Have we gotten observability backwards?

The last few years have seen digital transformation dramatically increase the sprawl and complexity of enterprise IT environments. Today, the average employee will likely access a dozen applications before lunchtime, both in the cloud and on-prem. This increased complexity has simultaneously created greater interdependences between applications, while also making visibility much trickier for IT teams. 

The upshot of this is a world where there are more applications in use, which are less able to be monitored, and causing greater frustration when they experience issues. This is not a recipe for success.

To mitigate the problem, enterprises have invested heavily in ‘observability’ solutions, largely based on application performance monitoring (APM) tools. These tools have vast amounts of data about every aspect of the technology backend (application code, databases, and server/network/cloud infrastructure) executing a given application. Yet for all this investment, businesses are still operating with gaping blind spots across their application stacks. So, is it time to ask the question; have we gotten observability backwards?

Hitting the limits of APM

There is no doubt that high cardinality APM tools and their extension to server and network technologies constitute an important and valuable part of the way modern enterprises operate their technology stack, and further digital transformation initiatives will only enhance that importance for critical, custom web applications. However, despite this, there are two key limitations to APM that mean they cannot be the silver bullet that businesses are looking for to monitor full-stack employee application experience.

The first is cost. While strategically deploying APM across select critical applications is a smart strategic move, doing so across the entire IT environment and every application is prohibitively expensive and technically complex. The second -- and perhaps more important -- is that while APMs are fine working in conjunction with custom web applications, they are blind when it comes to large-scale commercial SaaS applications. Deploying APM means altering the code, which businesses cannot do with commercial, packaged web applications like Zoom, Slack, SAP, Workday -- yet the need to monitor experienced performance is still there.

This is even more true for web-based communication and collaboration tools because, they multiply the effect of individual poor experience. If the presenter on a video conference call is having to deal with a buggy and slow app, those glitches create a problem for every single person on the call. 

Taken together, these factors present a serious conundrum. At the same time as application observability is becoming an increasingly important priority for businesses, the tools designed to provide it are only able to work on a partial basis -- from the application code backwards -- with no visibility to the employee experience.

What are we measuring?

Resolving this issue requires businesses to take a step back, pause, and reassess what they need to be measuring. Currently, observability is built on measuring the performance of applications and their supporting infrastructure, but this isn’t really the metric that is of most concern – outside the technical support teams. Measuring the performance of the application code and infrastructure is a poor proxy for understanding the end user experience. While APMs will show statistics such as the detailed performance of application code/steps and intersections with infrastructure metrics like CPU usage or network error rates and throughput, these metrics at best offer a partial understanding of the real employee experience.

By shifting the focus to the quality of the user experience, organizations gain insight into the most important element of any application, i.e. how productive it’s allowing an employee to be. After all, even if there may be problems on the backend, if these aren’t interfering with the end user, then they are of secondary importance. This allows overstretched IT teams to more effectively prioritize their workload to ensure optimal productivity at all times.

Going beyond the application itself

Moving beyond application performance to user experience has another key benefit. The way in which each user accesses a given application varies tremendously, depending on the version of the browser or operating system they have, the configuration they are using, or the quality of network connectivity. Application providers cannot hope to test against every conceivable set of conditions and so the onus is on companies to constantly measure and adapt.

In an environment where hundreds of (constantly updated) applications are being used in thousands of different combinations, the ability to have such granular detail and easily prioritize the most pressing issues for any given subsets of users is invaluable to IT teams.

Therefore, businesses need to properly equip their IT teams with the ability to consolidate experiential visibility over all the applications in a single space and sort rank applications according to experienced performance, as well as drilling down to see which users are having issues and what the common factors might be.

There’s no such thing as partial observability

Application performance and reliability is essential to almost all digital experiences. However, as important as they are, they are ultimately only an enabler to unlock greater human productivity.

So, while application performance is vital to a good user experience, it is the experience, not just the performance of the application, that businesses should be looking to measure. It’s a simple addition but one that can fundamentally change the entire way that organizations understand observability for the better.

David Wagner, DEX Strategist at Nexthink.

Image Credit: Ffennema / Dreamstime.com