A data-driven approach to optimizing IT services

BY Kristof Kloeckner, CTO and General Manager, Technology, Automation & Innovation, IBM Global Technology Services

With nearly all industries undergoing a digital transformation, using information technology for competitive advantage is quickly becoming a critical success factor for enterprises. At the same time, as business cycles shorten and IT environments are becoming more complex, managing these systems for business advantage is increasingly difficult.

The automation of IT processes and tasks promises to help minimize human errors and deliver more consistent services.
However, such projects often have limited impact unless they become part of a continual improvement effort that applies deep analytics and cognitive technologies to the vast amount of operational data an enterprise or service provider collects.

We have now reached a point in which operational data can fuel a cognitive feedback loop that steers automation and augments human decision making. While our point of view has been formed by a multi-year effort of applying analytics and cognitive technologies within our automation initiative across the IT environments of our outsourced accounts, we strongly feel that the same approach can be applied by a CIO organization acting as an internal service provider. Here are a few recommended principles:

1. Create a data lake: This holds your operational data, like problem tickets, service requests, configuration data, change records, root cause analysis, resource consumption history, etc. It will provide the basis for discovering patterns through correlation of data from different sources that could not otherwise be detected. For instance, collecting historical data across hundreds of thousands of servers can allow you to predict (and preempt) server outages. Similarly, analyzing thousands of change records allows you to more accurately assess the risk of changes, which often is classified incorrectly by IT personnel.
2. Establish comprehensive monitoring and event management, feeding into the data lake: While implementing APIs for the data lake is foundational to growing content from different sources, the main conduit is the event management system. It is very important to reduce the proportion of events that turn into incidents that require attention.
   By fine tuning event correlation mechanisms, nearly 90% of events can be filtered out.
3. Let problem tickets tell you what to automate and to identify root causes of problems: Typically, a Pareto distribution applies, most tickets fall into a small number of categories, and thus automation becomes economically viable for about 80 percent of incidents. Additionally, applying machine learning technologies can automatically identify non-actionable tickets, and the use of text analysis can identify common root causes. Together, these technologies can enable full automatic resolution or auto-assist resolution of the majority of tickets flowing through the incident management automation system. For auto-resolved tickets, mean time to recovery can improve by up to 90 percent.
4. Establish a continuous feedback loop: Every action leaves a trace in the data lake, which should be routinely analyzed for improvement opportunities. It is important to make this an ongoing iterative process. For instance, as a service-provider you can continually assess the effectiveness of automation and compare across accounts to bring every client to best of breed.
5. Make data and insights visible and create a data-driven culture: Through dashboards and self-service analytics, you can make data the foundation of any conversation between stakeholders. These insights dashboards increasingly have a business perspective. For instance, you can group information pertaining to IT resources for the business service they support, so that IT issues are made visible in a business context. You can use dashboards to create transparency, and to drive a ‘get to green’ culture.
6. Establish a data science competency: In today’s data-driven era, the data scientist has a distinct profession and career path that needs nurturing. Consider hiring a Chief Data Officer and establishing a center of competency and a community for data scientists.
7. Take a holistic approach: Adopting a data-driven continual improvement approach has a profound impact on organization, process and skills. You have to address all aspects to make a data-driven transformation sustainable.

All the insights gained through a data-driven continual improvement process will feed and enhance your knowledge bases.
This links the data lifecycle to a knowledge lifecycle based on systematic curation of expert knowledge which you can leverage into assistant and advisor tools for all stages of the service delivery lifecycle. For example, you can use text analysis and concept extraction technologies to map client requirements to a client solution catalog, with the aim to speed up the process of solution co-creation between your clients and your solution managers.

At IBM, we make extensive use of Watson technologies to optimize IT services and have packaged our technologies into a platform, the “IBM Services Platform with Watson”. We have introduced elements of the platform into 800 of our clients so far, and are also planning to make it available in the context of all of our managed services.

Enterprises today are faced with rapidly evolving business environments, and IT services need to continually improve in order to support business needs. Operational data are valuable enterprise assets, and taking a disciplined data-driving approach to gaining insights from operational data is critical for success.

——————————————————————————————————————————-

As CTO of IBM Global Technology Services, Kristof Kloeckner develops the technology platform for IT service delivery, based on automation, analytics and cognitive technologies.