Home > Columns > CRM Columns

How AI Gets to the Root of the Customer Feedback Loop

Contributed article by Chris Martinez and Kevin Yang, co-CEOs, Idiomatic 

There’s something so frustrating about taking the time to provide feedback to a brand and feeling like you’ve fallen on deaf ears. This is especially true if this is an issue that you know is widespread and not specific to your interaction with the brand. However, with the number of ways that customers can now provide input to brands through—support inquiries, social media, app/product reviews, surveys, forums, and more—brands are often left drinking from a feedback firehose. Brands can barely process and respond to all the feedback, let alone understand the bigger picture and identify trends. The result? The loudest customers get the responses (you know, the ones that eventually start tweeting, leaving reviews on every forum imaginable,  calling several times a day, and even going so far as to threaten lawsuits), while slow-burn issues impacting many customers go unaddressed. This leaves the majority of their customers feeling like brands aren’t listening to or acting on their feedback, sparking much larger issues for brands to address when the fire eventually starts.

It’s not that brands don’t want to create tight customer-feedback-to-product loops or customer-feedback-to-operations loops. It’s that they can’t translate the millions of customer feedback data points from various digital sources into easily understandable insights. Most customer service processes still rely on manual analysis, keyword searches, and survey-centric customer feedback initiatives, which simply can’t give them the actionable insights they need to properly manage the continued customer feedback surge.

Top digital brands like Instacart, FabFitFun, Pinterest and more are tapping artificial intelligence (AI) to help them stay ahead of the customer curve to improve satisfaction, cut support costs/complaints/issues and get more from their customer data. How can you do the same?

Bridge the Gap Between How Customers Speak and How You Label/Describe Their Issues  

Historically, companies have had their support agents label cases. Companies often make these labels very high level to reduce the options for agents and keep things simple. The problem with simple labeling solutions is that accuracy comes at the cost of specificity. AI can make labeling both specific and accurate. 

Companies can act confidently to address customer feedback when they have a complete and real-time view of what customers are actually saying. Machine learning and AI tools (like Idiomatic) can be tapped to create a custom set of labels for unique data sets and calibrate sentiment analysis. Knowing there is negative sentiment associated with the “Login” category is not actionable; however, knowing that 100 users tried to reset their password yesterday and didn’t receive the email tells a much more specific story, with a clear problem to solve. With AI doing the labeling, labels become much more specific without the heavy manual lift of relying on agents to lead this initiative.

Transform Freeform Qualitative Feedback Into Quantitative Data  

Another way AI can be tapped to wring value out of customer data is by looking at all interactions from the ground up to assess the root cause of each individual interaction and identify trends and questions that should be asked, as well as supplementing product usage data with anecdotal behavioral information. This could mean looking at how often features are mentioned, and how customers are describing their problems. One example of this is looking at tweets to identify sentiment timed with a product launch. On a larger scale, FabFitFun needed a scalable, data-driven way to translate the voice of the customer cross-functionally. Idiomatic analyzed and categorized FabFitFun’s text survey responses and support contacts in real-time–leading to a 250 percent increase in product satisfaction.

Combat Call Center Connection Chaos   

Nothing is worse than spending time with a customer service agent and then realizing you have to be transferred to someone else to help with your questions. When users self-select ticket categorizations, it can lack the precision needed to connect them with specialized agents. Now picture you have four million monthly customer support contacts and users self-selecting their ticket category. That’s exactly what Instacart was up against.

Instacart integrated Idiomatic with Zendesk to categorize support contacts in real-time. Using AI-driven, real-time ticket categorization can optimize efficiency, improve the support experience and save costs. Instacart streamlined support workflows with ticket routing, agent specialization, and spike notifications and was able to uncover nuanced customer pain points in the process. By routing tickets to specialized agents, the company was also able to reduce support time and save $445k in annual support costs by proactively addressing specific customer pain points and driving down contact volume.

These are just a few of the ways that AI can help bring focus, free up agents’ time for more strategic uses, automate repetitive tasks and enable quicker responses to customers. These strategies also help companies move away from gut instincts and fragmented perspectives to draw conclusions from real customer conversations and data. Instead of focusing on reply times and customer satisfaction, empowering customer support teams with AI tools and focusing on metrics like Net Promoter Score (NPS) and reducing customer churn can help focus on the right metrics and outcomes, a win for the agent, the customer and the company’s bottom line. When the root cause of a customer issue is quickly surfaced and addressed, we all win.