Subscribe Bookmark RSS Feed

Best Visual?

Swalo66

Community Trekker

Joined:

Apr 28, 2015

All,

 

I am looking to see how you all might use JMP ro visualise the data below. 

Each year shows the rolling total required.

For example, Role A adds 2 head in 2018, then 2 more (4 total) in 2019, then 4 more (8 total) in 2020 etc.

 

Role/Year

2017

2018

2019

2020

2021

Role A

0

2

4

8

16

Role B

2

4

6

9

12

Role C

1

4

6

8

10

Role D

0

2

6

6

6

Role E

0

2

3

4

5

Role F

0

1

2

4

5

Role G

0

2

4

4

4

 

I can make really basic visuals in ppt (see pic attached), but am looking for more snap.

Should I consider transposing data? Would it work well as a tree map...?

 

I am completly open to suggestions...Headcount.JPG

4 ACCEPTED SOLUTIONS

Accepted Solutions
M_Anderson

Staff

Joined:

Nov 21, 2014

Solution

A couple of ideas  - run the scripts in the attached data tables.

 

M

pmroz

Super User

Joined:

Jun 23, 2011

Solution

How about a stacked bargraph?

StackedBar.png

 

Here's the JSL code:

Graph Builder(
	Size( 533, 448 ),
	Show Control Panel( 0 ),
	Variables( X( :Year ), Y( :Count ), Overlay( :Role ) ),
	Elements( Bar( X, Y, Legend( 11 ), Bar Style( "Stacked" ) ) ),
	SendToReport(
		Dispatch( {}, "Count", ScaleBox, {Label Row( Show Major Grid( 1 ) )} )
	)
);
Peter_Bartell

Joined:

Jun 5, 2014

Solution

I agree 1000% with my colleague @M_Anderson and JMP User and friend @pmroz. Think about your message first...get cute later. If your message is one of rate of change across roles over time, then perhaps the attached overlay plot or some variation within Graph Builder might be effective? Perhaps laying on a local data filter once the graph is created as well?

dale_lehman

Community Trekker

Joined:

Jan 29, 2015

Solution

I'm agreeing with others here - snap is out, understanding is in.  I've used M_Anderson's file and added four more scripts.  One (the treemap) shows how the overall size and roles shift over time.  The second is the cleanest display of the data - but a but too busy given the relatively large number of roles displayed on one graph (you could always put the roles into wrap, as the next script shows).  While it clearly shows the data without need for much effort, it is probably the furthest away from "snap" you can get.  The fourth is the stacked bar chart already shown - it is an efficient display, but I'm not too fond of stacked bar charts as only a few major features are readily interpreted by your eyes - much of the detail is simply not comprehensible given the varying sizes of the bars.  If you wanted to emphasize the makeup by roles, you could always standardize the totals and show the percent in each role.  In that sense, I think the treemap may be best here since it shows both the makeup and the overall size.

 

To get better displays, I think you need to spend the effort on thinking about the question you want the graph(s) to answer.  You may also be able to combine the roles into a smaller number since some of the role staffing seem to behave similarly and some seem less important than others.

8 REPLIES
M_Anderson

Staff

Joined:

Nov 21, 2014

Solution

A couple of ideas  - run the scripts in the attached data tables.

 

M

M_Anderson

Staff

Joined:

Nov 21, 2014

Also, as a bit of a follow-up: A really good rule of thumb is not to try and make a visual "snap." The experts in Data Vis will always tell you to focus first on the story. The best visual is the one that most effectively conveys the information with the least amount of confusion.

So, while I gave you a couple of ideas in Graph Builder, the better question would be: "What specific finding from the data are you trying to communicate with the visual?"

Best,
M
pmroz

Super User

Joined:

Jun 23, 2011

Solution

How about a stacked bargraph?

StackedBar.png

 

Here's the JSL code:

Graph Builder(
	Size( 533, 448 ),
	Show Control Panel( 0 ),
	Variables( X( :Year ), Y( :Count ), Overlay( :Role ) ),
	Elements( Bar( X, Y, Legend( 11 ), Bar Style( "Stacked" ) ) ),
	SendToReport(
		Dispatch( {}, "Count", ScaleBox, {Label Row( Show Major Grid( 1 ) )} )
	)
);
Peter_Bartell

Joined:

Jun 5, 2014

Solution

I agree 1000% with my colleague @M_Anderson and JMP User and friend @pmroz. Think about your message first...get cute later. If your message is one of rate of change across roles over time, then perhaps the attached overlay plot or some variation within Graph Builder might be effective? Perhaps laying on a local data filter once the graph is created as well?

dale_lehman

Community Trekker

Joined:

Jan 29, 2015

Solution

I'm agreeing with others here - snap is out, understanding is in.  I've used M_Anderson's file and added four more scripts.  One (the treemap) shows how the overall size and roles shift over time.  The second is the cleanest display of the data - but a but too busy given the relatively large number of roles displayed on one graph (you could always put the roles into wrap, as the next script shows).  While it clearly shows the data without need for much effort, it is probably the furthest away from "snap" you can get.  The fourth is the stacked bar chart already shown - it is an efficient display, but I'm not too fond of stacked bar charts as only a few major features are readily interpreted by your eyes - much of the detail is simply not comprehensible given the varying sizes of the bars.  If you wanted to emphasize the makeup by roles, you could always standardize the totals and show the percent in each role.  In that sense, I think the treemap may be best here since it shows both the makeup and the overall size.

 

To get better displays, I think you need to spend the effort on thinking about the question you want the graph(s) to answer.  You may also be able to combine the roles into a smaller number since some of the role staffing seem to behave similarly and some seem less important than others.

Swalo66

Community Trekker

Joined:

Apr 28, 2015

Treemapdata.png

Dear all,

 

Firstly, let me thank you for your insightful inputs.  This was a good challenge (personally) since it involved, IT, Marketers and analysts...pure heaven on a committee :-)

 

While the goal was straight forward (at least in my mind) we did get into the weeds of the visual...

 

The current candidate (for additional massage) is attached.  A good composite of several comments from @M_Anderson @pmroz     

@dale_lehman and @Peter_Bartell.  Thank you all...

 

 

 
Peter_Bartell

Joined:

Jun 5, 2014

A few other thoughts for you...that are my way of 'doing things'...I'm not saying it's the ONLY way...but it's how I roll.

 

When I create a graphic I start with two questions and one assumption: Assumption: The graphic should allow the ONE key audience member to become informed or make a decision. Then the two questions: 1. Who is the key audience member in the room? You cite that you have many people in the room...but there is always ONE person who is the top person...either the 'decider' or 'most influential person' in the room. Question 2: "What's the SINGLE message I need that ONE person to understand (or decide) based on the graphic?"

 

I don't try to co-mingle messages within a single visualization. Co-mingling messages for multiple constituencies leads almost invariably to chart junk, overcrowding, too much ink, and confusion for all...And if my visualization doesn't convey that message sans my presence...I need to go back to the drawing board to refine and modify.

 

I do like that you are using JMP's Graph Builder as your core creation tool...Graph Builder is in large measure designed to be a 'scratch pad' upon which one can quickly create a variey of visualization ideas...and find the perfect one.

Swalo66

Community Trekker

Joined:

Apr 28, 2015

@Peter_Bartell indeed...the identifiction of the Economic Buyer (EB) as we would term them.  The one who can still say yes, when many say no.

 

Thanks again.