The Case for Automating Technical Support

October 3, 2018 | Ajay SIngh
The Case for Automating Technical Support


Customer experience is the new sales


In a bygone era, focus on new customer acquisition far outstripped attention to the overall experience of existing customers. Today, hardly anyone needs to be convinced about the importance of a great customer experience. The proliferation of recurring revenue models requires management to obsess over customer retention and churn metrics. Even companies with traditional sales models recognize the ease and power with which customer “word of mouth” can influence future customer acquisition.


Technical Support: A keystone of the customer experience


Although digital transformation initiatives aim to rethink every aspect of the customer experience, a big opportunity remains that arguably dominates a customer’s perception of their overall experience. Of course, we expect every product to be easy to research, buy, onboard and use. But no matter how perfect each of these are, one thing can significantly enhance or completely sour our perception – how technical support issues are dealt with when something goes wrong. Intuitively we know what great technical support looks like.


Entering the Golden Era of Technical Support


Companies that build products or develop applications clearly recognize the value and importance of technical support. The sheer volume of support job openings on LinkedIn compared with other roles is a telling sign. US based software companies list almost 3 times as many openings for support roles on LinkedIn as they do for developer roles. Many organizations are experimenting with new role definitions and collaborative (or swarming) support models. They are investing in tools such as helpdesk and ticket management systems, monitoring software and chatbots and AI systems to better classify and match searches to knowledge repositories. They collect large volumes of machine data from their products, and are building internal tools, scripts, querying and analytics capabilities to leverage this data. All of these efforts aim to help customers solve their problems more quickly and with the least amount of frustration.


However, many challenges exist


At the risk of oversimplification, two of the biggest challenges in technical support are shared by every organization: the difficulty of quickly troubleshooting a brand new problem, and the surprising difficulty of quickly matching a problem’s symptoms to a known solution when it occurs again.


When a new technical problem is encountered, the answer can almost always be found by analyzing machine data (log files, configuration settings, stats, debug outputs). Typically, such analysis is manual and painfully slow due to the unstructured nature of the data and the lack of purpose-built tools in this space. Manual data inspection and regex-style text searches are commonly used, but these can be like looking for a needle in a haystack. One-off scripting to understand a problem can take hours or days and often requires ninja-like skills.


Once root cause is identified humans do their best to capture it in notes that others can use for future reference. Over time, organizations accumulate large repositories of knowledge base articles, case and bug notes, community forum articles and self-help guides. However, there is a fundamental problem: even when the answer exists somewhere, it can be slow to find a match,  which consumes unnecessary support hours and frustrates customers due to the delays (and often further aggravates them once they learn they’ve hit a known problem).  A big reason for this problem is variation in the way a problem is described. A given group of customers and support personnel describing the same issue often use different terms or focus on different symptoms, making it hard to match problems with solutions. NLP based AI tools and chatbots can only help in some cases.


The drudgery of slow brute force problem solving doesn’t just impact customer satisfaction while they wait for problems to be solved. Employee morale and retention also suffer because hours spent on tedious tasks crowds out more thoughtful analysis and impactful projects that could systematically enhance the product experience and customer satisfaction. Attrition is particularly painful given the sheer scale of the search for talent mentioned earlier.


The Solution – Machine Data Based Support Automation


There should be a much better way.


If machine data could be interpreted perfectly, quickly and automatically, we could achieve the dream of rapidly solving new problems and instantaneously matching symptoms to known root causes. This would eliminate the need for tedious tasks such as manual data inspection, crafting regexes and custom scripts. Deep understanding of the data would allow pattern recognition to learn normal behavior and highlight anomalous trends. A purpose-built interface could quickly guide a user to the root cause of a problem without someone manually sifting through reams of data. And the system would accurately capture problem signatures, enabling automatic detection of known issues forevermore.


Zebrium is making all of this possible today. We overcome the fundamental challenge of leveraging messy, unstructured machine data by using machine learning to automatically structure it. Once structured, automatic pattern recognition and intuitive workflows quickly guide a user to a problem’s root cause and capture the details as a signature.  This means complex problems can be solved in minutes, not hours and known issues can be automatically detected and resolved. And by doing this, we empower employees to spend far more time solving high-impact problems and driving product improvements – the perfect recipe for enhancing job satisfaction and retention.


Tags: support automation, customer experience