📢The Danger of Preconceived Conclusions in Data Interpretation🪧

In the world of IT and telecommunications, data is everything. But what happens when your interpretation of that data is skewed by a preconceived notion or bias? You can easily find yourself bending facts to fit a predetermined conclusion instead of letting the data speak for itself. This phenomenon often leads to flawed decisions and missed opportunities for improvement. Let’s explore this through a few key examples and wrap up with how modern technologies like SD-WAN can mitigate these biases.

The Trap of Preconceived Conclusions

Confirmation bias is when someone already believes in a conclusion and interprets data in a way that supports it. This can happen to even the most seasoned professionals, especially when interpreting performance data. In IT, it’s easy to misinterpret network or server performance and be blinded by what you expect the results to be.

I once worked with an IT manager who had an interesting theory about optimizing his developers' work. His solution was to put them on a 64 kb/s Diginet link, believing it would force them to write more efficient code. While his reasoning had some merit—being in an environment that forced efficiency—the speed of this link was barely functional by modern standards. Diginet, provided by Telkom, capped out at 2 Mb/s, but most opted for the cheapest 64 kb/s, often without considering actual requirements. The result? A sluggish user experience that was rarely addressed until it was too late.

The Importance of Accurate Data Interpretation

This leads to a much broader issue: how data is interpreted when monitoring systems or performance. I encountered this while working with a financial services provider that had outsourced the monitoring of its servers to an external network operations center (NOC). They were using Concord Network Health, an advanced tool at the time. Yet, I kept noticing gaps in the data.

The service provider assured me that everything was functioning "normally" and showed me a graph of the link’s utilization, reporting an average low usage. However, I noticed something off—my own network management system was reporting unusually high error counts. This mismatch led me to investigate further using a tool called STG, which allowed for SNMP polling every second.

What I found was shocking: the link was hitting 100% utilization every minute, coinciding with the service provider’s polling intervals, causing packet drops. Their reliance on five-minute averages masked the actual problem, and the errors were lost in their data’s low-resolution averages.

When I increased the link’s bandwidth to 256 kb/s, the gaps in server data disappeared. It wasn’t the network utilization that was the problem—it was the momentary spikes that their tool simply wasn’t capable of seeing.

The Perils of Survivorship Bias

What I experienced was a form of survivorship bias—the service provider was only looking at the data that seemed healthy, ignoring the crucial moments of failure. They assumed that since most of the data points looked fine, the whole system must be fine. Survivorship bias occurs when you focus only on the "survivors" of a process, ignoring the failures or anomalies that don’t fit the expected pattern.

In my case, the provider was so confident in their averages that they missed the spikes entirely. Had I trusted their graphs without question, the root problem might never have been solved.

Why SD-WAN Reduces the Likelihood of Bias

This is where SD-WAN comes in as a game-changer. Unlike traditional, manually monitored networks, SD-WAN systems operate with a high level of automation. This automation reduces the human influence that often leads to misinterpretation of data or selective bias.

Automated Monitoring: SD-WAN constantly monitors network conditions in real-time, dynamically adjusting bandwidth and prioritizing traffic. This ensures that packet loss, jitter, and latency are addressed without human intervention—eliminating the likelihood of misinterpreting spikes or bursts in traffic.
Centralized Visibility: SD-WAN provides a centralized view of the entire network, regardless of the underlying technology. This allows you to see the network holistically rather than piecing together fragmented data from disparate systems.
Real-Time Data: Unlike traditional systems that may rely on outdated or infrequent polling intervals (like the five-minute averages I encountered), SD-WAN operates with real-time data, reducing the risk of important events being masked or averaged out.
Policy-Driven Approach: Instead of relying on manual configurations or human oversight, SD-WAN operates based on predefined policies that automatically adjust traffic priorities, reroute traffic, and manage resources. This minimizes the chance of human error or bias.

By eliminating many of the manual processes associated with traditional WAN management, SD-WAN provides a more accurate, unbiased view of network performance, enabling IT teams to make better decisions based on real-time data rather than flawed interpretations.

Wrap

In data interpretation, it’s easy to fall into the trap of confirmation bias or survivorship bias—seeing only what you expect or want to see. The example of the slow Diginet link and the service provider’s flawed monitoring illustrates how preconceived conclusions can blind us to the actual problem.

With SD-WAN, many of these pitfalls are avoided. Automation, real-time monitoring, and centralized control ensure that data is interpreted correctly without human biases. The result is a more reliable, efficient, and secure network that can adapt to business needs in real time. So, if you're tired of relying on outdated assumptions and missing the bigger picture, it’s time to embrace SD-WAN and let the data speak for itself.

Ronald Bartels ensures that Internet inhabiting things are connected reliably online at Fusion Broadband South Africa - the leading specialized SD-WAN provider in South Africa. Learn more about the best SD-WAN provider in the world! 👉 Contact Fusion

https://hubandspoke.amastelek.com/discover-fusion