Many companies limit the data to be analysed to save costs. The result: the small section leads to a distorted perception of the situation and, thus, wrong decisions. Instead, companies should collect all the available information and evaluate it.
More data is available today than ever before. Therefore, one might think it is not a problem to obtain and analyse the necessary information. But it’s not that easy. In practice, decision-makers limit the data’s determination, analysis, and evaluation to keep costs and management efforts low.
Only five out of 100 organisations can gain the necessary insights from the many monitoring tools to monitor their digital assets’ performance, availability, and security. That’s the conclusion of F5’s current State of Application Strategy Report.
Data Distortion Due To Cost Pressure
A fatal result, considering that there is neither a lack of suitable tools nor data, but the existing data is distorted. This is a subjective selection process: Decision-makers determine which information is recorded by systems and how the visualisations are displayed on dashboards.
One of the leading causes of data distortion is software agent austerity. These are usually linked to IT systems to collect data and forward it to an analysis platform automatically. While there are discounts for large deployments, companies often choose not to install agents across systems to reduce costs.
Subjective Selection Criteria
However, this distorts the data since information is not collected from all systems. The decision as to whether an agent is used or not is often determined by the subjective, cost-based opinion of the decision-maker – regardless of their experience.
The amount of data generated can also lead to limitations, along the lines of: Do you need all of these metrics, or can we limit ourselves to the three or four most important ones? There is no way to validate shifts in the data with such limitations. These changes would indicate a potential problem but are overlooked because the data was deemed insignificant.
Finally, decisions about the visual representation in the dashboard may further skew the interpretation. The selection is often based on one’s skills and experience, which other dashboard users do not have. Even the choice of graphics can lead to distortions. This is especially true for time-series-based operational metrics such as performance and uptime.
Bar charts are commonly used to show time-series data but are often not as descriptive as line charts. These decisions can have profound implications, such as when technicians rely on visualisation to gain insight into a system’s operational status.
All of these selection processes result in data being continuously skewed. And that translates into the ability to interpret them – and understand their actual meaning.
Combat Data Distortion
There should be as little data distortion as possible to make correct decisions in a digital world. OpenTelemetry is one way to do that. Standardising how telemetry data is generated and collected using open-source (and therefore inexpensive) agents addresses one of the primary sources of data distortion: IT budget.
When organisations ensure they can collect telemetry data from every system, not just a select few, they eliminate a significant source of data bias. That’s why edge computing solutions should embed telemetry generation in the platform itself, so it’s available anytime, anywhere.
Take Advantage Of Service Offerings
Data-Lake-as-a-Service can also address data distortions caused by limiting volume and storage costs over time. When organisations offload scale and capacity, they can ingest more telemetry. This makes it easier to uncover anomalies and patterns affecting performance.
In the past five years, the range of XaaS services has grown enormously – and companies are embracing it: In a study on XaaS, Deloitte found that almost 50 percent of companies will spend at least half of their IT expenditure on XaaS this year.
The distortion of data caused by visualisation can also be avoided. Rather than just showing snapshots of data points over time, Insights provide actionable guidance based on patterns and relationships found in the data. For example, if you see that the performance of an application is currently not optimal, you get a concrete suggestion to improve it at the same time. Ideally in the form of a clickable button, which then triggers the appropriate action.
Additionally, telemetry data should be analysed in the context of an overall user flow. It is then possible to understand when a failed component requires attention and when it does not. For example, moving from a model that relies on interpreting multiple visualisations to use rich insights can remove some biases associated with visualisation.
Companies should evaluate all data comprehensively for a reliable overall picture. For example, they can use OpenTelemetry to monitor their digital assets’ performance, availability, and security. The standardised generation and collection of telemetry data using open-source agents or functions integrated into the platform are easy on the IT budget. Data-Lake-as-a-Service can avoid unnecessary savings in storage. In addition, the analysis tool should offer concrete instructions for action instead of pure time series to gain real insights.