When it comes to Customer Intelligence, should we pay more attention to behavior or sentiment?
In our recent webinar about unlocking hidden revenue through the use of data science, the team talked about how data can inform us whether a customer is “happy.” Thoughtfully, one of the attendees asked the question:
When it comes to Customer Intelligence, does it matter more what a customer says, or what they do?
This question is so important. It acknowledges that:
Qualitative data isn’t perfect. People speak from a place of bias, have difficulty sharing difficult truths, and/or might not be representative of their company.
Quantitative data isn’t perfect. People can squish numbers into whatever narrative they wish to tell, and numbers don’t always reflect organizational changes or relationships.
So what's the answer?
Customer Intelligence means understanding customers holistically: It requires gathering and keeping an eye on both types of data, while letting historical outcomes tell us how much impact each actually has on customer health.
It's imperative to look at every type of quantitative and qualitative data that your organization can gather, from NPS to emails to support ticket sentiment, and from product usage to interaction frequency. By reviewing the impact each of these specific behaviors and sentiments have had on churn and upsell historically, you could theoretically determine how much weight such metrics should carry in predicting customers’ future health.
While neither quantitative nor qualitative data alone tells the whole story, a holistic view of both can deliver true Customer Intelligence.
But then what?
Of course, then the challenge becomes: How do you know which data sources are relevant, and how do you access them? And once you do, how do you analyze metrics to determine which data matters and how much? It's a big lift!
This is where involve.ai’s data science and machine learning teams come in.
Data science teams help advise organizations around which data sources host valuable information, help clean and validate data, and work directly with data administrators to pull it into a unified, customized view.
Then our machine learning models analyze that data to determine:
How to segment customers in a way that compares apples to apples.
Normalization for each and every metric and natural language processing for each and every word.
The weight each data category should have in health score calculations (based on historical impact on churn).
By combining data science and machine learning, organizations are starting to let their qualitative and quantitative data show them how to address churn risk, upsell opportunity, and referral sentiment. In other words, these two cutting edge fields (machine learning and data science), in conjunction with these two broad types of imperfect data (qualitative and quantitative), are helping us uncover hidden revenue with customer data.