# How to check for correlation when I have text?

I have a list of customers. Customers have between 1 - 20 "Jobs" that flow through a front end engineering effort to get details specified. The jobs are all similar but different in their details. We know that each customer has their own kind of profile in terms of how long they take to get answers back and to perform approvals. I have statistics about turn times of jobs for those customers:

Customer        Days
________        ____
Cust A, job 1   17
Cust A, Job 2   13
Cust B, Job 1   29
Cust B, Job 2   18
...
Cust Z, Job 999 32


I need to identify what customers (on average) have jobs that take too long. I understand how to determine a correlation when I am comparing two numbers like temperature (degrees F) and speed (MPH). I don't understand how to determine this when one of the variables is text (customer name).

How do I determine the customers who's jobs take longer on average?

• I'm afraid you need to give us more details. What are those "jobs"? What does mean that they take too long? How is it related to correlation? – Tim Oct 9 '19 at 4:54
• This doesn't have anything to do with correlation, you just wonder if the distribution of survival times differs by some combinations of person & task. Does everybody have the same number of each kind of job? How many people are there? Do the numbers of jobs differ? Do you have any censoring? How much data do you have in total? – gung - Reinstate Monica Oct 10 '19 at 18:04
• correlation measures linear relationships of the form: if x gets larger then y gets larger. As "getting larger" does not make sense for categorical data, correlation is not the appropriate method in this case. – Sebastian Oct 10 '19 at 18:10
• @gung - I have a couple hundred "customer" and each customer has from 1 - 30 jobs. – Eric Snyder Oct 10 '19 at 21:06
• @Sebastian - What would be the appropriate measure? – Eric Snyder Oct 10 '19 at 21:06