A week ago Tarun asked a question on the Chandoo.org Forums.
“I have got multiple names in each row and would like to have what name is repeated maximum number of times and how many times?
Eg. Ram, Amita, Obama, Ram, Willi, Ram, Amita, Chandoo, Ram, Willi
Ans: Ram (4 times)”
(The list and answers are edited)
Chandoo responded with a neat Array Formula:
=INDEX(B2:K2,MATCH(MAX(COUNTIF(B2:K2,B2:K2)), COUNTIF(B2:K2,B2:K2),0)) &
” (“&MAX(COUNTIF(B2:K2,B2:K2))&” times)”
Lets take a look inside this and see how it works
THE EXAMINATION
The formula has two parts separated by a &
=INDEX(B2:K2,MATCH(MAX(COUNTIF(B2:K2,B2:K2)), COUNTIF(B2:K2,B2:K2),0))
and
&
and
” (“&MAX(COUNTIF(B2:K2,B2:K2))&” times)”
Each part is separate and can be used independently, the & character simply joins the two parts together to make a single string which answers Tarun’s question, Ram (4 times).
Now, lets look at each part.
You can follow along with this forensic examination by downloading the Sample Data File.
=INDEX(B2:K2,MATCH(MAX(COUNTIF(B2:K2,B2:K2)), COUNTIF(B2:K2,B2:K2),0))
This is a single Index Function with 2 components, being:
a Range B2:K2 and
a Count MATCH(MAX(COUNTIF(B2:K2,B2:K2)), COUNTIF(B2:K2,B2:K2),0)
Typically an Index Function uses 3 components
=Index(Array, Row Number,[Column Number])
In this example the Range is a single Row, B2:K2
And so using the Counter in the Row spot has the effect of counting down the first Column and then continuing at the top of the second Column etc
So the formula used:
=INDEX(B2:K2,MATCH(MAX(COUNTIF(B2:K2,B2:K2)), COUNTIF(B2:K2,B2:K2),0))
Is equivalent to:
=INDEX(B2:K2,1,MATCH(MAX(COUNTIF(B2:K2,B2:K2)), COUNTIF(B2:K2,B2:K2),0))
Now lets jump ahead to the COUNTIF(B2:K2,B2:K2) bit
If you copy =COUNTIF(B2:K2,B2:K2) to a cell, Press F2 and then evaluate the Formula using F9
You will see that it returns an array. The array is highlighted by the squiggly brackets { } ‘s
={4,2,1,4,2,4,2,1,4,2}
This is the heart of the solution.
What this is showing us is that for each position in the range B2:K2, the count of how many times that cells value occurs in the range B2:K2
So the formula
=INDEX(B2:K2,MATCH(MAX(COUNTIF(B2:K2,B2:K2)), COUNTIF(B2:K2,B2:K2),0))
Is equivalent to
=INDEX(B2:K2,MATCH(MAX({4,2,1,4,2,4,2,1,4,2}), {4,2,1,4,2,4,2,1,4,2},0))
Looking at the MAX({4,2,1,4,2,4,2,1,4,2}) part, this simplifies to 4, the Maximum value of the array (Remember this line, we’ll come back to it later).
So our simplified formula is now: =INDEX(B2:K2,MATCH(4, {4,2,1,4,2,4,2,1,4,2},0))
Now looking at the MATCH(4, {4,2,1,4,2,4,2,1,4,2},0) part of the equation
You can see that Match is looking for the value 4, in the array {4,2,1,4,2,4,2,1,4,2}, which is the First value , Position 1, the 0 requesting that an exact match is found.
So that MATCH(4, {4,2,1,4,2,4,2,1,4,2},0) is equivalent to 1
So our equation =INDEX(B2:K2,MATCH(4, {4,2,1,4,2,4,2,1,4,2},0))
Is now simplified even more to =INDEX(B2:K2, 1)
Index will then look in B2:K2 and will return the first cell or “Ram” in this example.
& “(” & MAX(COUNTIF(B2:K2,B2:K2)) & ” times)”
The second part of the equation is responsible for counting the number of Times Ram occurs and displaying it with some text.
& “(” & MAX(COUNTIF(B2:K2,B2:K2)) & ” times)”
The parts displayed in Red above add the text ( and times) to the Count
Remember the section MAX(COUNTIF(B2:K2,B2:K2)) which was explained above and evaluates to 4 in this case
So the & “(” & MAX(COUNTIF(B2:K2,B2:K2)) & ” times)”
Part evaluates to: ( 4 times)
With the initial & adding it to the text of the first part Ram for the final result – Ram ( 4 times)
LEARN MORE ABOUT ARRAY FORMULAS
You can learn more about Array Formulas at the following links:
http://www.cpearson.com/excel/ArrayFormulas.aspx
http://www.databison.com/index.php/excel-array-formulas-excel-array-formula-syntax-array-constants/
http://office.microsoft.com/en-us/excel-help/introducing-array-formulas-in-excel-HA001087290.aspx
Chandoo.org has several articles on Array Formulas
http://chandoo.org/wp/tag/array-formulas/
FORENSIC FORMULAS
Would you like to see more “Forensic” examination of complex formulas ?
Let us know in the comments below and it may become a regular section at Chandoo.org.
13 Responses to “Using pivot tables to find out non performing customers”
To avoid the helper column and the macro, I would transpose the data into the format shown above (Name, Year, Sales). Now I can show more than one year, I can summarize - I can do many more things with it. ASAP Utilities (http://www.asap-utilities.com) has a new experimental feature that can easily transpose the table into the correct format. Much easier in my opinion.
David
Of course with alternative data structure, we can easily setup a slicer based solution so that everything works like clockwork with even less work.
David, I was just about to post the same!
In Contextures site, I remember there's a post on how to do that. Clearly, the way data is layed out on the very beginning is critical to get the best results, and even you may thinkg the original layout is the best way, it is clearly not. And that kind of mistakes are the ones I love ! because it teaches and trains you to avoid them, and how to think on the data structure the next time.
Eventually, you get to that place when you "see" the structure on the moment the client tells you the request, and then, you realized you had an ephiphany, that glorious moment when data is no longer a mistery to you!!!
Rgds,
Chandoo,
If the goal is to see the list of customers who have not business from yearX, I would change the helper column formula to :
=IF(selYear="all",sum(C4:M4),sum(offset(C4:M4,,selyear-2002,1,columns(C4:M4)-selyear+2002)))This formula will sum the sales from Selected Year to 2012.
JMarc
If you are already using a helper column and the combox box runs a macro after it changes, why not just adjust the macro and filter the source data?
Regards
I gotta say, it seems like you are giving 10 answers to 10 questions when your client REALLY wants to know is: "What is the last year "this" customer row had a non-zero Sales QTY?... You're missing the forest for the trees...
Change the helper column to:
=IFERROR(INDEX(tblSales[[#Headers],[Customer name]:[Sales 2012]],0,MATCH(9.99999999999999E+307,tblSales[[#This Row],[Customer name]:[Sales 2012]],1)),"NO SALES")
And yes, since I'm matching off of them for value, I would change the headers to straight "2002" instead of "Sales 2002" but you sort the table on the helper column and then and there you can answer all of your questions.
Hi thanks for this. Just can't figure out how you get the combo box to control the pivot table. Can you please advise?
Cheers
@Kevin.. You are welcome. To insert a combo box, go to Developer ribbon > Insert > form controls > combo box.
For more on various form controls and how to use them, please read this: http://chandoo.org/wp/2011/03/30/form-controls/
Thanks Chandoo. But I know how to insert a combobox, I was more referring to how does in control the year in the pivot table? Or is this obvious? I note that if I select the Selected Year from the PivotTable Field List it says "the field has no itens" whereas this would normally allow you to change the year??
Thanks again
worked it out thanks...
when =data!Q2 changes it changes the value in column N:N and then when you do a refreshall the pivottable vlaues get updated
Still not sure why PivotTable Field List says “the field has no itens"?? I created my own pivot table and could not repeat that.
Hi, I put the sales data in range(F5:P19) and added a column D with the title 'Last sales in year'. After that, in column D for each customer, the simple formula
=2000+MATCH(1000000,E5:P5)
will provide the last year in which that particular customer had any sales, which can than easily be managed by autofilter.
Somewhat longer but perhaps a bit more solid (with the column titles in row 4):
=RIGHT(INDEX($F$4:$P$19,1,MATCH(1000000,F5:P5)),4)
[…] Finding non-performing customers using Pivot Tables […]