# Cubicle Denizen Dilemma: the name for an “average” if the data are weirdApril 20, 2015 5:35 AM   Subscribe

Our company has decided to measure the jobs we do by assigning points to tasks. The problem here is that departments don’t do the same thing, or the same amount of things. What concerns me is that the company wants to publish a company-wide average, against which members in each department will be compared.

Without going into detail, let’s imagine there are 4 departments: A, B, C, & D.
Staff in departments A & B answer internal phone calls. This gets each staff member 50 points, regardless of how many calls there are. Department C answers internal calls (50 guaranteed points) but can also answer external ones. Anyone answering an external phone call gets an extra 10 points. Department D handles internal calls (total 50 points), has the possibility of external calls (10 points per call), and can also receive calls from abroad (also 10 points per call).
To me, the system seems flawed from the outset (whether there are external or domestic calls or not).
With even a few calls taken by staff in Departments C & D, staff in A & B will appear to be below “average”.
It looks like some sort of systemic flaw, but what do I know?
My questions:
Q1: Is there a name for this sort of situation?
Q2: Is there a name for an average where, technically, yes, it is the average, but it was not possible for all members to score equally?
posted by Prof Iterole to Work & Money (8 answers total)

Not sure of the terminology, but agree that grouping these into a single set will give biased data. The best way to calc individual performance would be across each set distinctly (staff member in A is compared to the others in group A, B in B, etc.).
posted by Gee Your Hair Smells Terrific at 5:45 AM on April 20, 2015

It's a weighted average--they're computing the average number of tasks completed, but the tasks have different values.

It's not necessarily inherently flawed, if you get the weights right. You need to pick weights where the "typical" employee ends up with X points (call it 1200 for the rest of this), regardless of their department. So You might expect someone in department A to get 24 calls, all internal, but someone in department C to get 20 calls, but theirs were all external (for some wacky reason, namely making my numbers easy). In both cases, you expect a worker to end up with 1000 points , even if they got there differently depending on which department they're in.

You're trying to come up with a scale where someone in department A doing lots of quick tasks will have the same output (in "adjusted tasks", i.e. points) as someone in department Z who does a handful of highly complicated tasks when they're performing their jobs equally and to that, you have to first figure out what 100% output should be in each department rather than valuing tasks and hoping you end up with comparable numbers across departments.
posted by hoyland at 5:56 AM on April 20, 2015 [1 favorite]

The way to do this would be to calculate an expected or target number of points per department, based on the tasks that people in that department perform, the desired productivity, and their point values. Then compare between departments the % of target points achieved.
posted by Nothing at 5:59 AM on April 20, 2015

This does seem weird. Not just the arbitrary, complex point system, but the fact that you need a single central authority in the company to measure the performance of everyone. Normally that's what your management structure is there for.

If Joe in accounting is not getting enough paperwork done, I'd expect Joe's direct supervisor to be responsible for reporting on his struggles, communicating to Joe that he is under performance review, and providing useful feedback to give him a chance to improve. Are you saying in your company you're going to have some Central Performance Intelligence System that churns through the numbers and issues warnings automatically? Sounds like an.. um.. interesting idea.

Anyway.

Is there a name for this sort of situation?

Bad management? Orwellian nightmare state? The plot of Brazil? I'm not sure there is a label for this.

Is there a name for an average where, technically, yes, it is the average, but it was not possible for all members to score equally?

Normal distributions are such things. I'm not really sure what you mean by that second clause, though, and the answer depends on what you mean by "average." If we scored up every person in your company and lumped their scores together on a graph, I would expect something like a normal distribution, with most scores clustered around a central point and then smoothing out toward the sides. Then, if you were going to use this scoring mechanism to say "this person is doing just okay" I would find the mean and say anyone within one standard deviation of it is "average."
posted by deathpanels at 6:11 AM on April 20, 2015 [1 favorite]

So, people in A and B will always have exactly 50 points, no matter how many calls they take, with no other way of earning points? Comparisons with other departments aside, that seems like a useless incentive system: nothing they do can change their score.
posted by xris at 6:51 AM on April 20, 2015

Is there a name for an average where, technically, yes, it is the average, but it was not possible for all members to score equally?

"Average." Nothing about an average requires the possibility for all members to have an equal score. It's just the sum over N. This is just a silly and uninformative thing to take the average of, is all.
posted by ROU_Xenophobe at 7:00 AM on April 20, 2015

Some folks call this GIGO (Garbage In, Garbage Out). (The link focuses on this being a computer term, but it is also a math term -- I know it primarily as a math term.)

You might try looking up mean, median, and mode (or even read the book "How to lie with statistics"), which are three different ways to talk "average" (these terms all have different defnitions..blah blah blah). It may not make any difference in how they handle this but it might make you feel better about the fact that statistics often do not say what a lot of people think they say.

"There are three kinds of lies: lies, damned lies and statistics."
- Mark Twain's Own Autobiography

posted by Michele in California at 10:28 AM on April 20, 2015

Hi, I solve problems like yours every day for a living. Your system is broken, you need to change the points.

If all departments *always* get 50 points, irrespective of how many internal calls they take, this value becomes meaningless and can be zeroed out.

This leaves you with Departments B, C, D which can earn points, and Department A which cannot.

The simplest solution is that each internal call is also valued at some level, probably lower than the external or foreign calls. Options to determine value:

(1) Arbitrarily assign 5 points per internal call... OR

(2) Estimate how many internal calls are occurring in each Department (or only in Department A, would need more information to decide), divide by number of employees who take those calls to get average number of calls per person per week (or month or day). Let's say it's 10 internal calls that each employee is doing per week, this gives you 50 (your original points amount per week) / 10 = 5 points per call.
posted by Ender's Friend at 6:15 PM on April 20, 2015 [1 favorite]

« Older What might a job recruiter/headhunter be looking...   |   Can three wireless routers become one? Newer »