Predictive Analysis

for marketing

Predictive analytics enable:

Assess customer CLV looking at their behaviour and going beyond socio-demographic classifications.

Predict customer churn in order to activate a protocol to retain them.

Build customer profiles on their estimated willingness to buy to plan targeted sales.

create a strategic database using combined data from different channels.

Increase approval rating and satisfaction index of the customer base.

Pythía – our platform

Through Pythía it is possible to manage: 

  • The file transfer process.
  • Data preparation for the AI engine.
  • Model training and validation.
  • Delivery of output to the dashboard.

 

Dashboard

This dashboard allows users to visualise data extracted by the AI engine together with more standard statistical data.

Features:

Customer: the user can visualise all the statistics related to single customers: CLV, churn rate and probability of purchase on the basket of selected products. This is in addition to other information that might be available regarding asset liabilities, credit card usage and online activity.

Products: by selecting a product the user can visualise the segmentation of the client base.

Data sources: this feature allows you to configure the path where files are uploaded and contains coding and decoding hash/obfuscation keys.

Marketing using AI

 

  1. We analyse all the customers and their transactions.
  2.  

  3. We normalize their behaviour using economic cycle variables.
  4.  

  5. Using our advanced AI techniques we produce a strategic database for customer profiling, with information on expected returns and willingness to buy specific products.
  6.  

  7. Using this data it is possible to set marketing strategies to targeted campaigns that predict and anticipate needs or customer churn.

Predict need using AI

Artificial intelligence in predictive analytics

AI is the most effective way to obtain accurate, automated predictions on a large volume of data that cannot be obtained using traditional methods.

Deep Neural Embedding

Customer profiling happens inside a neural network that renders it specific for the target of the network itself.

better prospective profiling

RNN - Reti Neurali Ricorrenti

A model of a temporal dynamic behaviour of customers with automatic detection of relevant variables.

better prediction accuracy

Inverse Reinforcement Learning

Extraction of the maximum amount of information from the interaction with the customer through learning processes similar to those of humans.

better marketing campaign effectiveness

Predictive analytics for client profiling

CUSTOMER LIFETIME VALUE

Assessment of expected profitability.

COLLABORATIVE FILTERING

Assessment of winning product/customer combinations from a profitability standpoint.

Profiling

1. Customer segments that are probably interested in purchasing a given product.

2. Customer segment that could leave for a competitor.

From predictive analytics to marketing strategy

1.

Segment

Segments to invest on are identified.

2.

Target

Given the segment, a target is set: 

  1. Prevent customer churn.
  2. Example  predict a need.

3.

Message

The correct message for the customers is identified.

4.

Action

Communication channels needed to reach the desired target are used.

An example of digital strategies for retention/upselling

Applied to customer segments processed using AI.

Contact sources

Available in the database

  • Email.
  • Mobile phone.
  • Address.
  • Customer profile on the home banking platform.

Obtainable

When the user that is part of a given segment browses in the home banking platform while being logged in, it is possible to install two cookies on its browser. Those cookies are provided by Facebook and Google and identify them as being part of that specific segment when they will browse on those platforms in the future (see following slide).

Contact means

Message differentiation

Using social networks and Google’s display network it is possible to create an additional profile of the customer, so as to create a bespoke message.

As an example, it is possible to target users that are part of the segment who is potentially interested to a new mortgage using specially-made multimedia advertisements

Are you interested to this solution?