Amplified Examination Revealing the Effect of Product Features on User Retention

In the realm of product analysis, understanding the performance of various features is crucial. However, traditional methods often lack the necessary clarity when it comes to combining both quantitative (user count) and qualitative (retention) metrics. To address this issue, a 50/80 percentile scatter plot has emerged as a powerful tool.

The scatter plot, with user count on the X-axis and retention on the Y-axis, provides an effective way to visualize product feature performance. To make this chart even more informative, consider the following key practices:

Visual encoding enhancements: Distinct marker shapes, sizes, and colours can represent different categories or percentile groups. For instance, colour-code points by retention tiers (high/medium/low retention), and size markers by user count. This differentiation helps viewers immediately distinguish qualitative retention levels alongside quantitative user counts.
Apply the 50/80 percentile rule for segmentation: Classify data points by whether they fall below or above the 50th and 80th percentiles for each metric. Use the resulting segments to group features visually, such as clustering or contouring regions in the plot. This highlights feature performance tiers clearly and facilitates strategic insights.
Customize marker transparency and borders: Alpha values (transparency) can prevent occlusion where many points overlap, improving visibility of dense clusters. Adding edge colours or borders to markers separates points visually for readability.
Labeling and annotation: Include concise, non-overlapping labels for key areas or outliers. Position labels strategically near relevant points without cluttering the plot. Tooltips or interactive labels can offer dynamic contextual information without crowding the visual.
Use intuitive axes and grids: Clearly label axes with units and metrics. Adding gridlines aligned with 50th and 80th percentile thresholds makes the percentile rule explicit on the plot, aiding interpretation.
Select colorblind-friendly palettes and contrast: Choose colours with strong contrast that are accessible to all users. Dark text on a light background or vice versa improves readability, maximising clarity.

By combining these elements—careful encoding of both quantitative and qualitative data, marking percentile boundaries explicitly, and ensuring clean labeling with colour and size distinctions—the scatter plot becomes a powerful tool for analysing product feature performance in line with the 50/80 percentile rule.

This approach aligns with best practices in data visualization for clarity and interpretability, particularly when dealing with mixed metric types and percentile-based segmentation.

Moreover, this method can help avoid suboptimal or incorrect decisions by providing a comprehensive view of product features. Cluster centroids analysis reveals that some features can move from cluster to cluster over time due to factors such as new user acquisition efforts, UX changes, and user base maturing.

The analysis showed that only ~20% of product features fall into the Core + Power clusters. The Core features, which are used by more than 80% of users and have a return rate of more than 80%, are the true essence of the product, typically numbering 2-3. Power features, delivering about 80% of the regular value, can be further divided into two categories: Power1 features, as popular as Core features but less valuable, and Power2 features, bringing as much value as Core features but less popular. There are usually 3-5 features in each Power cluster.

Interestingly, niche features, used by just 11.3% of users, have the highest retention, even higher than the Core cluster. These features, while used by a limited number of users, can bring enormous value. Casual features, used from time to time, are supporting features that bring some value to users but are not as central as Core or Power features.

In the next post, another perspective on feature retention definition will be discussed. Stay tuned for more insights into product feature analysis!

Amplified Examination Revealing the Effect of Product Features on User Retention