How Marketing Automation is transformed by AI and Data Science

You already know Machine Learning

Machine Learning is already used by many big companies. You might have experienced its potential when using Netflix. It provides you with movies recommendations based on Machine Learning itself. Facebook, Spotify, Google Maps or Uber are also taking advantage of this technology on a daily basis.

SALESmanago Copernicus Machine Learning& AI engine

SALESmanago Marketing Automation has developed its own AI engine – SALESmanago Copernicus Machine Learning&AI. Just now companies such as New Balance, Yves Rocher and Sizeer are using it to provide their customers with tailored and intelligently personalized content.

Copernicus is one of the most advanced recommendation engines across MA systems. It’s designed especially for eCommerce to guarantee real-time personalization and segmentation. It allows:

Big Data Analysis with most advanced machine learning frameworks and algorithms
Predictive product recommendations
Real-time omnichannel personalization
Gather and analyze all data in one platform
Report, analyze and improve results

Marketing Automation is now used to assign Lead Scoring, create Lead Nurturing campaigns, Workflows, Automation Rules, to run segmentation and in overall campaigns. The problem here is a huge amount of data to be analyzed.

The data analyzed in SALESmanago include:

Website visits
Products bought
Products added to cart
Conversion paths
Conversion sources
Buyers personal and demographical data (CRM)
Purchased products’ attributes
Reactions to direct marketing
Search terms used
Chat conversations
Products displayed
Cart value
Offline behavior

Today’s technology is based on following instructions, algorithms. It is a simple mechanism performing actions deriving from our primary input. Machine Learning is a completely new approach. Similarly to a human being its learning potential is coming from experience or to be exact from continuous gathering and analyzing data.

Types of Machine Learning

Supervised Machine Learning

This kind of learning is possible when inputs and outputs are clearly identified and algorithms are trained using labeled examples.

Regression – predicts the continous-response value.

Classification – predicts the categorical response value where the data can be separated into specific “classes”. It’s potential is based on learning from examples.

Unsupervised Machine Learning

Unlike Supervised Machine Learning, Unsupervised learning is used with data sets without historical data. An unsupervised learning algorithm explores surpassed data to find the structure.

Clustering – grouping similar elements together.

Association – the goal is to identify rules that define large portions of the data.

Reinforcement Machine Learning (Deep Learning)

The most sophisticated of all types. A system interacts with a dynamic environment in which it must perform a certain goal. The system is provided feedback in terms of rewards and punishments as it navigates its problem space.

This kind of Machine Learning is adapted to solving most complex problems such as:

robotics and industrial automation
automotious vehicles
health and medicine
text, speech and dialog systems.

Stages of implementation

The project consists of 5 steps:

1. Determining objectives, metrics and constraints

You need to focus on results you want to achieve and imagine constraints of the project.

2. Assessing data, data collection

With multiple sources of data, not always provided by the customer the next step is assessing and collecting data.

3. Model training

Building and training model for your project.

4. Integration and testing

5. Model monitoring

By monitoring the results you are able to adjust it to reach the expectations and get better results.

Machine Learning in Marketing Automation

Smart Segmentation

Product Images

Segmentation is the process of dividing a broad data set into sub-groups (segments) based on some type of shared characteristics.

Your database will be segmented into sections based on multiple attributes. For example if your customer is interested in red shoes, you can provide him with red shoes recommendations. If they’re however looking for red shoes with three white stripes, the machine learning will allow you to deliver such recommendations of fitting products, so in this case red shoes with three white stripes you have in your offer.

Number of purchases

You can analyze the number of purchases and subdivide the database based on purchases made by customers from different groups such as women who performed at least 10 purchases or people in their thirties with no history of a purchase and then adjust your actions to the characteristics of a specific group.

Basket Value

Using segmentation you can try to find specific characteristics of given segments – like basket value in age, frequency of purchases (RFM) or city size segments.

Sentiment Analysis

Analysis providing with information about customer’s interest in particular topic, product.

Dynamic Pricing

Manual discount decisions can be replaced with AI engine calculating discount based on probability of purchase maximizing income across all customers.

Retention Improvement

Client churn prevention – complex correspondence, behavior and interactions analysis identifying customers of high churn potential.

Product Recommendations

Unlike product recommendations engines available in most Marketing Automation platforms, AI recommendations are not only based on the product data itself or adjusted to specific user’s behavior in 1-to-1 model.

ongoing analysis of data about visits and transactions
results are calculated continuously and change according to changes in customers’ behavior
statistical analysis computes the probability of co-occurrence of events (if A, then B)
to calculate the strength of data dependence system uses weighing scales and amount of occurrences