Asking whether it was good.
WHAT THE COMMON APPROACHES ARE
A rating after each interaction A periodic relationship survey A likelihood-to-recommend question
WHAT PER-INTERACTION RATINGS MEASURE
That interaction, not the relationship.
WHY THAT DISTINCTION MATTERS
Someone can rate support highly and still cancel.
WHAT TO ASK
One question, immediately.
WHY ONE
Response rates collapse with length.
WHAT TO ADD
An optional comment.
WHY
The comments contain everything useful; the score is only a filter.
WHAT BIAS TO EXPECT
Strong opinions respond; the indifferent do not.
WHAT THAT MEANS
Treat the score as a signal, not a measurement.
WHAT TO DO WITH EVERY NEGATIVE RESPONSE
Read it, and follow up.
WHY FOLLOW UP
It recovers customers, and it explains the score.
WHAT TO AVOID
Tying individual pay to scores.
WHY
It encourages asking for good ratings, which corrupts the data.
WHAT TO COMPARE
Trends over time, and across categories.
WHAT LOW SCORES IN ONE CATEGORY INDICATE
A product problem, usually, not a support problem.
WHAT TO REPORT
The trend, with the reasons from comments.
