UCCE 8.5 - Calculating answer wait time from TCD records

[Mark Knepper]

Having difficulty making the numbers add up when comparing results to the Call_Type_Interval table results. From the TCD I am currently trying to add together NetQTime + RingTime + NetworkTime for the associated CallTypeID, CallTypeReportingDateTime, using the call leg that has a value for RouteID.   However, for each call in the interval, but I am sometimes off by 1 to 2 seconds.

I started down this path to answer the question -- When does the AnswerWaitTime timer start, as I have a 10 second greeting prior to my assignment of my CallType and I get AnswerWaitTimes for that CallType under 10 seconds?

As always appreciate the amassed wisdom to point me in the right direction as I am stumped.

Thanks,

Mark

[Jameson Gagnepain]

Mark,

As I said in the other thread:

All timers in Call Type data begin when the Call Type is applied. Times within the TCD are generally for the whole call up to that point. The "CTVRUTime" from the Call Type which is active before your 10-second message should provide an accurate count of the seconds callers are spending listening to that message. Compare those values to the 50-150 seconds that you are over by.

Some more to add:

NetQTime, RingTime, and NetworkTime are probably the right items to look at in TCD. Can you post some specific data to look at for this, or a sample query to diagnose issues with?

-Jameson

[Mark Knepper]

Thanks Jameson, I have attached my sample data that shows two ways of calculating in the TCD and a comparision to what "should be" according to the CTI table.

Thanks,

Mark

[Jameson Gagnepain · Accepted solution]

Mark,

Your first calculation is probably the way to go with this one.  Keep in mind that the AnsWaitTime from the CTI table is a total from a continuous counter, while the three statistics you are adding are storing discrete values in seconds. There will often be a variance of up to a second when you compare your TCD total with the AnsWaitTime due to this. Here's an example to explain:

Let's say you have a call with the following:

NetQTime of 4.499 seconds (stored as the integer 4)

RingTime of 4.499 seconds (stored as the integer 4)

NetworkTime of 4.499 seconds (stored as the integer 4)

AnsWaitTime of 13.497 seconds (stored as the integer 13)

Now, in this situation, NetQTime+RingTime+NetworkTime = 12 seconds, because you're adding the integers. You have no access to the value with milliseconds, so you can't get more accurate than that. This is an extreme example... In general, for a single call NetQTime+RingTime+NetworkTime should be within 1 second of the equivalent AnsWaitTime, and on aggregate things will tend to even out as you should have approximately the same number of stats being rounded up or down as your sample size increases.

Based on the above logic, Call #7 in your excel file is the main culprit to push me towards thinking your first method is more accurate - 64 is too far from 62. The second reason I would go with the first method is this: You're able to get the results for a call by looking at a single call leg. This should make the WHERE statement on your report simpler, and the query as a whole more efficient.

-Jameson