In ‘The Executive Decision Series,’ I explore key decision science concepts that empower leaders to make smarter, data-driven decisions. Today’s focus is on leveraging decile tables to rank and segment audiences using predictive model scores. In part two, we’ll take this further by calculating ROI per segment to optimize marketing spend and drive maximum impact.

What are Decile Tables?

In the world of predictive modelling, understanding how to act on a model’s insights is as critical as the model itself. How can leaders confidently allocate resources to maximize impact in a data-driven world? Enter the decile table—a simple yet powerful tool that helps rank, segment, and prioritize audiences for optimal results.

This post explains the purpose of decile tables, how to interpret their metrics, and how to turn insights into action for better decision-making.

Decile Tables are Executive-Friendly

During a recent review of a predictive model used for segmentation, I was asked to find a more “Executive-friendly” way to communicate model performance than a decile table. This surprised me. Decile tables are already an incredibly simple and powerful summary tool.

Decile tables are designed to be both data-driven and intuitive, condensing complex machine learning outputs into actionable insights. They require no technical expertise, take just one slide to present, and apply universally across machine learning models. In today’s AI-driven world, understanding decile tables should be considered essential data literacy for Executives.

Maximizing Returns Through Improved Targeting

A decile table is a summary report that divides a dataset into 10 equally sized groups (deciles) based on a predictive model’s probability scores (e.g. probability to buy (Yes/No)). Each decile represents a ranking, with Decile 1 containing the highest scores and Decile 10 the lowest.

Decile tables allow us to answer critical business questions:

• Which customer segments are most likely to respond to a campaign?
• How does the model perform across different groups?
• Where should we focus resources to maximize ROI?

A typical decile table includes key metrics such as:

• Conversion Rate: the percentage of responders in each decile.
• Lift:  a comparison of a decile’s response rate to the overall average, indicating how much better the decile performs than random targeting.
• Cumulative Metrics: aggregated percentages of customers and responders to show cumulative gains.
• Probability Range: the confidence level of the predictive model for each decile, showing the likelihood of customers responding.

Interpreting the Decile Table: A Step-by-Step Guide

Below is an example decile table from a predictive model. Let’s walk through how to interpret it.

1. Focus on the Top Deciles

The top deciles, especially Decile 1, represent the highest likelihood of response.

• Example: In Decile 1, the average conversion/response rate is 22.4%, and the lift is 2.28. This means customers in Decile 1 are 128% more likely to respond compared to the overall average.

This insight demonstrates that targeting Decile 1 will yield the most significant results for your campaign. As you move to lower deciles, lift decreases. By the time you reach Decile 10, the lift is 1.00, indicating no added value compared to random selection.

2. Understand Cumulative % of Responders

Cumulative metrics help quantify the total opportunity within the top-performing deciles.

• Example: the first 3 deciles capture 55.7% of all responders, while targeting Deciles 1–5 captures 76.7%.

By focusing on just the top 30% of customers, you can capture over half (55.7%) of all potential responders.

3. Assess Conversion Rates Across Deciles

Conversion rates drop significantly in lower deciles.

• Example: the average conversion rate in Decile 1 is 22.4%, while Decile 10 has a conversion rate of just 0.8%.

This demonstrates diminishing returns as you target lower-probability customers.

4. Consider the Probability Range

The probability range provides context for the model’s predictions.

• Example: Decile 1 includes customers with a probability range of 0.71–0.77, reflecting high confidence in their likelihood to respond.

By comparing ranges across deciles, you can understand how prediction certainty changes, a measure of risk.

The Cumulative Gains Chart

The Cumulative Gain Chart visually represents the effectiveness of a predictive model by comparing it to random targeting. It connects directly to the decile table by plotting cumulative percentages of responders for each decile, providing a clear way to assess the added value of using the model.

Interpreting the Cumulative Gains Chart

1. The X-Axis (Deciles): the chart divides the dataset into ten equally sized groups (deciles) based on the predictive model’s probability scores. Decile 1 represents the top 10% of the dataset with the highest probability scores, while Decile 10 represents the bottom 10%.
2. The Y-Axis (Cumulative % Responders): this measures the cumulative percentage of responders captured as you target each decile sequentially. Higher cumulative percentages in the early deciles indicate a better-performing model.
3. The Model Curve (Cumulative Gain Line): the upward curve reflects how the predictive model prioritizes responders in the top deciles. The steeper the curve at the start, the better the model is at identifying high-probability responders.
4. The Random Baseline: the diagonal dashed line represents random targeting. If the model were not used and customers were targeted randomly, responders would be evenly distributed across deciles. For example, targeting 30% of customers would yield only 30% of responders. A predictive model should outperform this line.

Conclusion: Why Decile Tables Matter

Decile tables bridge the gap between predictive models and business decisions. They transform complex machine learning outputs into simple, actionable insights, enabling leaders to identify high-value segments, allocate resources strategically, and achieve optimal results.

No technical jargon is needed—just one slide of powerful, reusable insights. As businesses continue to embrace AI and machine learning, decile tables remain an essential tool for driving data-driven decisions that deliver measurable impact.

In part two, we’ll take this further by calculating ROI per segment to optimize marketing spend and drive maximum impact.