Data Science & IA

Product Recommendation Model

The challenge

Our client, an automotive company, has different business lines: Automobiles, spare parts, accessories, technical service, among others. They had achieved, through marketing efforts, customer service, and people’s loyalty to the brand, to reach the goals in all their business lines except those related to accessory sales. In order to boost the sales of this line of business, it was defined that the generation of an automated model for recommending accessories to support salespeople could be an element to meet the goals and facilitate the process.

The use of 3 complementary models provides a pragmatic solution to product recommendation, allowing all possible cases to be covered

The strategy

At the time of designing the model that would best fit the project, we began by exploring the data, which allowed us to understand that to meet the agreed objectives it would be necessary to develop three different models, which would generate recommendations for 3 types of situations that would be frequently seen at the time of making the sale (e.g. customers with history, customers without history). This was validated by accompanying the salespeople in their process, discussing the different areas of the company and also understanding the systems and tools available.

The process of accompanying the salespeople also allowed us to understand their doubts regarding the solution that was being built, seeking to take their inputs and contribute to the adoption of the tool. The combination of developing an automated model, understanding the reality of field work is fundamental in every project we undertake, and this was no exception.

Inheritable SKUs model

for customers looking to change their car and add accessories

Model clone customers

for new customers, looking for similar characteristics with other

Ranking by model

for new customers who cannot be categorized

The achievements

The customer was provided with 3 automated models that would contribute to the sales process:

SKU recommendation model based on the history of that same customer, using variables such as previously purchased car models, previously purchased accessories, compatibility of the accessory with the model of interest among others.

Recommendation model based on similar customers, associating the customer to an identified customer segment within the company, and recommended according to customer preferences of the same segment.

Ranking model, to be used when no historical customer information is available to be used as input for the above mentioned models, and to establish recommendation priorities according to sales behavior of different accessories.