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  
jrodriguez

Registered:
Posts: 18
Reply with quote  #1 
Hey guys,

We're having an internal debate in my company over which graph type to use....a YOY graph, or an LTM graph. With the information given, which one would you find more useful?
https://promachinc.co1.qualtrics.com/SE/?SID=SV_9AGewBMz8ER5R09

These are the 2 graphs in question....thanks for the feedback guys.


Attached Images
Name: YOY.png, Views: 56, Size: 63.83 KB

Name: LTM.png, Views: 56, Size: 58.99 KB


sfew

Moderator
Registered:
Posts: 802
Reply with quote  #2 
Joe,

As you might expect me to say, there is no right answer to your question without understanding the purpose of the graph. Both graphs are useful, but for different purposes. What do people do with this information? What are they expected to understand?

__________________
Stephen Few
jrodriguez

Registered:
Posts: 18
Reply with quote  #3 
Hey Steve, 

This is a tough one as there is some variation between users, but from my conversations with them so far, the graph will be generally used for driving action to address negative trends and maintain continuous improvement of this metric (I know, very vague). There will be some drill through functionality, tooltip use etc to help them do this. The users would also also like to use the graph for reporting print outs.

It's somewhat of a toss up from what I have heard so far. Both seem like they could do the job, but I think one of them delivers more/better information than the other without misusing data-ink. 

It was great meeting you last year.

Joe
sfew

Moderator
Registered:
Posts: 802
Reply with quote  #4 
Hi Joe,

Rather than giving you answers, I'd rather help you learn to see for yourself what works and why. Both of your graphs display the same exact set of time-series values, but in different ways with different best uses. One of the two excels as a way to compare the pattern of change in 2015 to the pattern in 2016. Can you see which one does this? One of the two graphs excels as a way to see the pattern of change across the entire period of time beginning in January of 2015 and continuing into 2016. Can you see which one does this? Can you see why the other graph does not do this as well?

I mentioned above that these two graphs display the same exact series of values. They should, but in fact they do not. There is a discrepancy between the two. Notice that in your first graph the highest value is approximately 60%, but in the second the highest is only 53%. What accounts for the difference?

__________________
Stephen Few
sfew

Moderator
Registered:
Posts: 802
Reply with quote  #5 
Joe,

In case it isn't obvious, let me make another point. You are not forced to choose between these two graphs. If both are useful for two different purposes and people need to use them for both of those purposes, then both should be provided.

__________________
Stephen Few
jrodriguez

Registered:
Posts: 18
Reply with quote  #6 
I see what you mean Steve, thank you. YOY also makes it a bit more difficult to see a pattern during the first few months of the year. Something tells me that comparing years will be more useful, but these assumptions as you well know can lead to trouble.

In doing user research for something like this, would you recommend asking the actual users to select which graph they would find most useful in improving the KPI? To my earlier point, I think this will be a toss up... so I will consider developing both views so they can simply select YOY or LTM at the top of the page. There are several of these % metrics, so I won't be able to fit both views all in one page.
sfew

Moderator
Registered:
Posts: 802
Reply with quote  #7 
Joe,

I often hear the argument that only one view of the data can be provided because there isn't space for more. This argument is often invalid. Assuming that you really cannot fit both graphs on the screen or page, such as by making them smaller, why are you imposing the limit of a single screen or page on yourself? Graphs only need to appear on the same screen or page when they must be viewed together (e.g., for the purpose of making comparisons).

When doing the research to determine what to display and how to design it, you must learn what people do with the data. Ask questions that will lead to this understanding. Asking them how to design the graph would only be useful if they are skilled in graph design. If they were skilled in graph design, they wouldn't need you. You're responsible for bringing graph design expertise to the process.

You didn't answer my question about the discrepancy in values between the two graphs. Why are the values different? If your data is wrong, it won't matter how well you design the graphs.

__________________
Stephen Few
jrodriguez

Registered:
Posts: 18
Reply with quote  #8 
I'll need to start putting some of these together and see what it looks like. I am still at the beginnings stages, so I will keep your point in mind as I move along.

On a day to day basis, the idea is to draw attention (Through red dots like you use) to a good or bad current month. I am still determining if its vs previous month, prior year month or vs. a more robust/significant previous 3+ month rolling average. Either way, the user would click and land at a list of specific orders driving those results that month. The drill-through page would list orders with the lowest margins ranked up top. I did confirm this with the user group. They can do this with any given month shown in the line graphs.

I don't see a discrepancy here Steve, unless you're hinting at something I'm missing. I see Feb 2015, which was at about 60%, moving out of the time series in the LTM graph and the scale adjusting itself. This would be a great month to click on to understand what drove that for example. +1 for YOY? I do notice a slight mis-alignment (discrete vs continuous?) in the month scales between the two graphs.
sfew

Moderator
Registered:
Posts: 802
Reply with quote  #9 
Joe,

What I thought was a discrepancy in values turns out to be a difference in the time periods instead. The 60% value for February that stands out in the first graph doesn't appear in the second graph because it doesn't include January through May of 2015. I was mistaken when I said that the two graphs included the same values. The greater period of time in the first graph would offer advantages over the second graph if more than 12 months of history is useful. Rolling periods of time are useful, but when we use them we need to make sure that that extent of the period shown really is all that's needed. Assuming that only 12 months are needed is often a mistake.

You mentioned that you are causing red dots to appear next to displays the contain a good or bad current month, but that you're uncertain what you should compare the current month (previous month, same month last year, etc.) when determining if the dot should appear. Its good that you're thinking about this. Too many alerts, including alerts attached to items that don't actually deserve attention, cause people to ignore alerts, which renders them useless. Only alert people to conditions that matter and that they can do something about. Given the volatility of the monthly values that appear in these particular graphs, you definitely shouldn't cause an alert to appear merely because the current month is less than the previous month. Assuming that it is more appropriate to compare the current month to the same month last year, even then you shouldn't cause an alert to appear simply because the current year's value is less than last years. You should determine the degree of variation that constitutes a true signal of meaningful change and only cause an alert to appear when that threshold is exceeded. Don't encourage people to chase noise in the data.

__________________
Stephen Few
jrodriguez

Registered:
Posts: 18
Reply with quote  #10 
I'm glad you said that. The use of alerts is something that I have been improving at, in part thanks to you. Thanks again Steve, I'm humbled by your help and feedback. The learning never ends!

wd

Registered:
Posts: 167
Reply with quote  #11 
Joe, I have an idea, but I'd like to have the actual YOY data points. Could you post please?
__________________
Bill Droogendyk
jlbriggs

Registered:
Posts: 190
Reply with quote  #12 
Depending on purpose, I have very widely used both methods, and have also gone as far as to plot Previous 12 Months, with a second line comparing each month to its previous year.

I have most often found that it's best to pick a default, depending on the more common analysis tasks (or the common corporate reporting strategy), and allow a simple click to change the view, either on the specific chart if only one, or on the dashboard level.
Previous Topic | Next Topic
Print
Reply

Quick Navigation: