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  
adahbache

Registered:
Posts: 1
Reply with quote  #1 


Hi - I have a (hopefully quick to anwer) visualation question. 


When working on a dashboard, what is best practice around displaying coloured trending arrows?

In our current dashboards, a 'Down' arrow normally denotes negative trending.


Up normally denotes positive trending


 However, how about situations where an arrow pointing down is actually a good thing (like costs going down)?  Would you put an arrow pointing down highlighted in green indicated a downwards trend but favourable result?


My question is - have there been any articles written that summarise best practice for trending arrows?

jodwilso

Registered:
Posts: 21
Reply with quote  #2 
I'd use red for bad
acraft

Registered:
Posts: 51
Reply with quote  #3 
1. Hopefully the person using the dashboard understands the data well enough to know whether up/down is good/bad.

2. I would keep the arrows somewhat muted for expected changes; you don't want brightly-colored arrows all over the screen as important changes might not be as obvious (which is important on a dashboard). Then, make critical changes (you'll have to decide what constitutes "critical") obvious by coloring the arrows, red for bad and, only if it's useful, green for good.

3. If you can use sparklines that show an upward/downward trend over time, that may be another option to consider.
acraft

Registered:
Posts: 51
Reply with quote  #4 
So I completely missed that you had a specific question - I don't know of any articles offhand, but I do know that Stephen's book "Information Dashboard Design" makes note of trend arrows when introducing sparklines (Chapter 6). It seems the recommendation is to avoid them if possible.
jlbriggs

Registered:
Posts: 194
Reply with quote  #5 
Whenever I use trend arrows, which I do sparingly, I use a medium grey color regardless of the direction or the meaning of it.

As stated above, it should usually be clear whether up is good or bad (we all understand that higher cost is bad, higher profit is good, etc...).

"Would you put an arrow pointing down highlighted in green indicated a downwards trend but favourable result?"

No, never.  I also would not make the arrow red to indicate 'bad', because, again - that should be clear. 

The more green and red blobs of color on the page, the less attention anything important receives.
wd

Registered:
Posts: 167
Reply with quote  #6 
Colour may also have cultural implications.  We in NA & Europe would generally see red as bad.  If I'm not mistaken, red is seen as an indicator of good in China. 

But, why use a trend arrow at all?  A series of points (bars) going up or down is self-evident.  More importantly, is the "trend" significant (my statistical roots are showing!) in the face of the normal everyday(week/month/year) variation?  The Western Electric rules for statistical control chart interpretation do not trigger an signal until there are more than 5 consecutive points steadily increasing or decreasing. 

Given that a trend is "a prevailing tendency or inclination"  (http://www.merriam-webster.com/dictionary/trend ) and not the one game winning streak that so many like to attach an arrow to, I'd be more inclined to show that the trend is statistically significant (if it is) than to highlight its (self-evident) existence.




__________________
Bill Droogendyk
revans00

Registered:
Posts: 2
Reply with quote  #7 
I am new to this forum (..where has this been my whole life....? <GRIN>) and found this specific topic very apropos to a current effort of mine.

I am developing a visualization of a sporadic activity over time; specifically the work queue for a department of people.  The volume is low-to-medium and inconsistent, and I am interested in the time to complete a work request. There may be several days of no work completed or there may be 5-10 items completed in a single day.  The time to complete the work typically is in the 3-5 day range but may be a low as same day (zero: 0) to 20 days or more.

My approach (sorry, I can't post the visualization) is to calculate the 30-day moving average (MA) of the data and compare the current (today) MA relative to the MA of the last day of the prior month.

When presented as a line, the direction is clear, and in this case "down" (= smaller /faster times) is good.  However the rate of change (velocity) is not as obvious so I add a color indicating the trend:
- dark green = significant improvement (> 20% decrease in time)
- light green = moderate improvement (> 10%)
- grey = no significant change  (+/- <10%)
- light red = moderate degradation (> 10% increase in time)
- dark read = significant degradation (> 20%)

I've read several papers on treatments for visualization of sporadic / intermittent data.  They seem to be primarily around material inventory / stock and delivery which doesn't seem to apply here.

Any comments, observations or suggestions welcome!


jlbriggs

Registered:
Posts: 194
Reply with quote  #8 
I think you'll have to post an example to get any real feedback.

It's hard to picture exactly what you're describing.
revans00

Registered:
Posts: 2
Reply with quote  #9 
Understood, I'll see what I can do (when I get back from vacation).

Thanks!
whizzdome

Registered:
Posts: 3
Reply with quote  #10 
Here's my take on colours:
(1) The cultural aspect (as someone has mentioned elsewhere, in some cultures red is a good thing, for example).
(2) As Stephen Few points out many times, 10% of males and 0.5% of females are colour blind and would not be able to tell the difference anyway.

For what it's worth: in one screen-based report I created a few years ago, the client insisted on having trend arrows and wanted to be able to distinguish "good" from "bad", but /two/ of the stakeholders were colour blind. My solution was to use small triangles pointing up and down, and to make them solid for good and just outlines for bad, something like this:
Clipboard01.jpg 

The client was happy but, as someone else has said, the sparklines told a better story, and I would have preferred simply to have marked the bad (with a small filled circle or other marker) and left the good unmarked -- if it's a dashboard you want to be able to tell where the bad areas are at a glance, and don't necessarily want to be distracted by "good". If /everything/ needs to be marked then in my opinion it isn't a dashboard, it's a report.



wd

Registered:
Posts: 167
Reply with quote  #11 

Given the diffferences in work effort for the various jobs, number of jobs done in a day doesn't seem to be a useful signal. If the data could be normalized in some way, the plotted indicator would have more meaning. 

Are estimates made for time to complete before the job starts?  If so, a periodic report of proportion completed on time may be useful.

Is there a queue of jobs waiting?  Would length of time in the queue before the job starts be of interest? 

Number of jobs in the queue on given day of the week?  Number of jobs coming in per day?  An example of this is attached.  I don't have an original (it was better!).  The upper line is the queue, the lower is number of jobs coming in.  Hope this gives you some ideas.

Capture.JPG   



__________________
Bill Droogendyk
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.