In an effort to move away from mass marketing and towards personalization, marketers are ever more interested in their customer’s journey. Marketing professionals do not only want to know what types of products interest their customers but also when they will need the specific products in question. They want to be able to predict each customer’s next purchase. This benefits both parties as customers are not constantly receiving messaging that they are not interested in and companies are not wasting their budgets to send it.
Currently, companies run market basket analyses to determine which products are generally purchased together. This is very helpful when determining a store layout or which recommendations to put around which products online. Yet, this is not at all helpful in determining a future purchase. Today, companies only know that a customer is interested in a specific product if they show explicit interest, such as logging in to the website and clicking on an item or putting it in their basket. Otherwise, companies have no idea what customers might want or need next.
To determine what a specific customer might want or need next, companies need to study customer lifetime behavior to find patterns in purchase sequences. Companies that offer products that are highly influenced by the customer’s life cycle may benefit greatly from analyses that look at the average delay in purchase between the various product offerings. Customers in the insurance and banking industry, for instance, have products that are highly dependent on where the client is in their life. Such companies might benefit from finding patterns in the purchase patterns in order to help customers buy the right product at the right time.
Today, there is no out-of-the-box algorithm that will do this analysis. Even if one did exist, it would need to be trained and tuned for the specific industry that the company is in to give any sort of meaningful feedback. In order to spot complex, sequential patterns, an algorithm using Recurrent Neural Networks will likely need to be created to solve this challenge. What other industries do you think might benefit from such an analysis? Have you ever created a similar algorithm for another type of marketing analysis?