What is the best visualisation to quickly spot outliers in this two-variable dataset ?

Shahir
Updated on December 14, 2025 in

You’re working with a performance dataset from a rapidly growing digital platform that serves millions of users across different regions and device types. The dataset captures two core numerical metrics for every user session: processing time and resource consumption. These two variables often move together, but not always and the moments when they don’t align usually indicate deeper issues such as capacity overload, inefficient requests, or poorly optimized devices.

As you explore the dataset, you notice that summary statistics alone can’t give you the clarity you need. The averages look normal, the percentiles look acceptable, yet some users are still reporting unexpected slowdowns. When you dig deeper, it becomes clear that the problematic behaviour only emerges when both numerical variables are analysed together. Patterns don’t show up in isolation; they show up in the relationship between the two.

  • 2
  • 66
  • 4 weeks ago
 
on December 14, 2025

In situations like this, I stop looking at the metrics separately and focus on how they interact. Plotting processing time against resource consumption often reveals clusters or outliers that averages completely hide. Those mismatched points usually point to specific regions, devices, or request types behaving inefficiently. Once you see the relationship visually, the root cause becomes much easier to isolate and explain.

  • Liked by
Reply
Cancel
on December 8, 2025

This is exactly the kind of scenario where numbers behave perfectly fine on paper, but reality is hiding in the space between the variables. And it’s such a classic trap in performance analytics—everything looks “within limits” until you start exploring how two seemingly normal metrics behave together.

Averages, percentiles, even distribution curves—they’re all comforting because they suggest stability. But they also flatten nuance. They strip away the messy interactions that real-world systems thrive on (or break because of).
So you get this odd situation where every metric looks healthy individually, yet users are still complaining. And it’s only when you put processing time and resource consumption side-by-side that the story starts to reveal itself.

  • Liked by
Reply
Cancel
Loading more replies