Fatigue Segmentation models

The Fatigue Segmentation model is a ready-to-use data science model that classifies customers into different levels of fatigue based on their profile and engagement levels.

Fatigue Segmentation model explained

When customers receive too many marketing messages through various channels, they eventually become "fatigued". This results in decreased engagement and interactions, and eventually decreased purchases and conversions for different campaigns. The Fatigue Segmentation model addresses this issue.

The model measures the message fatigue of every customer profile, which is based on the customer's engagement, history of campaigns received and opened, and most importantly the persona the customer profile belongs to. This analysis allows you to control how many campaigns to send to each customer profile. You can also determine the optimal number of messages to send to each customer profile to avoid "fatigue".

Parameters of the model

When creating the model, you will need to define the following parameters for the model:

  • Algorithm: The algorithm is the piece of code that runs the model.

  • Queries: The queries selected for the model generate a data set for model training and scoring.

  • Inputs: The inputs are query attributes from the Unity data model that are used for model training and scoring. You can't make changes to model inputs.

  • Outputs: The outputs are data objects and attributes from the Unity data model that are used to store the output values of the model. You can make updates to the default mapping of model outputs.

Model inputs

To generate values, the Fatigue Segmentation model uses the following data.

For the model to successfully run:

  • Data needs to be ingested into all the input attributes below.
  • It is recommended to ingest data at the MasterCustomer level to analyze fatigue. However, you can modify the query to ingest data at the CustomerID level and run the model.
  • Only the following channels will be considered for the model:  Email.

     If you wish to run the model for other channels, you can map the medium to Email in the query and run the model.

  • For each channel, the following Medium and Type attribute values will be considered.

    • Channel: Email

      • Medium: Email

      • Type: Open and Sent

  • Based on these requirements for event data, you can review and make the necessary updates to Events in the data model.
Data object Attribute Data type Description
Event SourceCustomerID String The unique identifier for the customer.
Customer SourceID String

The unique identifier for the source.
Event CampaignID String The unique identifier for the campaign.
Event SourceEventID String The unique identifier for the event.
Event Medium String The channel for the message, such Email, SMS, and Push.
Event Type String The type of event, such as View, Purchase, Buy, and Click.
Event EventTS Timestamp The date and time when the event occurred.
Customer SourceCustomerID String The unique identifier for the customer.
MasterCustomer MasterCustomerID String The unique identifier for the MasterCustomer.

Model outputs

Model output values are stored in the FatigueSegmentation data object.

Attribute Description Data Type Is Key attribute?
SourceCustomer_FatiguePersonaID The unique identifier that is a concatenation of SourceCustomerID and ListID. String Yes
Channel The channel that the event occurred. String No
FatiguePersona The persona assigned to the customer profile: Inactive, Saturated, Over saturated, Just right, and Under saturated. String No
MasterCustomerID The unique identifier for the MasterCustomer. String Yes
SourceID The unique identifier for the source. String Yes

Fatigue aspersions are assigned for each customer profile and channel (SMS, Email, Push). For example, the same customer profile can be Saturated for the SMS channel, but Just right for the Email channel.

The following personas are available for the Fatigue Persona attribute.

Fatigue Persona Description Example
Inactive Has not responded to a sent message in the last 180 days.
  • The customer has not responded to any sent messages in the last 180 days (6 months), so is therefore Inactive.
  • No messages have been sent to the customer in the last 180 days, so is therefore Inactive.
Saturated

Customers who meet the following:

  • Are active (responded in last 180 days).

  • A marketing message has been sent recently.

  • Experience fatigue, which is interpreted as a consistent decrease in engagement.

  • Are not being sent too many messages

These are customers who are at risk of becoming fatigued, but not necessarily due to a large volume of marketing messages.

  • Responding consistently then stops responding in less than 180 days: If a customer receives five messages a week in January 2020, but in February 2020 doesn't respond, then they are most likely Saturated.

  • Responding before, but not responding as often: If a customer usually opens one message per day from January 2020 to December 2020, but in January 2021 starts to open only one message every two weeks, then their engagement local mean (average engagement recently) is lower than their global mean (average engagement for the entire time they have been around). If the ratio of the two values is lower than one, then the customer is Saturated.
Over saturated

Saturated customers who meet the following:

  • Are active (responded in last 180 days).

  • A marketing message has been sent recently.

  • Experience fatigue, which is interpreted as a consistent decrease in engagement.

  • Are being sent too many messages

These are customers who are at risk of becoming fatigued, due to an increase in marketing messages.

A customer opens one message per day in 2020 and is sent one message every day as well. But in 2021, they are sent four messages per day, and because of the increase in sent messages, they reduce their open frequency to one message every two weeks. This customer is Over saturated.
Just right

Consistently engaged customers who meet the following:

  • Are active (responded in last 180 days).

  • A marketing message has been sent recently.

  • Are not fatigued (neither Saturated nor Over saturated).

  • Have engaged consistently and recently.

  • A customer opens one message per week in 2020, then in January 2021 opens one message per day. This customer is Just right.

  • A customer opens one message per week in 2020, then in January 2021 continues to open one message per week. This customer is Just right.

Under saturated

Customers who engage more if they are sent more messages and have not been sent an SMS, email, or push message (either recently or in a long time). They may benefit from receiving more messages.

These are active customers (responded in last 180 days) who meet one of the following:

  • Old Sent: Have not been sent any messages recently or enough, so they need to be sent more messages.

  • High-Appetite: Customers who have been sent recent campaigns, but can still take more messages.

  • Low signal: Neither fatigued (not Saturated or Over saturated) nor Just right. They do not engage consistently, but they are not decreasing consistently enough to be considered fatigued. So more messages need to be sent to get a better signal.

  • A customer is only sent a weekly digest and responds once a week in 2020, but when they were sent one message a day in 2021, they responded twice a week. If the send frequency increasing causes the customer to increase their response rate, then they are Under saturated.

  • If the send recency threshold is 10 days, any customer who was last sent a promotional marketing message 10 or more days ago is Under saturated.

Create and use a Fatigue Segmentation model

To create and use a Fatigue Segmentation model, you will need to do the following:

  1. Follow the steps for Creating Fatigue Segmentation models.

  2. After creating the model, follow the steps for Running training and scoring jobs.

After the model runs and creates output values, you can do the following:

data science, data science model, analyze data, create data science model, how to create a data science model, fatigue, fatigue analysis, fatigue segmentation