=== How many reprompts===
**// Give callers two to three attempts at each question //**\\
The typical current practice is to give callers two to three attempts at a dialog step -- in other words, to offer one to two progressively more directed reprompts ([[references#rolandi2004a|Rolandi, 2004a]]). 
There is no published research that we know of on this practice, but it seems reasonable.

You don't want to move away from the dialog step based on the first unsuccessful attempt, because that would lose too many callers. On the other hand, you don't want to require too many attempts or you will frustrate callers who are experiencing significant trouble, and may well lose them if they hang up.

Note that most IVR architectures require you to set a certain number of no match reprompts and a certain number of no input reprompts. This means by combining them, the poor caller could be stuck in the same state for a while. In practice, the two are rarely missed. People are either trying (triggering no matches) or unsure of what to do or distracted (triggering no inputs), but rarely both.

**// In certain circumstances, fewer reprompts are appropriate //**\\
The appropriate number of retries is also influenced by what the failure action is. If you're simply asking the caller if they want to information repeated, a single reprompt is probably all that's needed. On the failure, acknowledge somehow that you didn't get what they said, but let's just continue onward.

There are some instances where you might not want to reprompt at all. These are very specific situations where a no match tells you something. Consider this prompt:

    * //Last question, your destination. You can say Las Vegas, New York, Florida, the Caribbean, Hawaii, Mexico, or say somewhere else. //

In this case, the intent is to pull off certain destinations. Suppose the caller says "California" instead of "somewhere else." That's a no match, but the effect is the same.

==== References ====
Rolandi, W. (2004a). Improving customer service with speech. Speech Technology, 9(5), 14.