A few weeks back in the Chandoo.org Forum Melvin asked about Apportioning Sales based on Division and Status to the current sales by store.
Today we will examine how this works and how to develop a solution for the problem.
Apportion/ing
Apportion means to assign or distribute.
In a court the Judge may apportion blame for an accident eg: 50% to the driver, 30% to mechanical failure and 20% to the road conditions, The Judge is assigning or distributing the blame as he deams appropriate.
This is what Melvin wanted to do with his sales. He wanted the sales distributed according to Division and Status based on the current sale by store.
Lets start simple and look at how we can distribute sales on a simple model first.
Let say we have a Distributorship and we buy and sell oranges.
We buy oranges from a supplier and distribute them to 3 stores, Store A, Store B & Store C
We received 1,000 oranges and they were sold as follows
We can see that each store received differing percentages of the original supply:
50% = 500/1000
30% = 300/1000
etc
A week later the supplier give us another 200 oranges and we want to distribute them based on the previous sales
So the new batch of 200 oranges will be distributed according to these previous percentages
100 = 200 x 50%
60 = 200 x 30%
etc
That is as simple and as complicated as apportioning is.
Melvin’s Problem
When we look at Melvin’s problem he has a more complex set of data
You can follow along using a sample file: Download Sample File
We can see that Melvin has 14 stores located in 4 Divisions (N, S, W & C) and each can have a status of Open or Open1
But if we simplify this and look at one set of data we can devise a formula which will adjust to each set of data
Lets develop a formula for cell F11 which is dealing with Store 1 in the N Division and has a Status of Open
We see it has sales of 100 (Cell E11)
Total sales of Division N and status Open are 600 (100+100+100+150+150) highlighted below (Lower table)
Looking at the Upper Table we can see that we need to distribute 200 units based on the Division N and Status Open (Cell C4)
So we need to distribute 200 units across the 5 stores with Total sales of 600
We know that Store 1. had sales of 100 in cell E11
The Total Sales of Stores in Division N and Status Open can be obtained using a Sumifs function
=SUMIFS($E$11:$E$24, C11:$C$24, $C$11, $D$11:$D$24, D11)
=600
So the proportion of Store 1’s sales 100 to Total Sales ( Division N and Status Open ) 600 is 100/600 = 16.66%
This is calculated by
=E11/SUMIFS($E$11:$E$24, $C$11:$C$24, C11, $D$11:$D$24, D11)
=0.1667
=16.67%
Note: We leave the references to C11, D11 & E11 variable, so that when the formula is copied down it will refer to the next row
We can use an index/match formula to get the 200 based on the criteria from row 11
=INDEX($C$4:$D$7, MATCH(D11,$B$4:$B$7,0), MATCH(C11,$C$3:$D$3,0))
What this is doing is doing a 2D Lookup in the Range $C$4:$D$7
It is looking up the Division Row no. MATCH(D11,$B$4:$B$7,0)
and looking in the Status Column No. MATCH(C11,$C$3:$D$3,0)
Note: Once again we leave the references to C11 & D11 variable, so that when the formula is copied down it will refer to the next row
So the proportion of the 200 sales attributable to Store 1 is:
=Distribution Qty * Actual Sales / Total Sales
=INDEX($C$4:$D$7, MATCH(D11, $B$4:$B$7, 0), MATCH(C11, $C$3:$D$3, 0)) * E11 / SUMIFS($E$11:$E$24, $C$11:$C$24, C11, $D$11:$D$24, D11)
= 33.33
We can now copy this down to all the cells matching our criteria of Division N and Status Open
Notice that the total matches the total to be distributed 200 showing that the formula is working
Although we copied the formula down to the cells that had matching criteria each part of the formula was setup to work on the appropriate criteria for the store in the current row
If we now copy F11 down to the other stores you will see that in fact all the stores sales have been apportioned according to the correct criteria.
eg: If we look at Stores 7, 8 & 9 we can see that they are in the W Division and have a Status of Open1
The distributed Proportions are each 16.67, totaling 50, which matches the distribution in the Upper table.
You may also notice that Division C has not been accounted for.
I assume that Melvin has sent us a subset of the data and that is why it is missing.
Download
You can download a copy of the above file and follow along, Download Sample File.
A Challenge
Can you solve the problem another way ?
Post your solutions in the comments below.
Other Posts in this Series
The Formula Forensics Series contains a wealth of useful solutions and information specifically about how Normal Formula and specifically Array Formula work.
You can learn more about how to pull Excel Formulas apart in the following posts: http://chandoo.org/wp/formula-forensics-homepage/
If you have a formula and you want to understand how it works contact Hui and it may be featured in future posts.
12 Responses to “29 Excel Formula Tips for all Occasions [and proof that PHD readers truly rock]”
Some great contributions here.
Gotta love the Friday 13th formula 😀
Great tips from you all! Thanks a lot for sharing! bsamson, particularly you helped me on a terribly annoying task. 🙂
(BTW, Chandoo, it's not exactly "Find if a range is normally distributed" what my suggestion does. It checks if two proportions are statistically different. I probably gave you a bad explanation on twitter, but it'd be probably better if you fix it here... 🙂 )
Great compilation Chandoo
For the "Clean your text before you lookup"
=VLOOKUP(CLEAN(TRIM(E20)),F5:G18,2,0)
I would like to share a method to convert a number-stored-as-text before you lookup:
=VLOOKUP(E20+0,F5:G18,2,0)
@Peder, yeah, I loved that formula
@Aires: Sorry, I misunderstood your formula. Corrected the heading now.
@John.. that is a cool tip.
Hey Chandoo,
That p-value formula is really great for a statistics person like me.
What a p-value essentially is, is the probability that the results obtained from a statistical test aren't valid. So for example, if my p value is .05, there's a 5% probability that my results are wrong.
You can play with this if you install the Data Analysis Toolpak (which will perform some statistical tests for you AND provide the P Value.)
Let's say for example I've got two weeks of data (separated into columns) with the number of hours worked per day. I want to find out if the total number of hours I worked in week two were really all the different than week one.
Week1 Week2
10 11
12 9
9 10
7 8
5 8
Go to Data > Data Analysis > T-Test Assuming Unequal Variances > OK
In the Variable 1 Box, select the range of data for week 1.
In the Variable 2 Box, select the range of data for week 2.
Check "Labels"
In the Alpha box, select a value (in percentage terms) for how tolerant you are of error.
.05 is the general standard; that is to say I am willing to accept a 95% level of confidence that my result is accuarate.
Select a range output.
Excel calculates a number of results: Average (mean) for each week's data, etc.
You'll notice however that there are two P Values; one-tail and two-tail. (one tail tests are for > or .05), the number of hours I worked in week two is statistically equivalent to the number of hours I worked in week one.
So here’s a way you might want to use this. You put up a new entry on your blog. You think it’s the best entry ever! So you pull your webstats for this week and compare it to last week. You gather data for each week on the length of time a visitor spends on your website. The question you’re trying to prove statistically is whether there’s an average increase in the amount of time spent on your website this week as compared to last week (as a result of your fancy new blog post). You can run the same statistical test I illustrated above to find out. Incidentally, it matters very little to the stat test whether the quantity of visitors differs or not.
Anyhow, the Data Analysis toolpack doesn't perform a lot of stat tests that folks like me would like to have access to. In those cases I have to either use different software, or write some very complicated mathematical formulas. Having this p-value formula makes my life a LOT easier!
Thanks!
Eric~
Fantastic stuf..One line explanation is cool.
Thanks to all the contributors
OS
Take FirstName, MI, LastName in access (you can fix it to work in excel) capitalize first letter of each and lowercase the rest and add ". " if MI exists then same for last name:
Full Name: Format(Left([FirstName],1),">") & Format(Right([FirstName]),Len([FirstName])-1),"") & ". ","") & Format(Left([LastName],1),">") & Format(Right([LastName],Len([LastName])-1),"<")
I teach excel, access, etc etc for a living and i have my access students build this formula one step at a time from the inside out to show how formulas can be made even if it looks complicated. Yes I know I could just do IsNull([MI]) and reverse the order in the Iif() function but the point here is to nest as many functions as possible one by one (also I illustrate how it will fail without the Not() as it is)
Extract the month from a date
The easiest formula for this is =MONTH(a1)
It will return a 1 for January, 2 for February etc.
if in a column we write the value of total person for eg. 10 if we spent 1.33 paise each person then how we get total amount in next column and the result will in round form plzzzzz solve my problem sir................... thank u
@Anjali
If the value 10 is in B2 and 1.33 paise is in C2 the formula in D2 could be =B2*C2
If the values are a column of values you can copy the formula down by copy/paste or drag the small black handle at the bottom right corner of cell D2
kindly share with me new forumulas.
How to convert a figure like 870.70 into 870 but 871.70 into 880 using excel formula ? Please help.