ML-Enabled Developer Forums Can Find Quick Solutions to Developer Queries

Traditional community forums and Internet search have been unable to solve the many questions of developers quickly or accurately. Robin Purohit, the co-founder and CEO of Peritus.ai, explains how machine learning tools can transform community forums to quickly and accurately address developer queries. 

A significant fallout of the COVID-19 pandemic is a rise in digital transformation investments, including increased B2B customer service spending. While many of the building blocks are available today, AI and machine learning approaches promise to radically improve customer service at a large scale with continuous improvement. As in any transformation, it’s crucial to implement the best approach to meet customers’ requirements. 

AI chatbots for customer support receive a great deal of hype and investment in industries ranging from retail to financial services to employee help desks. These have all shown early promise for automating high volume, simple questions typical in call centers. These approaches are enhancing and ultimately replacing call-handling systems and online chat. 

An important area that has suffered from neglect is support for engineers and developers driving IT innovation. These technical users have little patience for pre-wired chatbots without topical context and view opening a support ticket as the last resort. Instead, after an initial Google search, they typically look for insights from trusted experts on community forums, such as Stack Overflow or vendor-sponsored communities. When done well, community forums can create both user loyalty and deflect costly support cases.

Unfortunately, answers on forums are on a best-effort basis leading to unpredictable results. A recent bench market study of over 50 IT industry forums revealed the following:

• Only 31% of reported posts on forums are resolved;
• Just 19% of posts are resolved in less than 24 hours;
• A mere 14% of posts are resolved on the first response.

Learn More: Reinventing Modern Enterprises for a Post-COVID World

The Limitations of Internet Search

In a perfect world, Internet search would provide answers to all technical problems. Today, most of us search for answers using “keywords.” This method amounts to using a few carefully-selected words to retrieve a document or link of interest. This works well when the user is looking for specific facts and has the time and expertise to sift through results. It is less effective when the user is not sure about the information they are seeking. 

Also, search typically considers all public information as potential candidates for retrieval. This is not ideal when most of the data is either noise or irrelevant to the specific need and context of a technical problem. New machine learning search techniques can optimize search results for particular keywords, but they are not suited to technical questions beyond known facts. 

Learn More: How Technical Debt Can Negatively Impact Customer Experience

Why Community Forums Work

Technical people gravitate to public question-answering systems to get advice from trusted experts. This is today’s reality with 125 million monthly active users on Stack Overflow. Community forums allow technical people to post questions on forums organized to address specific domains. This is ideal for problems without a clear fact-based answer when the rate of technological change is outpacing available documentation and where expert human judgment is required.

Community experts who generate these answers are motivated by pride and public recognition, which improves their reputation in the IT industry. In the benchmark study referenced earlier, it became clear that the best communities significantly outperform the average by a factor of two:

• Over 60% of reported posts on forums are resolved;
• 42% of posts are resolved in less than 24 hours;
• 33% of posts are resolved on the first response.

A more striking contrast is that 72% of community members answer 80% of the questions on the best forums. In contrast, only 20% of community members address questions across all forums on average. A broadly engaged expert community capability is clearly the key to best-in-class performance. 

However, community experts have many demands on their time with no tools tailored to answer questions. This means they are left to rely on their knowledge base and problem-solving skills in their spare time. This can be particularly intimidating for high-potential community members who invest significant time to come up with credible solutions.

Learn More: 3 Ways IT Leaders Can Build Trust with Other Stakeholders

Use Machine Learning To Make IT Forums Best-in-Class

New machine learning techniques can rapidly enable IT forums to meet and beat best-in-class performance. Forum conversations and related content for specific topics provide a dataset to train ML tools to help community experts answer more questions more accurately. These tools can also expand the total number of active community experts by enhancing their confidence in answering questions. 

With the help of ML tools, community forums can increase the volume and speed of responses to new questions and the number of experts who can participate.  As a result, more people can turn to community forums to find solutions to technical problems, reducing the volume of support cases and increasing customer satisfaction for the specific products or technologies that the forum addresses.

Let us know your thoughts on LinkedInOpens a new window , TwitterOpens a new window , or FacebookOpens a new window . We would love to hear from you!