Banking
Bank customer satisfaction surveys
The hard part of a bank survey is that the same customer meets you in three places: the branch, the app and the phone. Those are three different experiences, and a single satisfaction score averages them into something nobody can act on. Ask which channel they used before anything else, and split the results by it. Measure trust as its own question: in banking it is what predicts who leaves. And never ask for an account or card number.
The templates
Branch, app and phone separately
Before any rating, ask which channel the customer used this time, and never publish the overall score without breaking it down by channel. An average of 8.2 can hide an impeccable branch and an app that fails at payment, and the team that has to fix it will not know you are talking to them. Each channel also has its own questions: in the branch, the wait and the staff; in the app, speed and errors.
Ask about trust separately
Satisfaction and trust are not the same thing, and in banking it is trust that predicts who moves their money elsewhere. Someone can be happy with the app and still not trust you with a mortgage. Ask it outright, in the same words every time so you can track the series: whether they trust how you look after their money, and whether they believe you would resolve a problem in their favour. Read that series after every fee change.
Data you must never ask for
No survey needs an account number, a card number, a PIN or a balance, and people will type them if you ask, which is exactly the habit fraud later exploits. If you need to know who answered, identify them through the link, never through data the customer types. The same restraint applies to selling: asking about products they do not hold turns the survey into a pitch, and it shows.