Saturday, January 17, 2009

4D Graphing technique for KPI visualisation

Data visualisation techniques have been an area of research since so many years. Veterans like Dr.Edward Tufte have written several books on these subjects.

Most effective data visulaisation technique will use minimum ink, represent maximum amount of data and convey the message EFFECTIVELY to the viewer with out third party explanations. In other words, the graph will speak for itself. Simplicity and pleasing appearance (Shapes & Shades) are icing on the cake.

Graphs are to be chosen on case-to-case basis. It is up to the data visulaiser to decide what he wants to convey and choose the graph accordingly.

Those who are lazy and those who don’t care about the end users will supply them with a standard table. Most users, especially senior management won’t be interested in reading and apprehending long data tables. Even if they go through the table table taking lot of time, they may not be able to come up with some conclusions, which fail the basic intention of reporting itself!

Lets take a business case in Project Management. Earned value management (EVM) is one of the greatest tools in Project Management which allow project team to assess project performance, draw conclusions and implement remedial actions for slips and overruns. It is kinda ‘Early Warning System’. CPI (Cost Performance Index) and SPI(Schedule Performance Index) are two major indices measured in EVM.

Business Case:
In company XYZ Inc, there are lots of projects executed simultaneously. Contract value of these jobs differs significantly. Profitability of the company greatly influenced by few major jobs. Now XYZ Inc has got a very good system to monitor progress & costs for each project. CPI was reported on weekly basis for all the projects by the Project Control Engineer in the following format:
There were 100’s of ongoing projects in the company and hence the table was too big. A small decrement in CPI of a big job has serious effect on overall performance of the company. How ever such critical jobs lost somewhere in the crowd of 100’s of jobs and often left un-noticed! Several bad performing small jobs also affect the profitability if not controlled on a 1-to-1 basis.

So what is the solution, which shows the ‘SIZE & SEVERITY’ of each job in single graph?

PM Karma has a solution for this. Take look at the 4D-Graph for KPI’s given below:
The bubble plot is an X-Y graph , but here it is not just showing 2 –parameters like an ordinary X-Y graph or 3 –parameters like a bubble graph. Each point in the above graph bears 4 parameters:

1.    Project ID (Y-Axis)
2.    Progress % (X-Axis)
3.    Budget value in $ (Size of bubble)
4.    CPI (Colour of bubble)

Result :
Now the management will never miss a big job with a bad CPI , since it will be represented as a ‘Big’ bubble with red colour ! The relative sizes of bubbles & colours aid the senior management to zero in ‘real trouble makers’.

While the first graph (Bubble graph) represent snapshot ‘at the moment’, the line graph given at the bottom provides trend over a period of time (weekly, as given here).

The above graph is generated using Microsoft® Excel ™. I have given different colours to bubbles by creating ‘Series’ with ranges.

The above technique can be used for other KPI’s also like SPI, Productivity ...etc. Spread your imagination, you may be able to find better ways. If you have something better , please share.
Of late, I have been doing a lot of research on unconventional data visualisation techniques. This again goes with my strategy 
“If everyone were doing things in the same way others did, the world wouldn’t have been progressed , ever!’
Disclaimer: XYZ Inc ,job numbers and budget values are imaginery and real life situations may need alternate representation/customizations.


  1. Life in the macro world isn't so easy !
    It is better to manage 1 Huge project than managing 100 small projects!

  2. Good Work Sreejith!
    Very nice explanation, it really helped alot.
    Now i'll start using this technique in my office work!

  3. Hello, Sreejith! I also need to graph four different sets of data in one graph.. I'm just a high school student though, so could you please tell me what you did step-by-step? The more detailed, the better :D Thanks a lot!


Please post your comments here. I read them personally and reply , if needed.