Is a team member pulling their weight as far as work is concerned?

[Darren Bartrup]

Hi all,

This is more of a statistics question rather than Excel or Access related and as such I've cross posted on http://www.talkstats.com/showthread.php/56705-Is-a-team-member-pulling-their-weight-as-far-as-work-is-concerned?p=160170#post160170


I'm building an Access database at work which stores how much work a specific team has been supplied with and has completed from day to day. I'm looking to find out if team members are doing more or less than is expected of them.


The problem I'm having is that a team performs a variety of work types, each of which takes a different amount of time to complete and each team member can only complete the amount of work they're given which may be more or less than is expected of the team member.


The data structure is as follows:

• A team has a number of processes numbered 1 to N. Each process has a an expected time to complete.
• The total number of items for each process received by the team as a whole is recorded.
• Each team member has a total of hours & minutes worked per day.
• Each team member is allocated a number of items to work on through the day - this comes from a variety of work types.
• The number of items each team member completes is recorded.

So with the raw data listed above can I calculate that a particular team member is reaching/exceeding the expected target per process - shown as a percentage or any other method I could understand.


Can I also then boil the separate achieved targets down into a single figure that says the team member is getting a green thumbs up, a red thumbs down, or a yellow 'alright' thumb?


Any help will be greatly appreciated in explaining stats to me.


Regards,
Darren.

 

[Dendro]

It depends on whether the workload(=would define the time a team member needs to complete a task) per item is equal. Or you would have to measure somehow how the workload of these items compare to the other items of the same process.

Now lets assume the workload is equally divided in a process under X items. The expected time(Y) to complete the process is known as you said yourself, so the expected time to complete one item would be Y/X. Note this is the expected time per item, we will call it Z(doesn't really matter what, just for future reference)

Now take a look at what a team member does. How many x items can/does that person complete per time y. This is were I'm not totally sure what you mean by "The number of items each team member completes is recorded." is this each time per day?
Let's assume it is measure within one day and you have a table containing dates with the amount of items completed on each date. We could than calculate the average time per item (z), out of all the items completed(create a sum formula here to get this value) and dividing the total hours worked(sum of nr. 3 in your list).

So now we have 2 values which we can compare: the expected time per item (Z) and the average time per item of a certain team member. So now we can compare if this value higher or lower than the expected, but is this within the range of the expected? If I remember correctly we could use a t-test with unknown variance to compare these values. But how would you determine how far this average time/item can deviate from the expected time/item to conclude whether this person is reaching/exceeding the expected value? I have no clue about that, but you need a set a certain power in the test to decide whether it is within the expected or outside of the expected...You should think about that.

Something that is also worth thinking about is whether this expected time is realistic(where do you get this value from, is it just a very hard process the team is working on?). But you might want to see if one team member is underachieving or the opposite against the other members of the team. For this i would suggest you look into ANOVA( to simplify it a bit, it compares the variation within groups(= variation in time per item for a member) and the variation between groups(=the variation of the averages of the teammembers)). However as in all statistic tests you will have to set your p-value again as i stated above.

Anyway working with statistics has been a while a go, so it might be a bit confusion. I hope i got you thinking in the right direction though

 

[Darren Bartrup]

Hi Dendro,

Thanks for the reply - think I'll have to read your post a few times to fully understand it.

Now lets assume the workload is equally divided in a process under X items.

Not 100% sure what you mean here - the time an item takes to complete will be the same in a single process.
e.g

• A team member may have to forward 100 emails in a day, each item here (each email) takes only a couple of seconds to forward, they may also have 10 emails to respond to which will take approx 10 minutes per email (except the dodgy one which took 30 mins).
• Another team member gets 30 emails to forward in a day, 15 emails to respond to and 5 payments to process.

I don't however have the times take to complete those 100 emails - I just know that in the 7 hours of working they completed 100 email forwards and 10 email responses. It might have taken 6 hours to do the forwards and 1 hour to do the responses (but very unlikely).

Let's assume it is measure within one day and you have a table containing dates with the amount of items completed on each date

Yep, that's exactly it.

I'm now studying these pages to try and get some further insight:
Introduction to ANOVA / MANOVA
What are Basic Statistics for independent samples

Regards,
Darren.

 

[Dendro]

Ill' have to read your first post and my own reply again to answer your post

 

[Dendro]

Ok didn't quite understand the first post as good as i thought. I thought a process was some kind of project(eg make applesauce) with different items/tasks(eg. item1 peel the apple, item2 remove the core, item 3 cut in pieces,...) to complete, but your example makes it all clear now. The items is the number of times the process has to be done. So "the workload is equally divided in a process under X items" is true.

Maybe what you are searching doesn't have to be so complicated at all.

A simple method would be to simply take the sum of all hours worked off all days you have in your database and divide it by the sum of expected hours the team-member gets for all items of all processes he has completed. Multiply this number by 100
and you have a percentage that indicates the work-rate relative to the expected work-rate.
-A percentage above 100% is not reaching the expected time for all his work. So 120% would mean he/she needs 12min more than needed for 1hour of work
-A percentage under 100% is not reaching the expected.

Then you need to set a limit, depending on how strict you are, for your thumbs up or thumbs down. f.i 90-110 is alright, 110+ is green, and below 90 is red

However i think it is impossible to do this for particular processes, as you don't know if he/she has worked a certain amount of time on a process within a day. This could only be done if you could do the above for days when a team-member was assigned only one process on a day.

 

[Ben Nevis]

Hi Darren,

I used to be a work study practitioner so (I think) I have some insight here.
Dendro's suggestion is correct:
You need a "Standard time" per unit produced for each task, it is normal to state this in minutes and decimal parts of a minute.
So if the Standard time to produce a gadget is 10.25 minutes and the time to produce a widget is 14.75 minutes and Employee A has produced 18 gadgets and 16 widgets in a given day they have "earned" 18 * 10.25 + 16 * 14.75 = 420.5 Standard minutes. If their working day is 7 hours 30 minutes they worked for 450 minutes so their "Performance Index" = Earned minutes / Actual minutes = 420.5 / 450 = 93.4 (this is not stated as a percentage). However you also need to take "Ineffective time" into account so if Employee A spent 29.5 minutes waiting for the components required to produce the widgets and gadgets then their "Performance Index" = Earned Minutes / (Actual minutes - Ineffective time) = 420.5 / (450 - 29.5) = 100. A Performance Index (PI) of 100 is known as Standard Performance.

You can directly compare each person's PI to see who is and isn't performing. The statistical test you need to to show that Employee A's PI is significantly different from the mean PI (over time) is the F-test. You would normally calculate this to a 95% or 99% confidence level and a p-value < 0.05 or 0.01 respectively would give you "statistical" proof that their performance was less than average.

Please note that in the UK this comes under the remit of the Institute of Management Services and requires appropriate qualifications. There is a lot more to this than I have outlined above and you should be very careful about how you use such information. Employee tribunals can have a field day with comparisons of employee performance in cases of unfair dismissal so if you want to use your data to "prove" that someone isn't pulling their weight, get some professional advice.

 

[Darren Bartrup]

Thank you both for your insights into this and I'm thinking I can get this to work now.

Ben Nevis - your last paragraph is probably the most important. The calculations need to be correct as we'll have to prove the formula to the lawyer of the first person to disagree.

Thanks again!