One of the most asked questions within the posts and Forums at Chandoo.org is “How Does Sumproduct work ?”.
Rahul recently asked for an example in Excels Sumproduct Formula post; Comment No. 55.
So today in Formula Forensics we will take a look at just that with a few worked examples.
Sumproduct
Excels help defines Sumproduct as:
So what are these arrays referring to:
An array in Excel can be :
A manual Array: {10;20;30}
A Range: A1:A3
A Named Range: MyRange1
Where MyRange1 is defined as a defined range in the Name Manager.
A Named Formula: MyRange2
Where MyRange2 is defined as a Formula returning a range in the Name Manager.
Lets look at each
You can follow along in the Example file on Sheet1
An Array
In C2 type: =SUMPRODUCT({10;20;30})
Excel will display 60, which is the Sum of the array elements =10+20+30
A Range
C7: =Sumproduct(C4:C6)
Excel displays 60, which is the Sum of the cells from the range C4:C6 =10+20+30
A Named Range
In the Name Manager or Name Box define a Named Range
MyRange1: =Sheet1!$C$4:$C$6
Then in C10 type:
C10: =Sumproduct(MyRange1)
Excel displays 60, which is the Sum of the range elements =10+20+30
A Named Formula
In the Name Manager define a Named Formula
MyRange2 =OFFSET(Sheet1!$C$3,1,0,3,1)
Then in C12 type:
C12: =Sumproduct(MyRange2)
Excel displays 60, which is the Sum of the range elements from cells C4:C6 =10+20+30
You may be asking why use Sumproduct when we can use a simple Sum to add up 3 numbers?
The answer is to show you what Sumproduct is doing, it is Adding up each Array element.
What about the “Product” part of Sumproduct ?
Remember back at the start where we saw the Definition of Sumproduct,
SUMPRODUCT(array1, [array2], [array3], …)
Only Array 1 is required, Array 2, Array 3 etc are optional, that’s what the square brackets [ ] mean.
Multiple Arrays
Goto Sheet 2 in the Example file:
We will look at a simple example using two arrays
The data consists of Sales data.
Often we want to know what the total sales are
We do this by adding a Sales column
Which multiplies the Qty and Price columns
And then Sum (Add) up this new column
Returning our Total Sales of 15,000
Now we can manually check the above as the numbers are simple eg: 100*20 = 2,000 etc
And we can sum up the Sales and see that we in fact had total sales of 15,000
Well this is exactly what Sumproduct is made to do:
In a Blank cell enter: =SUMPRODUCT(D4:D8,E4:E8)
Excel will return 15,000.
So what is Sumproduct doing?
Lets look inside and see what’s going on
In the Example File, Sheet2, H1 there is a copy of the data laid out as below
Note that our formula =SUMPRODUCT(D4:D8,E4:E8)
Has two Arrays
Array 1: D4:D8
Array 2: E4:E8
Note that each corresponding Array Element is multiplied together
100 x 20
20 x 200 etc
These are the products of the two Arrays
Finally the Products are Added together and the correct answer 15,000 is returned.
So Sumproduct is the Sum of the Products of the Arrays
Of course we can extend that to a large number of Arrays, columns in this case, if we wish.
Sumproduct with Logic
In the above two examples we saw that Sumproduct can Sum a single Array and can Sum the Product of two or more Arrays.
We can use that to our advantage and build logic into the arrays, allowing us to optionally include some array elements and leave out others.
How?
Sumproduct will always add up the product of all Arrays.
So by including an Array where the elements within the Array that we don’t want to Sum are Zero and the Elements within the array that we do want to Sum are 1 we can control what is included in the final Summation.
Goto our Example File on Sheet3
Lets say we only want to include the Sales from our Northern Region
One way to do this is to purely delete the other entries
But what if we could do that without altering our worksheet or there are thousands of rows of data?
This is where Sumproduct comes into its own.
What we need to do is add some logic to our equation, effectively doing:
Lets try it with Sumproduct
In Cell F12: type =SUMPRODUCT(D4:D8,E4:E8,{FALSE;TRUE;FALSE;FALSE;TRUE})
Excel displays a –
Excel doesn’t know what to do with the True/False and so converts them to 0
We can force excel to evaluate these as numbers by adding a simple “1*”
In F14: Type =SUMPRODUCT(D4:D8,E4:E8,1*{FALSE;TRUE;FALSE;FALSE;TRUE})
Excel now displays 5,000 the total sales from the North
To see what has happened in F16 type: 1*{FALSE;TRUE;FALSE;FALSE;TRUE}, but don’t press Enter press F9 instead.
Excel displays ={0;1;0;0;1}
The use of the 1* has converted each of the Array elements from a True/False to a 1,0 respectively.
So our 3 arrays are now:
Now adding an Array of 1*{FALSE;TRUE;FALSE;FALSE;TRUE} every time we wanted to add some numbers isn’t a practical solution.
Excel has the ability to work construct an Array on our behalf!
In E18: enter =SUMPRODUCT(D4:D8,E4:E8,1*(C4:C8=”North”))
Excel will display 5,000
So 1*(C4:C8=”North”) is exactly equal to our previous array 1*{FALSE;TRUE;FALSE;FALSE;TRUE}
1*(C4:C8=”North”) = 1*{FALSE;TRUE;FALSE;FALSE;TRUE}
At the heart of this is that Excel is evaluating each cell in the Range: C4:C8 against our required logic =”North” and setting up an Array for us internally.
Simplify
The power of Sumproduct is therefore in that we can now simplify and extend
In cell E20 type: North
In cell F20 type: =SUMPRODUCT(D4:D8,E4:E8,1*(C4:C8=E20))
Excel will display 5,000
This simple addition allows us to vary the Summation based on the value in E20
We don’t need to multiply our logic array by 1, we can actually use any number or another Array.
In cell F22 type: =SUMPRODUCT(D4:D8,(E4:E8)*(C4:C8=E20))
This works as (C4:C8=E20) is returning an Array of True/False which get converted to an array of 1/0’s when subject to any maths.
The Math in this case is the multiplication by the 2nd Array (E4:E8)*(C4:C8=E20)
In Cell F24 type: =SUMPRODUCT(Qty, Price *(Region=SalesRegion))
Excel will display 5,000
But notice that by using Named Ranges/Formula how simple the logic of the equation has now become.
Rahul’s Question (Multiple Criteria):
In Comment No. 55: Rahul asked, “Can you give an example work sheet of above example”
Sheet 4 in the Example File is the answer.
In Cell C23: type: =SUMPRODUCT(- -(A2:A21=”Luke Skywalker”),- -(B2:B21=”West”),C2:C21)
Excel will display 141, which is the sum of the Sales made by Luke Skywalker in the West Region.
However using what was learned above, this is better simplified to:
C26: =SUMPRODUCT((Name=SalesMan)*(Region=SalesRegion)*Sales)
The Double Unary
In the formula above Chandoo has used what is known as a Double Unary, which is 2 – signs next to each other (I have inserted a space above to make it more legible).
Two – signs are the same as saying
– -(A2:A21=”Luke Skywalker”) = -1 x -1 x (A2:A21=”Luke Skywalker”)
-1 x -1 is 1
Technically this is the most efficient way for Excel to perform any maths on the Array
– -(A2:A21=”Luke Skywalker”)
So that the Array of true/Falses made by (A2:A21=”Luke Skywalker”) is converted to an Array of 1/0’s for use in Sumproduct.
At the slight expense of speed but for improved readability and understandability by others I prefer the use of 1* instead of – – and you will mostly see that convention in my posts.
Chandoo: – –(A2:A21=”Luke Skywalker”)
Hui: 1*(A2:A21=”Luke Skywalker”)
In fact any maths performed on the array will convert its contents to an array of 1/0’s, so long as the maths doesn’t change the Arrays values
For a real good discussion on this topic have a look at the post The Venerable SUMPRODUCT at ExcelHero.com
Other Links to Sumproduct
http://chandoo.org/wp/2009/11/10/excel-sumproduct-formula/
http://chandoo.org/wp/2011/05/26/advanced-sumproduct-queries/
http://chandoo.org/wp/tag/sumproduct/
http://www.excelhero.com/blog/2010/01/the-venerable-sumproduct.html
DOWNLOAD
You can download a copy of the above file and follow along, Download Here.
OTHER POSTS IN THIS SERIES
You can learn more about how to pull Excel Formulas apart and what makes them tick in the following post:
FORMULA FORENSICS NEEDS YOUR HELP !
I am running out of ideas for Formula Forensics and so I need your help.
If you have a neat formula that you would like to share and explain, try putting pen to paper and draft up a Post as Luke did in Formula Forensics 003. or like above.
If you have a formula that you would like explained but don’t want to write a post also send it in to Chandoo or Hui.
XMAS BREAK
This will be the last Formula Forensics Post for 2011, but rest assured that we will be returning in early 2012.
I’d like to take the opportunity to thank Chandoo for allowing me the space and freedom to post pretty much what ever I’ve wanted at Chandoo.org. I hope you have enjoyed my contributions to the Chandoo.org community over the past year.
On behalf of Eva and myself I’d like to wish you all a very Merry Xmas and a Happy and Safe New Year ahead
Hui…
15 Responses to “A Gantt Chart Alternative – Gantt Box Chart”
That's a great idea.
Maybe the planned End Date should be highlight more.
I don't know how it would look like (nor how to do it yet), but what if instead of finishing the bold line to the best case End Date, it finishes to the realistic End Date?
The idea is ok, I think other project management tools have this, already? Maybe not.
Gantt charts in my view are about the signal most unless thing in the world, theres no way you can look at one thats more that a little complex and understand what it's telling you. I'm going to write a diatribe on project management at some point, its one of my pet areas I think!! 😉
The issue I have with this chart Chandoo, is that Tasks need to be linked to each other, so they should inherit the uncertainty, which would mean the as you moved down chart the lines would be miles apart for later tasks, and you might have to add lots of lines for subsequent tasks to cover the various outcome of it's parents.
Having said that, for the high level board summary, it's a nice way to go, it it appeals to the management 😉
thanks Chandoo, great post.
Ross
Whoooa !!! That's a very clever idea Chandoo. I really love it.
I think i'll update my gantt project sheet with that idea soon (remember my template ?)
@ross : you can link start date to the end date of the previous task in your data. The only problem I still se is to which end date (real ? planned ? best ?) in order to have average amount of information.
If best end date, you'll tend to increase uncertainty at the end of chain, although if you link to real end date, uncertainty will be decreased too much, leading in both cases to wrong management direction.
Maybe planned till the task is finished then real will do the job ?
Hey chandoo, this looks good and this would definite add value in production planning / scheduling. Uncertainity in finishing a task is very high in production scheduling and this could give an insight or a bird eye view of possible shipments we can have....
I've always been frustrated by the limitations of gantt charts. Will definitely use this, I've always struggled with how to succinctly communicate the uncertainty of certain tasks without confusing stakeholders.
I like this, I think it's a very effective way of showing how a timeline can change and which parts of a project need close attention.
@Cyril / @Ross: I would intially link the the start date to the planned end date of the previous task, with the chart updating when a task has been completed to reflect the true end date.
Or what about giving a drop-down selection box to allow the user to see the chart based on planned/best-case/worst-case end dates?
Like the idea. Have found that Excel is more flexible than MS Project for graphical solutions. The "Best Case"\"Worst Case" metrics are theoretically appealing but once the project and\or phase commences their reliability diminishes. A chart like the above that showed Planned Start, Planned End, Replan End Start, Replan End Date, Number of Replans the Start and End Dates, and Actual would provide an active, actionable view of each task\phase. It would also highlight the areas which are riskiest.
It is always amazing how flexible excel can be.
My question is how would the chart show a scenario where the date moved up? If a task is dropped or the duration of the task is significantly reduced by applying more people or machinery to the task, the dates will move up.
The gantt chart has been around for a long time, but it is still quite useful to show progress.
Cheers,
B
I like the idea but seems bit complicated in case of long projects involving numerous activity.
Also, reading and explaining is required hence not feasible where plans are just send to audience for approval.
Cheers
SY
Great idea Chandoo,
When I was reading this idea regarding delivery dates, another thought popped into my mind, how can you show the uncertainty with MONEY!!
In this case, applies to cost management or even a normal budget, you think?
Would Box Chart and Gannt Chart help to understand the best case, middle case and worst case when money is spend or planned with these three risks are involved?
I imagine that this chart could help people who write their budgets get a better understanding of risks affecting their spending.
Peter
Chandoo,
I like it. How would you display an entry once it has been completed (actual)?
Thank you,
Matt
From what you have shown so far I think that this box Gantt chart is awesome! I think that this could be an extremely useful tool.
I can't wait to learn how to make my own charts in Excel.
Will the methods that you are going to teach us work in 2003 as well?
[...] Firday, we proposed a new chart for showing project plans. I chose an ugly name for it and called it Gantt Box [...]
You need to read Eli Goldratt's Critical Chain. The uncertainty you are looking for should be accounted for in a project buffer. Not at each task level.
Further you should spend time understanding Agile Development. This would have you plan only in 1-3week iterations. This allows you to embrace changes to work not yet started, and for your customer to re-direct your course at regular intervals (after each iteration) throughout your project. keyword search: Agile Scrum
These items will show you that you are solving a tracking problem for something that you can entirely avoid!
[…] Chandoo.org’s Gantt Box Chart. […]