CSAT lag and the earlier thread

Surveys are polite. Tickets are not. A utilities desk in Leeds watched CSAT dip eleven days after a billing copy change. Leadership wanted a workshop on empathy. The threads from day two already showed customers quoting the new sentence and asking whether the debit would still leave. That is not an empathy gap. That is a lag between the event and the instrument you worship.

We timestamp three things: the policy or copy change, the first cluster of threads that name it, and the survey that finally notices. The middle stamp is the one customer support signal analytics cares about. The last stamp is for people who only attend the monthly pack.

Early signals are often ugly. They live in half-finished chats and in emails that never received a category. If your QA programme only samples closed, five-star-eligible contacts, you have built a museum. The lag will keep surprising you because you keep refusing the room where it happens.

A limitation we state in class: if your survey vendor cannot join on ticket id, you will estimate the lag with a calendar, not a join. That is still better than pretending the dip “came out of nowhere.” It did not. The thread was already talking.