Excel - How to count pairs in two columns containing lists - excel

Question

I have a number of farmers registered on my database. Each farmer grows a few fruits and sells to a few counties.
For every fruit / county pair (e.g. apple, Warwickshire), how do I count the number of farmers that can supply that combo?
I have over 100 farmers registered on my database.
So my database has a row for each farmer, a column for fruits and a column for the counties they cover. The fruits and counties that each farmer covers are recorded as comma separated lists in the two cells on that farmer's row.
I want to create a matrix with fruits on the horizontal and counties on the vertical to count how many farmers cover that particular combo.
For the example in the screenshot, I've tried:
=COUNTIF(A2:B4,AND(ISNUMBER(SEARCH(G11,A2,1)),ISNUMBER(SEARCH(A13,B2,1)))="TRUE")
but with no luck.

Answer 1

IF you have Excel 2010+, you can do this with Power Query (aka Get & Transform in Excel 2016+).
Using Power Query allows you to update the table easily whenever any new products (or counties) are added. You just re-run the query after you add rows to the data table (or add a product or county to a given row).
Except for removing the extra spaces (Trim after splitting the columns), all can be done via the GUI. But you can just paste the M-Code into the Advanced Editor and then explore the GUI to study the individual steps.
let
Source = Excel.CurrentWorkbook(){[Name="Table1"]}[Content],
#"Changed Type" = Table.TransformColumnTypes(Source,{{"Products", type text}, {"Counties", type text}}),
#"Split Column by Delimiter" = Table.ExpandListColumn(Table.TransformColumns(#"Changed Type", {{"Counties", Splitter.SplitTextByDelimiter(",", QuoteStyle.Csv), let itemType = (type nullable text) meta [Serialized.Text = true] in type {itemType}}}), "Counties"),
#"Changed Type1" = Table.TransformColumnTypes(#"Split Column by Delimiter",{{"Counties", type text}}),
#"Split Column by Delimiter1" = Table.ExpandListColumn(Table.TransformColumns(#"Changed Type1", {{"Products", Splitter.SplitTextByDelimiter(",", QuoteStyle.Csv), let itemType = (type nullable text) meta [Serialized.Text = true] in type {itemType}}}), "Products"),
#"Changed Type2" = Table.TransformColumnTypes(#"Split Column by Delimiter1",{{"Products", type text}}),
#"Added Custom" = Table.AddColumn(#"Changed Type2", "Prod", each Text.Trim([Products])),
#"Added Custom1" = Table.AddColumn(#"Added Custom", "County", each Text.Trim([Counties])),
#"Removed Columns" = Table.RemoveColumns(#"Added Custom1",{"Products", "Counties"}),
#"Grouped Rows" = Table.Group(#"Removed Columns", {"County", "Prod"}, {{"grouped", each _, type table [Prod=text, County=text]}, {"counts", each Table.RowCount(_), type number}}),
#"Removed Columns1" = Table.RemoveColumns(#"Grouped Rows",{"grouped"}),
#"Pivoted Column" = Table.Pivot(#"Removed Columns1", List.Distinct(#"Removed Columns1"[Prod]), "Prod", "counts", List.Sum)
in
#"Pivoted Column"
Original Data
Results

Answer 2

Just for "fun" I built a solution with AND() and FIND() :
IFERROR(AND(FIND(B$7,$A$2,1),FIND($A8,$B$2,1)),0)+IFERROR(AND(FIND(B$7,$A$3,1),FIND($A8,$B$3,1)),0)+IFERROR(AND(FIND(B$7,$A$4,1),FIND($A8,$B$4,1)),0)
You could wrap this in an IF() so you only show results greater than 1 which may make it easier to spot the ones wanted.
IF(IFERROR(AND(FIND(B$7,$A$2,1),FIND($A8,$B$2,1)),0)+IFERROR(AND(FIND(B$7,$A$3,1),FIND($A8,$B$3,1)),0)+IFERROR(AND(FIND(B$7,$A$4,1),FIND($A8,$B$4,1)),0)<2,"")

Answer 3

Oh I get it . Its the way it has been tabulated by the form-filler(s) or has been given the data (I'm explaining it to others here so they get where I'm coming from). He probably wants to change the way these forms are filled to make it easier to read, follow and more efficient logical process going forward.
he has received confusing/badly compiled information/tables and wants to make them more straightforward/logical.
I think I got it working according to the way I understood the way you worded the question, what i know how to do in excel and your information given. Way I did it works like a "count the number of occurrences any specific word appears in a string." .
version 1:
=(LEN(lookupall("*"&B$8&"*",$A$2:$B$4,2))-LEN(SUBSTITUTE(lookupall("*"&B$8&"*",$A$2:$B$4,2),$A9,"")))/LEN(B$8)
and drag across and down.
or better:
version 2:
=(LEN(lookupall("*"&B$7&"*",$A$2:$B$4,2))-LEN(SUBSTITUTE(lookupall("*"&B$7&"*",$A$2:$B$4,2),$A8,"")))/LEN($A8)
[now 14:00 edited above - 5am written and it was off by some cells]
my results are :
version 1 results table:
&
version 2 results table: I think this is exactly what you wanted.
Notes: yes, in both source tables, A2:B4 I've called them those names (but the data is the same. war = Warwickshire. app=apples etc. )
Which one does what your seeking most?
lookupall is a UDF you can find on the net if you search around. It gives all vlookup results including duplicate lookups, concatenated together. It occurred to me that you can then look at the number of times your values in A (counties) appear in each of the results (fruit look ups), and then divide by the number of letters in that word (the counties im version 1, the fruit in version 2) to get precise number.
in version 1, I think you have to round down/up the results (because when I get rid of the glos (Gloucestershire) in b2 for instance, the result in b12 becomes 0 which would be precise given those numbers). But version 2 is better - more accurate.
Is this kinda going in the right direction for you? Might be worth more tweaking, but I think given the approximate nature of question (the way I read it), its the best I can do. It would have been more accurate to tie ...
Though I am sure there are better, more versatile, generic-scientifically accurate, other-similar-table-applicable and precise formulas out there which would do it better in just 1 formula or 1 single UDF.
The lookupall UDF I use:
Function LookupAll(vVal, rTable As Range, ColumnI As Long) As Variant
Dim rFound As Range, lLoop As Long
Dim strResults As String
With rTable.Columns(1)
Set rFound = .Cells(1, 1)
For lLoop = 1 To WorksheetFunction.CountIf(.Cells, vVal)
Set rFound = .Find(what:=vVal, After:=rFound, LookIn:=xlFormulas, lookAt _
:=xlWhole, SearchOrder:=xlByRows, SearchDirection:=xlNext, MatchCase:= _
False, SearchFormat:=False)
strResults = strResults & ";" & rFound(1, ColumnI)
Next lLoop
End With
LookupAll = Trim(Right(strResults, Len(strResults) - 1))
End Function
This actually does this and (many other) jobs actually and has been a life-saver for much of my own work. (p.s. nobody asks me questions in the office & nobody gave me anything ! its all found, researched and discovered or made by me to survive!).
my Correct results table I am pleased to say is exactly the same as Solar Mikes! So Version 2 is correct
with
=(LEN(lookupall("*"&E$7&"*",$A$2:$B$4,2))-LEN(SUBSTITUTE(lookupall("*"&E$7&"*",$A$2:$B$4,2),$A11,"")))/LEN($A11)
in cell B8 and dragged down&across