Why is NPS not dead yet?

For years, support organizations have used NPS as a post-transaction survey–in a very misguided manner, in my not-so-humble opinion. Why misguided? Because the context (post-case) is wrong and the audience (technical contacts) is also wrong. And we now know that NPS surveys are just not that predictive.

Scott Clark recently wrote a great article exploring why NPS has not disappeared yet, despite Gartner predicting it would. Here’s my take from the support perspective.

• NPS is deeply ingrained. Many organizations have used it for years and are reluctant to move to new metrics for which they don’t have a baseline.
• We don’t have a single, clearly better replacement yet.
• If we must rely on surveys, CSAT is more appropriate to the context and the audience. Technical contacts are absolutely able to tell you whether the interaction went well, and if not they can give you a perfect opportunity to get it right. Service recovery is a good reason to have a CSAT survey, even if return rates are low.
• The big limitation of surveys is the response rate. If you only hear back from a tiny portion of customers (a small portion of customers who log cases, which itself is a small sample of the overall customer base), you are not getting a general view of what customers are thinking.
• The future is to use the abundant data we have access to (case notes, product usage, digital interactions across the entire organization, not just support, public forums, etc.) and deploy AI tools such as sentiment analysis and pattern analysis to figure out what customers are doing with the product, how they drive value from it, and whether they are likely to churn, stay, or expand their relationship with us.

Short-term, I recommend discontinuing post-case NPS surveys entirely and using CSAT instead. To boost response rates, ask a single mandatory question (rate the quality of the support you received) and respond very quickly to any negative surveys.

Bonus thought #1: beware of averages. Most organizations average CSAT ratings so if you have one completely furious customer (with a rating of 0/10) and nine happy customers (with ratings of 9/10), you would get a respectable overall average of 8.1. But the respectable average hides that very unhappy customer, which may well be at risk. Treat very low ratings as your warning system instead of hiding them.

Bonus thought #2: look into sentiment analysis and other AI tools. They are the future.

How do you capture customer satisfaction? Please share in the comments.