Dubai — For decades, IT has had a visibility problem, one that has only intensified as enterprises moved workloads to the cloud, embraced hybrid architectures, and expanded across increasingly distributed environments. It may therefore come as a surprise that today, one of the most persistent performance challenges organisations face is not too little visibility, but too much of it. More specifically, too many alerts, too few of which carry real value.
Over the past five years, enterprise telemetry has increased by up to ten times. And yet, according to Enterprise Management Associates, as much as two-thirds of alerts offer no actionable value. Together, these trends point to a growing imbalance at the heart of IT operations. The ability to detect issues has advanced rapidly, but the ability to understand them has not kept pace.
As visibility has expanded, clarity has diminished, leaving IT teams navigating increasing volumes of data without a corresponding improvement in decision-making.
This disconnect is not simply a byproduct of scale, but the result of how telemetry is captured and managed across modern IT environments. Two structural factors in particular are shaping this challenge: fragmentation and sampling.
Fragmentation and the breakdown of context
Over time, organisations have built their observability strategies around specialised tools, each designed to monitor a specific domain such as networks, applications, infrastructure, or end-user experience. While these tools provide valuable insights individually, they rarely operate as part of a unified system. The result is a fragmented telemetry landscape, where data is distributed across multiple platforms, each offering a partial and sometimes conflicting view of reality. Metrics are captured in different formats, at different intervals, and presented in separate interfaces. Rather than delivering a cohesive picture, this creates multiple versions of truth.
Sampling and the loss of critical insight
Alongside fragmentation, sampling introduces a second layer of complexity. In order to manage the scale of modern telemetry, many systems rely on sampled or aggregated data rather than capturing events in full fidelity. While this approach reduces storage and processing overhead, it comes at the cost of completeness. Sampling inevitably introduces gaps. Critical moments, particularly those that occur within short time windows at the onset of an incident, may be missed entirely. What remains is a partial reconstruction of events, rather than a complete and accurate record.
Just as with fragmentation, this loss of fidelity has direct consequences for decision-making. When data is incomplete, analysis becomes less reliable. Patterns are harder to identify, anomalies are more difficult to interpret, and the distinction between signal and noise becomes blurred. In combination, fragmentation and sampling create an environment where data is abundant, but insight is constrained.
IT teams may have access to vast amounts of telemetry, but lack the continuity and context needed to act on it with confidence.
The actionable one third
Within this landscape, the concept of the ‘actionable 33 percent’ becomes increasingly relevant. While IT environments generate large volumes of telemetry, only a small proportion of that data typically drives meaningful outcomes. These high-value signals are those that provide clear, contextual insight into what is happening and why. They enable teams to move quickly from detection to understanding, and from understanding to resolution.
When effective filtration and contextualisation are in place, this dynamic shifts. High-signal insights are surfaced quickly, noise is suppressed, and teams are able to focus on what truly matters. Instead of being overwhelmed by alerts, they are guided by them, with clarity replacing ambiguity and speed replacing delay. In these environments, the value of observability is fully realised, as data is not only collected, but actively used to drive precise, confident decision-making.
AI without a foundation
The fact that only a fraction of alerts carry real value raises another irony. Increasingly, organisations look to artificial intelligence as the solution to signal overload. Yet in practice, the same issues that challenge human operators also limit AI. AI systems depend on complete, consistent, and contextualised data to deliver accurate insights. In environments where telemetry is fragmented or sampled, that foundation is compromised. Critical signals may be missing, relationships between events may be obscured, and the broader context needed for interpretation may be incomplete.
As a result, AI struggles to move beyond surface-level analysis. It may identify patterns, but without full context, those patterns are harder to trust and act upon. Rather than eliminating noise, poorly structured data can cause AI to amplify it, reinforcing uncertainty rather than resolving it. This is why the vision of fully autonomous, self-healing networks remains out of reach for many organisations. The limitation is not the capability of AI itself, but the quality of the data it relies on.
From noise to clarity
As digital environments continue to expand, the volume of telemetry will only increase. What is required is a shift in focus toward the quality, coherence, and usability of data. Extracting the actionable 33 percent depends on creating a telemetry foundation that is both complete and connected. This means reducing fragmentation by unifying data across domains, and minimising reliance on sampling to preserve fidelity. Equally important is intelligent filtration, ensuring that noise is reduced and high-value signals are surfaced clearly and in context. When these elements come together, organisations are able to move beyond reactive operations, improving both the speed and confidence of their decision-making.
This opinion piece is authored by Charbel Khneisser, SVP Solutions Engineering, Global, at Riverbed Technology.
