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  
peter_mikael

Registered:
Posts: 18
Reply with quote  #1 
Hi

I am doing an analysis of sales in different retail channels. I want to create one chart per channel. For each product in the chart I want to display two things:
1. Total sales in the channel
2. How much of the products total sales (across all channels) are sold in the channel

I want to emphasize the outliers in each channel.

We have 23 products which I need to include (not all products are sold in all channels). How can I visualize this in a good way? My best shot so far is below but I am not happy with it e.g. with the colors for the products. 

  sales.PNG   
Data below

ProductTotal sales 2015% of the products total sales sold in channel
Product 11,16E+080,313269
Product 297271430,309436
Product 3329494680,288946
Product 4234574420,316991
Product 5129694100,384818
Product 61,62E+080,306165
Product 7660356790,336052
Product 800
Product 914506190,291565
Product 103,03E+080,17428
Product 11157203060,234025
Product 1200
Product 131,2E+080,250032
Product 143,79E+080,24864
Product 15767198460,327106
Product 16243347370,306768
Product 17760958630,25429
Product 181,43E+080,467577
Product 1934673110,015547
Product 2000
Product 2100
Product 2200
Product 2300

Jeff

Registered:
Posts: 53
Reply with quote  #2 

Hi, peter_mikael. If you want to emphasize outliers, this is a good start. One idea would be to abandon trying to identify every product in each graph. Make the clustered dots one neutral color and the outliers a bolder one. Add labels directly on the graph, next to the data points, just for the outliers. Then remove the grid lines to increase overall readability and the now-unnecessary legend.

Hope this helps.

Jeff

sfew

Moderator
Registered:
Posts: 812
Reply with quote  #3 
Peter,

If you want the identity of the products to be known, your scatter plot won't do the trick. Some of the colors in the legend are duplicates and some that aren't are too much alike to distinguish. Here's a suggestion: display sales revenues in a horizontal bar graph and sort the bars from the highest value at the top to the lowest at the bottom. Display the percentage sold through the channel either by highlighting the portion of each bar that was sold through the channel (you can do this using a stacked bar in Excel) or as a separate horizontal bar graph to the right of the first.

__________________
Stephen Few
jlbriggs

Registered:
Posts: 194
Reply with quote  #4 
My first thought was the same as what Stephen outlined.

I think that with 23 products, and presumably a small number of channels, and not all products being sold in all channels, you're just not going to have enough of a pattern visually or statistically to start calling out outliers in any meaningful way.

Another benefit to the multiple bar chart approach is that you can add another column containing a bar chart for each (or at least several) channels - that way you can see the relationship between items in terms of overall performance, and their performance within each channel all at the same time.

peter_mikael

Registered:
Posts: 18
Reply with quote  #5 
Thanks for good suggestions. I went with the show only outliers suggestion. What do you think?

Chart v2.png 

sfew

Moderator
Registered:
Posts: 812
Reply with quote  #6 
Peter,

Are you suggesting that all of the data points in your scatter plot are outliers or only the ones that are labeled? Either way, how are you determining what qualifies as an outlier. "Outlier" is a statistical term. Values qualify as outliers when they fall outside of the range that has been designated as routine based on some statistical measure variation.

__________________
Stephen Few
jlbriggs

Registered:
Posts: 194
Reply with quote  #7 
I have to be honest, I am struggling to find a lot of good value in this method of display.

It doesn't show you which items have zero sales, or which items have (nearly?) zero sales, and yet have 4-5 % of those sales in the channel.

And as has been discussed, determining what qualifies as an outlier by identifying which items are far enough apart to have their own label isn't really going to pan out in the long run.

I put together a quick random example of the direction I would try to go with this data, in case it may be helpful in illustrating why I find it to be a better method:

product sales example.png 

For me, the clear benefit is that it allows you to compare items to each other, and across sales channels at the same time.
It also allows each item to be fully visible and identifiable, without needing a cumbersome colored legend.

In a web format, or even in a table in excel, it could be made to be sortable by any of the columns to allow different prioritization.



peter_mikael

Registered:
Posts: 18
Reply with quote  #8 
Quote:
Originally Posted by sfew
Peter,

Are you suggesting that all of the data points in your scatter plot are outliers or only the ones that are labeled? Either way, how are you determining what qualifies as an outlier. "Outlier" is a statistical term. Values qualify as outliers when they fall outside of the range that has been designated as routine based on some statistical measure variation.


My idea was to plot all products but only label the ones of interest (what I called outliers but that is probably not the correct term).
peter_mikael

Registered:
Posts: 18
Reply with quote  #9 
Excellent suggestion jlbriggs. A scatter plot is maybe not the way to go.

The main point I want to make is this:
1. Are there any products which are sold in significantly higher or lower volumes in supermarkets vs other channels.

2. If so, do those products represent significant revenues?

Would it be advisable to use another color to highlight the channel/product combinations I want to draw attention to (in your small multiples suggestion)?
peter_mikael

Registered:
Posts: 18
Reply with quote  #10 
I like jlbriggs suggestion. The problem is that we have 7 channels and it may be hard to know which product a certain bar belongs to for channel 7. What is a good way to make it easier to know which bar belongs to which product for the channels to the right? More white space between products? Repeat labels to the right? A thin grey line below each row?
jlbriggs

Registered:
Posts: 194
Reply with quote  #11 
Peter - being a quick example made in excel, the image I posted is not ideal in all of its details :)

The primary thing that I would do to make it more easy to see what bars belong to what product as to add a little more white space between rows.

In most cases, white space is really all that is needed. 
If it proves to truly not be enough, I would consider adding very faint, very thin, separation lines between rows - but just like axis grid lines, it's very important that the lines are not visually prominent. It doesn't take much to guide the eye along the line, and it doesn't take much more to become distractingly prominent.

Alternately colored rows is not generally something I would suggest, though when necessary, and also done very subtly, it can be an option.

If supermarkets are the channel that you want to draw attention to, then it's perfectly reasonable to make the bars of that channel a separate color, though I would again advise a subtle variation (a darker grey than the others, or perhaps a more subtle blue than the total sales).  I would definitely not make each channel a different color.

If you want to highlight a specific product that meets your criteria of standing out, I would probably prefer to see an indication in the form of a marker next to the product name, or perhaps the product name in bold, or something like that.  But I guess I'd have to see the whole set up, and what qualifies as standing out to really judge.








Previous Topic | Next Topic
Print
Reply

Quick Navigation: