1
$\begingroup$

I'm analyzing employee performance at a call-center. I'd like indicate whether a particular employee's conversion (success) rate is significantly different from the average/expected conversion rate. Some factors for consideration:

  • there is a wide range of events for each employee. i.e. some employees have only taken 200 calls, while others have taken over 2,000 calls.
  • there are 2 types of conversions. the first type has an average conversion rate of around 15%, while the other is around 80%. That may be irrelevant because I can just separate them into two separate tests.

Initially, I was using a chisquare test from scipy.stats, but now I'm wondering if I should use the chi2_contingency test or something else. I'm worried that the chisquare test does not take into account the number of calls an operator takes, only the difference between the observed and expected amount of conversions (i.e. expected = (average conversion rate)*(total operator calls)). Would it be wrong to do a normal t-test to test each operator against the whole population?

Additionally, would conversions count as poisson distribution and does that make a difference to what test to use?

Improve this question
$\endgroup$
1
  • 1
    $\begingroup$ What do you mean by conversion rate ? What is some employees ? $\endgroup$
    – user10619
    Commented Oct 25, 2017 at 4:29

1 Answer 1

Reset to default
1
$\begingroup$

You can look at the proportion. SO, for example, if the average employee converts 25%, you can see if a specific employee who only converted 22% was statistically meaningful.

There is a whole literature on how to do this. Take a look at: https://onlinecourses.science.psu.edu/stat414/node/268

Improve this answer
$\endgroup$
2
  • $\begingroup$ So this package? statsmodels.org/dev/generated/… $\endgroup$
    – Hiram Foster
    Commented Oct 25, 2017 at 13:48
  • $\begingroup$ Don't know it and so can't vouch for the package, but it looks like the answer is "yes". This problem has a fairly stylized setup: suppose we send batters to the plate to hit during a game. On average, 33% strike out in the game. Buster Newplayer struck out 40% of the time at plate. Is this meaningful? While you are (rightly) measuring by odds (maybe other players got 10 times at bat, but Buster only got 5) and so doing percentages, the question is this: on average, was he meaningfully above average? The challenge is that his hitting is part of what gave the 33% originally. $\endgroup$
    – eSurfsnake
    Commented Oct 28, 2017 at 6:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.