Discussion


Note: Images may be inserted into your messages by uploading a file attachment (see "Manage Attachments"). Even though it doesn't appear when previewed, the image will appear at the end of your message once it is posted.
Register Latest Topics
 
 
 


Reply
  Author   Comment  
quetzalc0atl

Registered:
Posts: 9
Reply with quote  #1 
Hello,

I've inherited some reports with visualisations that I have attempted to improve and would appreciate any feedback whether there are any further improvements that I have missed.

The first graph shows the number of contacts received by the organisation before and after a date on which some organisational change/restructure took place.

Completed Weekly Contacts Prev Graph.jpg
 
  The second graph shows the outcomes of those contacts.

Contact Outcomes Prev Graph.jpg 

I've decided to break each graph down into multiple graphs so that they are less cluttered. As below, the contact types with the highest number are in the top graph and lowest in the bottom graph.  I am a little concerned about the gradient of the lines in the lowest graph compared to the highest.  Could this be misinterpreted?

Completed Weekly Contacts New Graph.jpg 
I used my limited abilities to do something similar with the second graph again as below. 

Contact Outcomes New Graph.jpg 

Again I am concerned with the gradient of the lines in the bottom graph.  Are axis scales in the bottom sets of graphs too small?  Also think my choice of colour in the Outcomes of completed contacts could be better.  I'm also a little worried by the amount of page real-estate that 6 graphs take up instead of 2 graphs.  Would it perhaps be better to place the two sets of graphs side by side?

Any feedback would be greatly appreciated.

Thanks,

quetzalc0atl

danz

Registered:
Posts: 186
Reply with quote  #2 
quetzalc0atl,

I have a few questions.

Can you provide the data behind these graphs? xls or csv would be just fine.

Besides completed contacts, the rest of them: nonauthorized, cancelled contacts, contacts in draft have any positive sense, or ideally they should be zero?
Is it any sense to stack any values (first original graph) other than to read totals and completed contacts? (are partial totals relevant in that context?)
Is it any logic order of interes for contacts types?
Grouping of contact types, as you did in partial graphs, had any other reason than a clear design?
Are they the totals for 16 months relevant for sorting A, B, C...?


Dan
quetzalc0atl

Registered:
Posts: 9
Reply with quote  #3 
Hi danz,

Thanks for taking the time to respond.

Ideally the contact types other than completed contacts should be zero.  But for example "contacts cancelled" tend to happen from time to time due to human error.  "Bad contact types" that we would wish to be alerted to are mainly "contacts in draft - overdue".

With regards to the stacked bar graph the original idea I'm told that they wanted to see proportionately the variation in contact types over time (week to week).  I guess I had only used the line graphs because I felt it would be easier to read - however it doesn't necessarily show the variation of contacts  each week as easily. So I'm now questioning whether the line graphs I've used are suitable.

I guess it would be useful if the order of the contact types flowed in some form of logical way i.e. Contacts Completed
Contact completed by not authorised
Contacts in draft - ideally the overdue draft contacts should be prioritised over the not overdue ones
Contacts cancelled

The grouping of the contact types was just done to make it a little easier to read. Similarly I grouped the contact outcomes by similar numbers/totals again just to make it easier to read.  Apologies I had to generalise the actual contact outcome names due to their sensitivity.  It would be useful to see the variation in change between A, B, C and D - however I thought any line graph would look too frightening for non data colleagues to infer anything from which is why A and B were grouped together and C and D. Again E, F and G were grouped together due to small numbers.

Sorry not sure what you're asking with your last question "Are they the totals for 16 months relevant for sorting A, B, C...?"?

Thanks,

quetzalc0atl
danz

Registered:
Posts: 186
Reply with quote  #4 
A, B, C,... are values of a nominal variable, which by default has no signifficant order. However when we need to display values of a nominal variable (as scale of a graph or as table), in most of cases we have an associated quantitative variable we use for primary sorting purposes. When we deal with a part of a whole situation (as completed contacts), is always a good idea to display the components in order of the values, largest first. Can be the total amount of completed contacts (for whole period) a default order for A, B, C?

jlbriggs

Registered:
Posts: 194
Reply with quote  #5 
I think that if the goal is to show the change that happened as a result of an organizational change at a specific time, then that is what you really need to focus on - instead of breaking categories up by how big their numbers are, look at each series individually, and analyze the difference pre-change vs post-change.

There are various means tests and other statistical analysis that should be part of any look at this difference, which can be displayed as a set of box plots for each series.  You could also plot it according to "proportion of total before change", and "proportion of total after change"  or something similar.

quetzalc0atl

Registered:
Posts: 9
Reply with quote  #6 
danz - let's go with what you've suggested, the total amount of completed contacts.

jlbriggs - thank you for the input - I hadn't thought of analysing pre-change vs post-change. I'm a little worried that box plots might be seen as a little to "advanced" for colleagues who are going to be interpreting the data as some struggle with line and bar graphs as it is.
jlbriggs

Registered:
Posts: 194
Reply with quote  #7 
The "advanced" concern is certainly understandable.

I have personally found that the best way to address the problem is to simply use the charts. :)

A box plot usually takes less than a minute to explain in simple terms, and the more users see them, the more they are discussed in meetings, the faster they become an extremely useful tool.

There isn't usually a more succinct way to display a distribution, so sticking to lines and bars alone often won't tell the story.

Obviously how well this works is very dependent on your organization.

Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.