analyzing both past behavior (e.g. what you bought or watched) and the behavior of others (e.g. what people who bought or watched the same thing bought ghana whatsapp data or watched). The key to the success of these algorithms is the breadth of data available. By analyzing the past behavior of similar consumers, these algorithms are able to make recommendations that are more likely to happen.

Group filtering
Collaborative filtering (CF) and its modifications are one of the most commonly used recommendation algorithms. Even novice programmers can use it to build their own movie, product, or ad recommendation system .

The algorithm uses association rule learning principles .

When we want to recommend something to a user, the most logical thing to do is to find people with similar interests, analyze their behavior, and recommend products that similar users prefer. On the other hand, we can look at items similar to those the user has bought before and recommend similar ones to them.

These are two basic approaches. CF or user-based collaborative filtering and item-based collaborative filtering in this case products. In both cases, this recommendation algorithm has two steps:

Find out how many users/items in your database are similar to a given user/item.
Rate other users/items to predict what value you would give to a user of this product, taking into account the total weight of users/items that are more similar to it.
Unfortunately, these types of algorithms have a flaw. It's about the so-called information bubbles. Relying solely on recommendations, we will lock ourselves in these information bubbles. We will never try anything new, anything that does not fit the patterns assigned to us. Personally, for me, this is a big flaw.

What does “most similar” mean in this algorithm?
We only have a vector of preferences for each user (a row of matrix R) and a vector of user ratings for each product (the columns of matrix R).

User Batman Star Wars Titanic
Olek 3 3 -
Charles - 2 4
Ola - 5 -
Next steps:
First, let's leave only the elements for which we know values ​​in both vectors.

For example, if we want to compare Olek and Karol, we can mention that Olek has not watched Titanic and Karol has not watched Batman until this moment, so we can measure their similarity only in Star Wars.

The most popular techniques for measuring similarity are cosine similarity or correlations between user/item vectors. The last step is to take a weighted arithmetic mean according to the degree of similarity, this will allow us to fill in empty cells in the table.