The functionality of the NBO model

The NBO model is a probability and prioritization model, that can be utilized to guide sales and marketing investments to most likely converting and realizing product and service sales. It consists of several different prediction models, that are combined into one bundle, NBO model. The NBO model prioritizes the prod-uct recommendations for each customer.

FIGURE 4 Illustration of the NBO model components

“The NBO model is a product recommendation model, a frame, which consists of several different, independent prediction models. It is used to recommend products and services for customers and to optimize marketing activities.” Analyst

“With NBO model we aim to find the most interesting service for each customer. On the background there is modelling and calculations about how interested each customer is about each service and what is the most interesting one for the customer.” Digital sales manager

The NBO model predictions include customers’ predicted interest towards the products and additionally, predicted acceptance rate of loan products. The inter-est prediction indicates, which product should be advertised for the customer and the acceptance model indicates, whether the customer’s loan application will be accepted or not. As a result, the model recommends each customer a product or service, that they are most likely to take in use or purchase in the following two-month time frame.

“The two-month time frame makes the recommendations comparable to each other and enables that the recommended products can be arranged to most likely, second likely and so on. With the help of the model, each marketing manager can target the marketing ac-tions to the customers that are most likely to be interested of the product they are adver-tising, without using the best customer potential of other products.” Data scientist The technical functionality of the model is however more complicated, as the NBO is a so-called black box machine learning model.

“The models are quite technical and hard to explain. So, it works as it works and generates the results, but it is not easy to explain what the model actually includes.” Analyst

The data scientist stated, that the NBO model predictions were based on cus-tomer data from the previous 12 months from the moment the prediction is made.

The model utilizes 700 background variables including for example demographic data, postal code area data, behavioral data and transaction data, of which the model selects the best variables to explain each customer’s interest. As a result, each customer gets an interest rate for all the products in the NBO model.

The NBO model includes specific interest limits for each product, which is a calculation based on machine learning. All customers whose interest rate is the same or above the NBO limit, would then be in the target group of that spe-cific product. The NBO model then prioritizes the recommendations based on the customer’s interest rate. The NBO recommendation prioritization is illustrated in figure 4.

FIGURE 4 The NBO model limits and interest rates generate the recommendations for each customer

The data scientist stated, that the optimal NBO limit is found through testing dif-ferent limits in marketing activities. The limits can be adjusted for difdif-ferent chan-nels separately, for example digital marketing and customer service, depending on each channel’s characteristics. The data scientist stated, that choosing the lim-its also include lots of business logic and decisions about how big target groups are pursued. Determining the limits is balancing between the quality and the

quantity of, for example, loan applications. If the NBO limit is lower, the bigger target group can be reached, but the conversion rate will be lower. On the con-trary, with high NBO limit a smaller target group will be reached, but the quality is better, thus, the conversion rate is higher. The digital sales manager acknowl-edged, that they also sometimes adjusted the NBO limits to achieve sales goals.

“We have sales goals and if we are reaching one goal, and we know, that that product has a big target group potential in NBO which we don’t need in advertising, we can advertise some other product which sales are lagging behind the target, for those customers. Then we get a larger target audience for that product’s marketing and we test what is the im-pact.” Digital sales manager

The marketing director pondered which should be prioritized – business goals or the NBO model.

“I think business goals should come first, but how we get to the goals, there we should primarily follow the NBO model. Business goals should start from setting a profit goal and then think about the different options how we get to the set goal. In the best case, the NBO model could guide us about how many customers from our customer base could get different products in use. It could help inform the sales goals – what sales distribution and customer relationship development gets us to the desired profit goal.” Marketing director