App: Predictive Analytics

Spotler Activate offers a powerful tool to improve audience targeting with Predictive Analytics app. This feature uses predictive models in collaboration with Cross Engage to help create audiences based on predicted purchase behaviors and churn risk. It adds a new layer to the targeting strategies already possible in Spotler Activate.

Predictive 1.jpg

360 profile predictive1.png

What is Predictive Analytics?

The Predictive Analytics model learns from a merchants' historical purchase data and makes predictions, which it then validates using the same data to improve over time. This process continues until it generates accurate predictions for the next 30, 60, and 90 days. These predictions include details about the customer's decile (retargeting group), their churn probability, and the expected purchase value for each time frame.

Currently, audiences are created using historical or manual data. While this works for many situations, targeting based on future behavior often comes down to educated guesses. Another option is our RFM module, which categorizes customers based on their transaction history. This offers a more structured way of targeting, but when it comes to predicting future behavior, it still relies on intuition. With Predictive Analytics, we move beyond guess work and make targeting more data driven.

Key features

CLV prediction

Predictive Analytics helps you identify customers who are likely or unlikely to make a purchase in the next 30, 60, or 90 days. By analyzing Customer Lifetime Value (CLV) trends, you can view predictions at the profile level and compare them to historical data.

Profile with a rising prediction vs historic data: Good candidates for upselling and loyalty program campaigns, for example.

Key 1.png

Profile with descending prediction vs historic data: Good candidates for re-engagement and churn prevention campaigns, for example.

Key 1 1.png
Churn probability

Predictive Analytics sorts customers into three churn risk categories: Low, Medium, and High. This helps you identify which customers are at risk of churning, allowing you to take targeted actions to improve retention.

Low probability: Customers in this category are highly engaged and likely to continue purchasing without you intervening. Focus on maintaining their loyalty, for example.

Key 2.png

Medium probability: These customers show early signs of reduced engagement. Timely, personalized re-engagement campaigns can help retain their interest.

Key 2 1.png

High Probability: Customers in this group are at a high risk of leaving. Proactive customer retention strategies, such as exclusive offers or direct outreach, can be examples of how to win them back.

Key 2 3.png

Order History

At the profile level, Predictive Analytics gives you insight into customer purchase patterns. By looking at both the average order value and time between orders, you can spot trends across different customer groups.

Average Order Value: Understand the typical spend per transaction for various customer segments. You can use this to tailor promotions or recommend products within a similar price range.

Average Time Between Orders: Analyze how often customers make purchases. This can help you identify patterns like regular buying cycles or extended gaps, which might indicate that customers are either engaging more or disengaging with your product.

Key 3.png
Predictive Segmentation

Customers are divided into 10 deciles (retargeting groups) based on their predicted Customer Lifetime Value (CLV) for the next 30 days. This segmentation lets you focus your efforts on the groups that are most relevant to your campaigns.

Decile 10: Represents the top 10% of customers with the highest predicted CLV. These are your most valuable customers, perfect for targeted offers or loyalty programs, for example.

Decile 1: Represents the bottom 10% of customers with the lowest predicted CLV. These customers may need extra attention, such as re-engagement campaigns, to retain them and boost their value.

Key 4 1.png

The average 30-day CLV for the entire decile is displayed as well. The predictive analytics filters can also be used with the custom audience creation tools. You can apply a range of filters based on these predictions to create audiences more tailored to your campaign goals.

How does Predictive Analytics work?

Once a day, purchase data from Spotler Activate is automatically uploaded to the model. This ensures the predictions are always based on the most up to date information, while also providing continuous data for the model to learn from.

Once the data is received, it’s processed through a standard predictive model that analyzes customer behavior, including purchase frequency, order value, and engagement patterns. Using these insights, the model generates predictions about future behavior, such as the likelihood of a customer making a purchase or their risk of churn.

To ensure accuracy, the model constantly validates its predictions by comparing them to actual customer outcomes. This allows the system to refine itself over time and improve its accuracy. The process works as follows:

  1. Learning: The model learns from new data sent each night, refining its understanding of customer behavior to get smarter over time.
  2. Predicting: The model then predicts future behavior, including purchase likelihood (within 30, 60, or 90 days) and churn risk.
  3. Validating: Predictions are validated by comparing them with actual outcomes. If a predicted purchase doesn't happen, for example, the model adjusts and learns from the discrepancy.
  4. Improving: Based on validation results, the model improves by identifying any gaps between predictions and actual behavior. This continuous cycle ensures that the predictions get more accurate and reliable.

Activate the Predictive Analytics App

In Apps you will find the Predictive Analytics app. First, check if your situations suits the required conditions of use. Then, click on the Activate button. 

predictive 3.png

This is a premium app

For Premium apps an additional license is needed. Click on the Activate button, there you will find the additional monthly costs. 

Predicitive 4.png

Conditions of use

To ensure accurate predictions, merchants must provide/have historical purchase data that meets the following criteria:

  • Minimum of 2 years of purchase data: This allows the model to identify meaningful trends and patterns in customer behavior.
  • Maximum of 4 years of data: Keeping the data within this range ensures relevance, as older data may not reflect current customer trends.

The model requires a sufficient volume of data to build a reliable dataset:

  • At least 5,000 purchases per year: This ensures the model has a diverse range of customer behaviors to analyze, improving the accuracy of its predictions.

For the top 10% group, this is different from the deciles, the following applies:

  • Minimum 5,000 customers: if there are not enough customers for the group it fills up to a minimum of 5000 by adding customers
  • Maximum 20,000 customers: if there are more then 20k profiles in this top 10% it cuts to a maximum of 20k profiles.

Use Predictive Analytics data in Audiences

Naturally predictive analytics can be used for all sorts of use cases, the following are a few general examples:

Send everyone who is likely to churn an incentive

Use Case Objective: Retain customers at high risk of churning by sending them tailored incentives to re-engage.

Step 1 Segmenting High Churn Customers

Filter your audience based on the Churn label or Churn probability. Choose the appropriate label or probability. For example:

use case 1.png

  • Identify customers with a high likelihood of churning.
  • Focus on those who need additional engagement to retain.

Step 2 Create a Targeted Campaign & Activate Your Campaigns

Prepare your campaign with personalized content, Journeys, Audiences, mailings, or ad campaigns. Use the relevant audience segments for maximum impact.

Step 3 Measure the Impact

Analyze campaign performance to determine its effectiveness in reducing churn. Adjust incentives or communication strategies based on results to optimize engagement.

Exclude likely buyers from expensive ad campaigns

Use Case Objective: Optimize your advertising budget by excluding customers who are already likely to make a purchase from high-cost ad campaigns.

Step 1 Segment Likely Buyers

Filter your audience based on a high decile or a range of deciles that represent customers with a high probability of purchasing without additional prompting. These are the customers to leave out of campaigns.

use case 2.png

Step 2 Exclude this Group in Ad Campaigns

Exclude these audiences from your paid ad campaigns. This ensures that your ad spend is focused on less engaged customers or those at medium or high churn risk, where advertising can have the greatest impact.

Step 3 Track Performance

Compare ad spending and ROI before and after excluding likely buyers to assess cost savings. Optionally run unpaid campaigns (e.g personalizations, journeys) on these groups to see if this is enough to get the likely buyers converting.

Invite your top 10% customers to exclusives

Use Case Objective: Reward your most valuable customers by giving them exclusive access to a VIP event, early product access, or other meaningful rewards. This strategy helps strengthen customer loyalty and increases their lifetime value.

Step 1 Segment Your Top 10% Customers

Focus on your top 10% of clients for this initiative. Filter your audience accordingly to ensure only your most valuable customers are targeted.

use case 3.png

Step 2 Design Campaigns Tailored to These Customers

Create personalized campaigns that emphasize their exclusive status. Use your usual communication channels to highlight their importance and offer relevant incentives.

Step 3 Follow Up

After offering the incentive, follow up with a personal message tailored to this elite group. Ensure they feel valued and appreciated as your top-tier customers.

Change the tone of voice or selected products in automated communication based on deciles

Use Case Objective: Personalize your automated communication by tailoring the tone and product recommendations to the decile a customer belongs to. This ensures relevancy and maximizes engagement across different customer groups.

Step 1 Segment Your Customers into Separate Deciles or Filters

Divide your customers into deciles or other relevant filters. For example:

  • Group 1: Deciles 1-3 (low engagement or value).
  • Group 2: Deciles 4-6 (medium engagement or value).
  • Group 3: Deciles 7-10 (high engagement or value).

use case 04.png

Alternatively, use other filters such as churn probability or order frequency to refine segmentation.

use case 041.png

Step 2 Adjust Messaging and Tone

In campaigns that target all customers, customize the tone of voice and messaging for each decile or group. Use Journeys to adapt your communication strategies based on engagement level. Examples include:

  • A formal tone for high-value customers.
  • A friendly and encouraging tone for medium-value customers.
  • An urgent and persuasive tone for low-value or at-risk customers.

use case 4 1.png

Step 3 Track Performance

Monitor performance metrics for each version of your campaigns. Compare engagement rates and conversions across the different adaptations to evaluate the effectiveness of your personalization strategy.