Know-How Exchange : Research/Metrics

I am not sure I understand the question. As a rule of thumb for consumer interviews - it takes as long to find a willing respondent as it takes to administer the survey. So if the survey itself takes 10 minutes and it takes 10 minutes to find a new respondent you can complete 3 surveys an hour. This is not a perfect formula and you will need to up the time to find a willing respondent if you are dealing with a low incidence audience.

Perhaps there is more to life than Call Time and Administration. I imagine there are visits to the water cooler, bathroom and discussions with the supervisor.

You need to break down all the activities undertaken and allocate the time spent to each.

Call time must be available as a system statistic... As must system log-in time. How bothered are you about tracking the other statistics?

WHY are you doing this exercise?

Is it to try to increase performance, or for disciplinary measures of some sort?

ChrisB

Dear Cathal.Kerr

We seem to be in danger of suggesting some much better ways of measuring things than your company currently appears to provide, but without answering your question in the first case!

I think that you are putting the cart before the horse here as there do not appear to be enough measurements already taken to establish that which you want to measure in the way you state it. It doesn’t mean that you can’t come up with some useful figures through a survey though!

Firstly, 50 people, 40 hour each give you what you’ve termed a population of 2000 hours. You say confidence is 95%. I guess from your clarification that you mean Confidence Interval is 95%. But measuring what? The confidence interval simply states what % of a measured parameter will fall between two values. You can only know that once you’ve taken some measurements and set an upper and a lower limit!

Can I therefore assume then that the measured parameter is Time on Admin (Ta) that you wish to measure to give a confidence interval of 95%?

Should you measure in hours or minutes? Use Minutes, because you want to have uniform units of measurements and measuring in decimals of an hour is a pain and is hard to translate back into real life.

To take the form of your example of what you would expect an answer to look like, you would want to see something like, “95% of the measured user admin times (Ta) to include 90 minutes + / - 10 minutes.” i.e. 80 is less than or equal to Ta which is greater than or equal to 100

On the face of it, you might assume that by taking a sample of 5% of the hours worked in a week that you could have the level of confidence you are looking for, but this is a no-goer as far as statistical methods are concerned.

As you havn’t done any measurements yet, you have no idea of the range of values of Ta, thus you can’t know the mean or the standard deviation. That infers that you can’t calculate a confidence level.

If you set a confidence level you want, you can’t determine the sample size you need because you don’t know what the distribution of the values of Ta look like, so you have to start to take some measurements before you can start to approximate to the figures you’re seeking.

Sorry not to be able to give you a specific answer, but there isn’t one based on the information you have given. On the face of it, this seems to be a very difficult way of sorting out something which should be really very easy to do, if the parameters you need are recorded in the first place. That’s where the other postings are really useful – why set out to do something which is going to be a) difficult and very time consuming and b) Initially, statistically meaningless.

As I can’t print equations here, if you want to run through the problem in more detail, please email me directly and I can try to run through the base figures you need and the formulae you need to plug them into.
Regards

Steve

Your measurements should be in minutes. Start having your call center agents log in and out for admin (something your phone administrator should be able to assign to several phone pad keys) and then you can measure time in your CMS reports from the phone system. Since the log will be by call agent you can then analyse the agent, shift, day, etc.

Once you have the measuring tool, you can then tackle the question of population mean and standard deviation and can manage the call center accordingly.