Dynatrace Helped Us Optimize the Self-service for My O2

September 19, 2016

For O2, a key player in the Czech telecommunications market, online is the primary channel for reaching end customers. These customers, now more demanding than ever due to the continuous efforts of competitors to attract potentially dissatisfied users, expect high-quality services and the manner in which they are delivered. Thus, the smooth operation of the online self-service portal My O2, where customers can check the status of their services including billing, is a fundamental prerequisite for success.

We discussed the contribution of the Dynatrace application monitoring tool, deployed since 2014, to ensuring the smooth operation of the online self-service platform with Petr Venera, the head of Online Systems operations at PPF IT Services.

When deciding to deploy an APM for monitoring the online self-service, what were your primary expectations?

We faced a communication misalignment between the IT operations department and the marketing and business departments. There were frequent discussions about what exactly constituted an outage, what did not, and why information about ongoing issues was not reaching the customer care department in time.

A common, but not ideal, scenario is that in an attempt to gain a perfect understanding of the system’s status, each department acquires its own system monitoring tools, attempting to address critical situations independently. This approach often leads to departments presenting and comparing inconsistent data from different monitoring systems, which may not always be compatible.

Instead of agreeing on the nature of the problem, discussions shift to arguing over data compatibility and comparing different results. A typical example is when the business monitoring tool reports an outage. The business team then approaches the IT department, which, after checking their logs, claims it was a monitoring tool failure, and so the cycle continues, wasting time and increasing animosity between departments.

Our main expectation was to acquire a tool that would allow us to view the same data. Another requirement was to have a unified monitoring system where every alert is visible to IT, which can immediately start addressing the problem, and at the same time, the business and customer care departments are informed and can respond appropriately.

You ended up choosing a tool that offers more than just monitoring. What led to this decision?

When we began exploring the APM software market, we compared various options and realized it would be a missed opportunity to settle for a tool that offered only monitoring. For the My O2 self-service, we saw the potential to find a tool that could directly analyze what had happened.

This means directing us to the problem without needing to rally database experts, operating system administrators, and application managers. Without Dynatrace, we could quickly detect, for instance, that an application was down but not why it was happening. The analysis alone of why a problem occurred could occupy four to five people for many hours.

In contrast, Dynatrace can quickly focus on this phase, so we immediately know where the problem lies and where to start looking. We didn’t initially expect this functionality from the APM tool we wanted, but it proved to be a significant benefit during the final selection process. Was quick analysis and problem direction the deciding factor for implementation? Let’s say it was an optimal mix of several decision factors.

Price was also an important consideration. Of the three tools we were considering, Dynatrace offered the most benefits, and its cost was not the highest. After we had more or less decided in IT, we still needed to justify our choice to the decision-making committee with a compelling use case. This is somewhat difficult without data from real operations, but we managed to do it in the end.

What were the main benefits of deployment?

The business gained much greater control over the functionality of the My O2 application, and problem analysis became simpler for us. With hindsight, I see another significant benefit for developers. Areas that were previously difficult to investigate are now easily accessible through Dynatrace dashboards.

For example, if SQL database queries significantly slow down, we can now simply look at the dashboard, and the whole problem is detectable with just a couple of clicks. Previously, the same task involved going through a lot of code. Just analyzing the logs entailed selecting all calls to the database, sorting them by response time, averaging those, and only then drawing some conclusion.

Did any problems arise during the implementation itself?

Surprisingly, the entire implementation was smooth. One day, a consultant from Adastra showed up, and the next day, we were already looking at the first data. What we originally hadn’t anticipated was the need to acquire new server hardware. However, this requirement was common to all the solutions we considered. No other significant problems occurred.

In the post-implementation phase, of course, it took some time to understand exactly what we wanted to measure and where else we could effectively utilize Dynatrace’s potential. We also spent some time getting everyone, especially those in operations and development, accustomed to actively solving problems with Dynatrace. These are more psychological nuances than a problem with the software itself.

Are you using Dynatrace today exactly as you initially envisioned?

Yes and no. The original plan was that by saving time on analyzing application errors, we would save time and resources on the staff involved in analysis and subsequent resolution. However, in practice, it doesn’t work like that; Dynatrace actually reveals new and critical situations that we would otherwise discover much later, usually during a significant failure in front of the customer.

Therefore, the workload is ultimately the same. The difference is that we can address individual failures more quickly and often much earlier before they become real problems. It might not be immediately apparent, but this leads to a significant improvement in the quality of the entire application and thus the service for the end customer.