AI-powered omnichannel customer care. Philip Morris Sales and Marketing experience

The key trend in modern communications with the consumer is omnichannel - seamless interaction with the user through several "touch points". Artificial intelligence technologies are increasingly being used to improve customer experience in this area. Yulia Kobyzhskaya, an IT analyst at Philip Morris Sales and Marketing, talks about how to improve the quality of service in offline sales channels using technology.

Artificial intelligence and data analysis

To personalize our approach to the client, we use the most modern tools, including AI-based solutions. Instead of developing our own AI platform, we settled on using the enterprise-level SalesForce CRM system, which we use in Russia.

What does the overall picture look like? First, we analyze the market and investigate consumer preferences – we collect data. Then we process this data using Einstein Analytics, an advanced AI–based analytics tool from SalesForce. This service builds predictive models and puts down "estimates" of the level of customer satisfaction. Based on these findings, we help customers solve their problems, as well as determine who to send personalized offers to at points of sale and online. In the process of work, we use a single corporate domain-oriented database DWH (Data Warehouse).

Forecast data

In the process, the Einstein Analytics tool creates self-learning predictive models. Our version of the analytical system allows you to take into account more than 20 parameters including: the model of the client's mobile device, the level of involvement in communication, socio-demographic characteristics and other factors.

This way we get the opportunity to improve customer service and, among other things, identify dissatisfied customers who are in no hurry to report a problem to the support center. And all this without creating complex data models and writing code for each new data analysis model. It's easier for us this way:

• Build forecasts and business strategies based on historical customer behavior data.
• Create customized offers for customers to meet their expectations as much as possible.
• Automate routine tasks to focus on improving the customer experience.

The Low code vs Data Science model

Before the large-scale implementation of Einstein Analytics, we studied its effectiveness and conducted testing. His goal was to determine the level of customer service satisfaction using a predictive model. The target audience includes only adult users.

During the tests, we wanted to determine the accuracy and reliability of the forecasts that the program makes. In parallel, for the purity of the experiment, our data processing and analysis specialists wrote a separate competing predictive model in Python. A comparison of the results showed that the use of SalesForce Enterprise technologies gives reliable results.

The next stage was the creation of a prototype of a self-learning predictive analysis model. To improve the accuracy of forecasts, we determined the threshold values (baseline) within which the model could be considered effective. We linked the results of the model with customer cards in the form of estimated values. Now the data in the model is updated daily, and we have the opportunity to quickly find out that one of our consumers may be dissatisfied with the service.

Communications "right on target"

We have figured out how to assess the level of customer satisfaction. But how to increase it? It is not enough to have huge amounts of data, it is important to use them to look for points of development. To do this, we use two tools.

We transfer the received satisfaction ratings to another SalesForce tool – Marketing Cloud. A universal communication cloud platform for communicating with customers automatically sends a message to "dissatisfied" consumers, which helps to identify and eliminate the cause of the negative experience.

It is especially important that the system chooses the communication channel itself (email, messenger or SMS), depending on what is more convenient for the client. If after that the team does not see confirmation from the client that the problem has been solved, a certain automation trigger is triggered in the CRM and the consumer is transferred to a call center to address the difficulties that have arisen. Our problem solving group comes to the rescue.

This approach helps to improve the customer's consumer experience through interaction in communication channels convenient for him, and if the difficulties are not eliminated, we do not abandon the user, but help him solve the problem.

Improving the offline customer experience

It is important that we have learned to use the tools of targeted personalized communication not only online, but also in the "real" world. When a customer addresses employees at points of sale, the Next-Best-Action marketing tool helps them. Technically, it is an information system that instantly finds and displays the customer's card, allowing the employee to understand what the current level of satisfaction of this customer is and what can be offered to him in order to increase this level.

Instead of conclusions

Consumer care is "digitized", automated and works smoothly. Even if half of the company's managers go on vacation, customers will not be left without attention. SalesForce tools have helped us provide a comprehensive approach to customer care and improve their user experience. For us as a business, we see an important advantage of this approach: we no longer need to spend huge sums on calling all customers every day in an attempt to understand how satisfied they are.

This is not the only example of an individual approach to the client. Predictive models, the relationship with CRM and Marketing Cloud, tips for employees at points of sale form an omnichannel communication environment in which any client receives messages adapted for him on programs and services of interest to him.