Dot plots are easily created and offer simple, fast insight when comparing two or more data sets, without performing a single calculation.
Dot plots are easily constructed: each dot represents a data point along an axis, and dots of the same value are stacked on top of one another. Dot plots look very similar to histograms, but are easier to construct and can be more valuable for generating clues into potential causes.
Creating a Dot Plot – 5 Easy Steps
Dot plots are most easily created by hand, or an excel template can be used. Here are the basic steps for creating a dot plot:
- Assemble your data, which will comprise two or more groups (in the above case, the two data groups are “Day Shift Downtime” and “Night Shift Downtime,” in minutes.
- Note the minimum and maximum values of each data group. In the example above, the day shift data ranged from about 32.2 to 40 minutes, and night shift ranged from 26.5 to 37 minutes.
- Note the overall minimum and maximum value for the plot, which will be the smallest minimum and the largest maximum from step 2. In the above case, the overall minimum is 26.5 minutes, and the overall maximum is 40 minutes.
- Draw a horizontal line the full width of the chart (piece of paper, easel pad, white board, etc), and note the overall minimum value from step 3 on the left-hand side, and the overall maximum value from step three on the right hand side.
- Add a “dot” for each data point, and stack the data groups above one another, as in the above example.
Dot plots are an excellent tool that can be utilized by Green Belts through Master Black Belts, and are especially useful in the Analyze and Improve phases of DMAIC, for generating clues and quickly comparing proposed process changes.
Dot Plot Example
The example above came from a team working on a downtime reduction project (we’ve also seen dot plots used for comparing production tooling, machines, operators, etc.). One of the team members suggested that the night shift crew generally experienced less downtime than the day shift crew, so the team collected downtime data and compared the two shifts.
One can see that the suspected downtime difference between the shifts was real. A hypothesis test could have been conducted to ensure that the observed difference was statically significant, but there was so such much separation in the data that a hypothesis test was not needed in this case:
Take another look at the above dot plot – do you see any clues that could ultimately lead to less downtime? Here are a few possibilities:
- Night shift does outperform day shift consistently, and finding out why and confirming the findings through an experiment could create a major breakthrough for the project team.
- While night shift has significantly less downtime most of the time, there are a few outliers in the data where night shift experienced unusually high downtime – it would be worth going back through the data to understand the downtime-causes on those nights.
- Day shift shows some outliers on the “good” side of the distribution (days with unusually low downtime). These are equally valuable in terms of understanding why and possibly applying some lessons-learned to standard work practices, etc.