Satisfaction scores get used for coaching, and on any team with handoffs they are attributing the score to the wrong person often enough to change how people are managed. The survey fires when the conversation closes, and closing belongs to whoever happened to be holding it.
The attribution problem, in one example
A customer waits 6 hours on a first agent, gets escalated, and a second agent answers in 12 minutes. The customer scores the conversation 3 out of 5, which is a fair score for a 6-hour wait. The 3 lands on the second agent. Run that for a quarter and your escalation specialists sit at the bottom of the leaderboard, which is exactly backwards. Lucia Ferrari found her two most senior people there for eleven months before anybody questioned the number rather than the people.
Two things to change
Attribute the score to every agent who sent a customer-facing message on the thread, not just whoever closed it. And carry the conversation’s total wait time next to the score, so a 3 on a 6-hour thread reads differently from a 3 on a 20-minute one. Neither change needs new tooling. Both need somebody to stop treating the default export as the truth.
What CSAT is genuinely good for
It is a weak per-agent metric and a strong per-topic one. Group the scores by tag instead of by person and the picture becomes useful immediately: refunds score 4.4, integration setup scores 3.1, and now you know which help article to rewrite this month. Individual scores mostly measure how hard the questions were that happened to land in that person’s queue.
Read the response rate first
A 12 percent survey response rate is normal and a 4 percent one is a warning, because at 4 percent you are hearing almost exclusively from customers who were delighted or furious. Before comparing this quarter with last, check whether the response rate moved. A good share of the CSAT swings that get escalated to leadership are a change in who bothered to answer.
Leave a Reply