I recently had to perform some analysis on a set of insurance companies in certain geography. After searching the net I found such list of companies operating in that continent. But the problem is,
- These companies are listed multiple times, one time for each of their geographical area of operation.
- I cannot count each company only once since there is geography specific operational data in that list.
- But at the same time, information pertaining to organization like total sales, strategy etc. are common to all the subsidiaries / separate entities of a company.
- There are too many companies to do the manual grouping of companies
I think this type of problems are fairly common in business analytics. So here is a relatively simple solution for getting unique list of companies without losing any information or writing macros. See the example for yourself.
Lets take an example of employee data. You have fields like Person Name (sometimes unique), Previous employment, Previous workplace, Previous Salary. Now, if person Anand worked in more than one place earlier, there will be more than one rows with his name. But for details like DOB, SSN etc. there wont be multiple rows. So you need to know how many uniques rows are there in that huge list. Detailed Steps:
- Enter / copy the data
- Sort the list on person name
- In the column next to prev-company enter the following formula in E2
=IF(OR(LEFT(B2,SEARCH(" ",B2))=LEFT(B3,SEARCH(" ",B3)),LEFT(B2,SEARCH(" ",B2))=LEFT(B1,SEARCH(" ",B1))),LEFT(B2,SEARCH(" ",B2)),B2)
Copy – paste the formula in all the cells in the list for the column. You will see the followin result.
- Now create one more column next to Unique Name with heading Unique?. Paste the following formula in the F2. Apply the formula to all the cells in the list.
=IF(E2=E3,0,1)
Now, you would see the list like this.
- Now sellect the top row and apply “auto filters” [Data->Filter->Auto-Filter] And select 1 in the Unique? column. You will see all the unique names. You can copy paste this list in another sheet/work-book and work with it or assign corresponding codes to each of the unique items in this list.
To summerize:
1. Sort your list first.
2. Get the unique parts like first word / first number etc in another column. In my case I had to use first word.
3. Compare consecutive rows and mark the differences. Now you know how many unique items you have.
Comments / Any better ways of doing this are always welcome.
14 Responses to “Group Smaller Slices in Pie Charts to Improve Readability”
I think the virtue of pie charts is precisely that they are difficult to decode. In many contexts, you have to release information but you don't want the relationship between values to jump at your reader. That's when pie charts are most useful.
[...] link Leave a Reply [...]
Chandoo,
millions of ants cannot be mistaken.....There should be a reason why everybody continues using Pie charts, despite what gurus like you or Jon and others say.
one reason could be because we are just used to, so that's what we need to change, the "comfort zone"...
i absolutely agree, since I've been "converted", I just find out that bar charts are clearer, and nicer to the view...
Regards,
Martin
[...] says we can Group Smaller Slices in Pie Charts to Improve Readability. Such a pie has too many labels to fit into a tight space, so you need ro move the labels around [...]
Chandoo -
You ask "Can I use an alternative to pie chart?"
I answer in You Say “Pie”, I Say “Bar”.
This visualization was created because it was easy to print before computers. In this day and age, it should not exist.
I think the 100% Bar Chart is just as useless/unreadable as Pies - we should rename them something like Mama's Strudel Charts - how big a slice would you like, Dear?
My money's with Jon on this topic.
The primary function of any pie chart with more than 2 or 3 data points is to obfuscate. But maybe that is the main purpose, as @Jerome suggests...
@Jerome.. Good point. Also sometimes, there is just no relationship at all.
@Martin... Organized religion is finding it tough to get converts even after 2000+ years of struggle. Jon, Stephen, countless others (and me) are a small army, it would take atleast 5000 more years before pie charts vanish... patience and good to have you here 🙂
@Jon .. very well done sir, very well done.
good points every one...
I've got to throw my vote into Jon's camp (which is also Stephen Few's camp) -- bars just tend to work better. One observation about when we say "what people are used to." There are two distinct groups here (depending on the situation, a person can fall in either one): the person who *creates* the chart and the person who *consumes* the chart. Granted, the consumers are "used to" pie charts. But, it's not like a bar chart is something they would struggle to understand or that would require explanation (like sparklines and bullet graphs). Chart consumers are "used to" consuming whatever is put in front of them. Chart creators, on the other hand, may be "used to" creating pie charts, but that isn't an excuse for them to continue to do so -- many people are used to driving without a seatbelt, leaving lights on in their house needlessly, and forwarding not-all-that-funny anecdotes via email. That doesn't mean the practice shouldn't be discouraged!
[...] example that Chandoo used recently is counting uses of words. Clearly, there are other meanings of “bar” (take bar mitzvah or bar none, for [...]
[…] Grouping smaller slices in pie chart […]
Good article. Is it possible to do that with line charts?
Hi,
Is this available in excel 2013?