Implementing and monitoring metrics is critical to measuring conversational artificial intelligence (AI) performance.
Self-service resolution rate, costs saved and positive customer feedback are some (but not all) of the top metrics CX professionals can look at to determine whether their conversational AI technology is connecting with customers.However, these key metrics heavily depend on the type of conversational AI application in use.
At a high level, there are three categories of metrics that organizations should consider:
- Business-related metrics: Key performance indicators (KPIs) that focus on overarching business goals and conversational AI-related objectives.
- User experience metrics: Focus on an experience that is useful, engaging and enjoyable, spurring users to return and/or recommend the product to others.
- Technical metrics: Ensure the conversational AI product works and adheres to the requirements for performance or latency.
Build a Team to Craft a Metrics Roadmap
"To build a comprehensive list of metrics, key stakeholders across the organization need to come together," said Inge de Bleeker, principal UX & conversational AI consultant, founder at outriderUX. "In terms of how to determine themetrics, a workshop or series of workshops that bring this cross-organizational team together is a good approach."
For each of the three categories above, the individuals or specific teams in an organization will know which metrics they need to measure based on their goals.
"However, it is only when all teams come together that the comprehensive view on the set of metrics can come together," de Bleeker added. "In larger companies especially, it can be tricky to create cross-organizational workingrelationships as silos can get in the way. Building a tiger team and raising awareness around the endeavor can be helpful."
Are you focused on driving customer retention or avoiding attrition? Are you striving for a better net promotor score (NPS) but don’t know how to influence it?
"Select metrics that are likely indicators of your business objectives," Brown said. "For example, if you aim to improve customer retention, make that your metric."
Next, she advised, use conversational AI to study what customers are complaining about and make changes to address those complaints. "If you’re successful, you should see your customer retention rate increase.”
Related Article: How Will Conversational AI Transform Customer Experience?
Align Metrics to Business Goals
To define a comprehensive set of metrics that can realistically be obtained, one needs input from all corners of the organization, according to de Bleeker.
She pointed out that business and executive teams have certain goals around products — such as monetization — set by investors and board members. Product managers, on the other hand, understand the product vision and can determine howto measure things like consumption and conversion.
"UX professionals know which user experience metrics to gather so that user-driven feedback is gathered and distilled into metrics as well as getting self-reported user ratings," de Bleeker noted. "Engineers and QA (qualityassurance) can help gather performance, latency and other measures of a more technical nature to ensure that the product is performing as expected in production."
Brown agreed with the notion that all business functions have a role to play in customer experience.
"That said,” she added, “the most common stakeholders we see are operations leaders, as well marketing and CX leaders. All need to come together to understand key learnings from conversational AI and create action plans that improvecustomer experience, and therefore the bottom line."
She cited the important concept of setting metrics that align with business objectives to measure the overall success of a conversational AI strategy. "Conversational data is meant to drive action. Monitoring results is the only way toknow whether or not the actions you’re taking are having the intended impact."
De Bleeker claimed it’s important to apply these metrics when assessing the performance of conversational AI for a number of reasons.
"Before the organization starts using metrics, they decide what to measure and determine the threshold for success for each measure," she explained. "Then they can assess the product throughout the development phase and onwards andtrack the progress against these thresholds to create a clear way to measure success."
Once a product is live, she said, metrics allow for tracking improvements over time and over different versions. "If applicable, a competitive analysis using the metrics can be useful to benchmark your product against that of yourcompetitor.”
Metrics Grow in Importance as Conversational AI Evolves
De Bleeker pointed out that immature products and product spaces may need more detailed metrics compared to mature spaces.
"Take the example of TTS (text-to-speech) voices in voice assistants," she said. "While TTS voices have come a long way and sound far better now than they did years or decades ago, they are still not at the point where we are certainthat the user will find the voice assistant a great experience."
Brown added that conversational AI doesn’t just capture the customer side of the conversation — it also provides valuable insight into how your contact center agents are interacting with customers. "Tracking quality and compliance trendrates shines a light on agent performance and training opportunities.”
She explained that as conversational AI becomes more advanced, it will be more and more predictive of business results and can be leveraged to solve many business problems across the enterprise. "For example, we expect to see greaterdepth and understanding of the nuances of customer emotions and how those emotions lead to varying business results.”