About 6 months ago, Fred asked a question on the Chandoo.org Forums: I need idea on a simpler formula
In column A I have the total sum product of C to AU
Line one has all the names.
A2 = (B2*C2)+(D2*E2)+(F2*G2)+(H2*I2)+(J2*K2)+(L2*M2)+…+(AT2*AU2)
A3 = (B3*C3)+(D3*E3)+(F3*G3)+…+(AT3*AU3)
Is there is simpler way by line to do this without clicking each cell on line 2?
I tried Sumproduct but I think I have received a wrong answer during testing.
Hui offered a Sumproduct Formula as a solution
=SUMPRODUCT((B2:AT2) * MOD(COLUMN(B2:AT2) -1, 2), (C2:AU2) * MOD(COLUMN(C2:AU2), 2))
and then followed up with a simpler Sumproduct Formula a day later
=SUMPRODUCT(B2:AT2 * C2:AU2 * (MOD(COLUMN(B2:AT2), 2) =0))
Let’s take a look at this second solution.
Setup the Problem
Copy the numbers 10,20 into alternate Cells A2:U2 or download the example file here: Example File (Excel 97-2010)
Copy this formula into B6: =SUMPRODUCT(B2:U2 * C2:V2 * (MOD(COLUMN(B2:U2), 2)=0))
Pull The Formula Apart
Lets take a look inside this formula and see how it works.
=SUMPRODUCT(B2:U2*C2:V2*(MOD(COLUMN(B2:U2),2)=0))
We can see that in the above formula the main function used in the formula is a Sumproduct Function.
=SUMPRODUCT(B2:U2*C2:V2*(MOD(COLUMN(B2:U2),2)=0))
Within the Sumproduct function there are 3 arrays, which are multiplied together
Array 1: B2:U2
Array 2: C2:V2
Array 3: (MOD(COLUMN(B2:U2),2)=0)
What’s in these arrays?
Array 1 is simply the range from B2:U2
Array 2 is simply the range from C2:V2, note that it is offset from the first array by 1 Column.
This is so that the first value of the Second Array matches the first value of the First Array. That is they are both in position 1 within there respective arrays.
Array 3 is where all the action is.
Enter =(MOD(COLUMN(B2:U2),2)=0) into a cell and press F9
Excel returns: ={TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE}
Lets look at what is happening here
=(MOD(COLUMN(B2:U2),2)=0)
Mod returns the remainder of the first value Column(B2:U2) after dividing it by the second value 2.
Column(B2:U2) returns the Column Number for each cell in the Range B2:U2.
Because this part of the formula is in a Sumproduct formula it is evaluated as an Array Formula and hence it does this for each cell in the range B2:U2, thus returning an Array as the answer.
We can see that if we enter =MOD(COLUMN(B2:U2),2) into a cell and evaluate it with F9
={0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1}
However in this case we want to convert this from an array of Numbers to an array of True/False
A simple addition of an =0 does the trick
=(MOD(COLUMN(B2:U2),2)=0)
Now causes the formula to return: ={TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE}
So How does this go together with the other 2 arrays?
We now have 3 arrays:
Array 1: 10,20,10,20,10,20,10,20,10,20,10,20,10,20,10,20,10,20,10
Array 2: 20,10,20,10,20,10,20,10,20,10,20,10,20,10,20,10,20,10,20
Array 3: TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE
When Multiplying arrays, Excel multiplies the first value of each array and then the second value of each array, etc, effectively doing
={10*20*True, 20*10*False , 10*20*True, 20*10*False , 10*20*True, 20*10*False , 10*20*True, 20*10*False , 10*20*True, 20*10*False , 10*20*True, 20*10*False , 10*20*True, 20*10*False , 10*20*True, 20*10*False , 10*20*True, 20*10*False , 10*20*True}
You can see above that when the 3 arrays are multiplied it will only be the Odd entries in Arrays 1 & 2 which are evaluated by the Sumproduct, as all the even entries are multiplied by False which is equivalent to Zero
So the above evaluates to
={200,0, 200,0, 200,0, 200,0, 200,0, 200,0, 200,0, 200,0, 200,0, 200}
Sumproduct then takes over and adds the values together returning, 2000, the correct answer.
Problem Extension
After solving the problem, Fred decided to add a column between each entry in the data set.
The solution is posted in the forum and is also shown in the example file.
You can work through that formula at your leisure, except to say that it is similar to the solution above.
Alternate Solution
Luke correctly pointed out that the data was poorly arranged and the solution would be much simpler had the data been more logically arranged.
However as an Excel practioner we are often called to solve other peoples dirty work.
HINTS
You will notice that in the solution of this problem I have done a few small things that make solving the problem easier.
Use Smaller Subsets of the Data.
Instead of putting numbers from Columns B to AU as Fred has I have used a set from Column B to U.
This way I can see all the data on one Excel screen without scrolling as well as the formula links and extents when pressing F2 on the cell containing the formula.
Use Simple Numbers
Use numbers that you can manually check. In this example I can easily, manually, check that the answer should be 2000.
Evaluate
Where ever possible, enter sections of a formula in a cell and evaluate its output:
=(MOD(COLUMN(B2:U2),2)=0) and press F9
={TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE,TRUE,FALSE}
See how easily we can check that this section is returning the correct values.
OTHER POSTS IN THIS SERIES:
You can learn more how to pull Excel Formulas apart in the following posts:
http://chandoo.org/wp/category/formula-forensics/
The link above, will show you all the posts in this series
WHAT FORMULAS WOULD YOU LIKE EXAMINED?
If you have any formulas you would like explained please feel free to leave a post here or send me an email:
If the formula is already on Chandoo.org or Chandoo.org/Forums, simply send the link to the post and a Comment number if appropriate.
If sending emails please attach an Excel file with the formula and data
One Response to “SQL vs. Power Query – The Ultimate Comparison”
Enjoyed your SQL / Power Query podcast (A LOT). I've used SQL a little longer than Chandoo. Power Query not so much.
Today I still use SQL & VBA for my "go to" applications. While I don't pull billions of rows, I do pull millions. I agree with Chandoo about Power Query (PQ) lack of performance. I've tried to benchmark PQ to SQL and I find that a well written SQL will work much faster. Like mentioned in the podcast, my similar conclusion is that SQL is doing the filtering on the server while PQ is pulling data into the local computer and then filtering the data. I've heard about PQ query folding but I still prefer SQL.
My typical excel application will use SQL to pull data from an Enterprise DB. I load data into Structured Tables and/or Excel Power Pivot (especially if there's lot of data).
I like to have a Control Worksheet to enter parameters, display error messages and have user buttons to execute VBA. I use VBA to build/edit parameters used in the SQL. Sometimes I use parameter-based SQL. Sometimes I create a custom SQL String in a hidden worksheet that I then pull into VBA code (these may build a string of comma separated values that's used with a SQL include). Another SQL trick I like to do is tag my data with a YY-MM, YY-QTR, or YY-Week field constructed form a Transaction Date.
In an application, I like to create a dashboard(s) that may contain hyperlinks that allow the end-user to drill into data. Sometimes the hyperlink will point to worksheet and sometimes to a supporting workbook. In some cases, I use a double click VBA Macro that will pull additional data and direct the user to a supplemental worksheet or pivot table.
In recent years I like Dynamic Formulas & Lambda Functions. I find this preferable to pivot tales and slicers. I like to use a Lambda in conjunction with a cube formula to pull data from a power pivot data model. I.E. a Lambda using a cube formula to aggregate Accounting Data by a general ledger account and financial period. Rather than present info in a power pivot table, you can use this combination to easily build financial reports in a format that's familiar to Accounting Professionals.
One thing that PQ does very well is consolidating data from separate files. In the old days this was always a pain.
I've found that using SQL can be very trying (even for someone with experience). It's largely an iterative process. Start simple then use Xlookup (old days Match/Index). Once you get the relationships correct you can then use SQL joins to construct a well behaved SQL statement.
Most professional enterprise systems offer a schema that's very valuable for constructing SQL statements. For any given enterprise system there's often a community of users that will share SQL. I.E. MS Great Plains was a great source (but I haven't used them in years).
Hope this long reply has value - keep up the good work.