The process of selling products is not only dependent on its quality, although it is very important, but on the timing that those products are served to interested customers. The time is of the essence, as it is the the behaviour of users, meaning the actions they have performed in the past and they are performing right now, to be identified in a potential buying mood.
This is very important, because there is nothing more annoying that the right recommendations at the wrong time or the other way around. It is frustrating too, and can lead to a mistrust of the service, ending to a churn which is not good for your business.
Thus, you have to catch your clients mood and identify the right time, defining when their attitude is crossing a certain threshold that defines the impact of products’ characteristics on clients’ desires and needs.
The behaviour of users is the sum of every action, weighted by a factor that is calculated in real-time.
The model is defined to detect a specific behaviour — which is the grey line in the chart on the left — that is calculated analyzing the sum of actions that each user makes.
But what is the use of this? This is to confront the user’s behaviours with the threshold, which works as a trigger to identify whether a user is in purchase mood or looking for suggestions.
Obsviously, both behaviour and threshold lines are subject to changes; the former is mainly based on personal actions, while the latter reflects the overall set of actions, from many users.
We consider the case of purchases. The threshold is determined by the pressure applied by different factors, on both sides of the fence: who wants to buy and what is up for sale.
The concept of pressure is inherited from other fields and we borrow to define how various forces shape the overall area of the underlying environment.
What is the area of the environment? It is the set of actions that define a behaviour or many behaviours. If we consider an ecommerce as an example, one of the environment could be defined by the set of purchases and the actions are the views, add to cart, remove from cart, modify the cart and the purchase itself.
#ecommerce #algorithms #startup #recommendations #machine-learning #algorithms