AI, a new ally for telemarketing

The acquisition of new customers is one of the most important and difficult processes for a company. Traditionally it has been necessary to resort to the purchase of contact databases and make calls to all of them or to the maximum contacts that could be called according to the available resources; with really low conversion rates. This process involves high costs both for the purchase of customer data and for the time and resources invested in making the calls.

The predictive models created by machine learning, which are a tool of the artificial intelligence increasingly present in countless processes, they can also be of great use in helping to attract new customers.

Focusing on the capture of businesses and companies, their information is available to any user through the Internet, and thanks to services such as Google Places, you can make sweeps in search of establishments in a particular area. The data collected includes both contact information, essential for making calls, and basic characteristics of each one of them. This would provide access to a very large number of potential customers, at a very low cost, and with very useful features to improve the conversion rate.

The information acquired ranges from the sector to which the point of sale is dedicated, to the valuation given by its customers, and can be used as input variables for the training of predictive models, focused on detecting those stores with the highest probability of becoming a customer of a particular service or product.

Solutions for telemarketing thanks to AI

GAMCO has implemented trade prioritization solutions as a module of its development (SAIL - Sales Artificial Intelligence Launch), increasing the conversion of calls made by more than 20%, which translates into being able to get the same amount of customers with less than half the number of calls. 

The use of artificial intelligence models makes it possible to group businesses that have similarities between them, and to try to detect those establishments that are similar to others and are already customers. The models can generate as output the probability of adhesion, and even the most appropriate way to contact each store (channel, day and specific time). In this way, the conversion rate can be improved, avoiding contacting businesses of little interest or with a low probability of adhesion.

The models are adjusted and adapted to the results obtained from previous contacts. In addition to using basic customer information, it will be vital to enrich the machine learning with information on the response given when contacting, being useful the information from both successful and unsuccessful contacts.

As model predictions become more widely used, new models can be trained to use all this information to detect which characteristics are shared by adhering customers, thus achieving increasingly accurate predictions. The greater the amount of data available, the greater the robustness of the models. 

On the other hand, customers can be tracked, and information on their behavior can be used to detect those customers who have the most sought-after results and who will normally bring the most value to the company. In this way, the IA I would not only look for those stores that are most likely to convert but also those stores that are expected to provide the most value once a customer converts.

An example of the above is a marketplace that seeks to increase the number of businesses that become members. This marketplace will also be particularly interested in businesses that once they join it will generate a greater number of sales. 

The process of creating predictive models consists of choosing the type of learning structure, training algorithm and, in particular, the selection of features from the available data, which are most useful for detecting the stores with the highest probability of adhesion. It should be noted that these features may be different depending on the specific objectives pursued: maximizing customer acquisition, incorporating customers that generate greater value, avoiding customer churn or abandonment, etc.

In addition to the above, the fact of having all customers characterized and described by their behavior patterns (a behavior pattern is nothing more than a set of characteristics per customer), and included in models, facilitates their monitoring and analysis, helping in the loyalty tasks, and recommending actions to improve the service, value provided by the customer and the value perceived by him.