In my last post, I started to talk about the free Analysis Platform application used by Denso and others (especially in Japan) to tackle big data. In this post I will go into much more detail on what Denso calls the Digital Native Seven Quality Control Tools (DN7 or DN7QC7). These data analysis tools go way beyond the usual things you can do in Excel, and are designed to tackle big data analysis for industry 4.0. Beyond that, this application also includes more tools, which I will also show briefly in my next post.
As a reminder, the tool is available for free online on Github, but its support site is only in Japanese, with only a bit of English documentation. But let’s have a look at the seven tools:
Full Point Plots (FPP)
The full point plot (FPP) is basically a control chart. It shows the variables’ behavior over time (or generally as a series). While it does not really use statistical tools, it does plot all individual data. The advantage of the Analysis Platform application is that you can plot a lot of different variables easily, also include a histogram, and have an interactive chart where a mouse-over shows you more details of the data. Below is an excerpt of the plot with the sample data. The actual display of the app in the web browser would go across multiple pages. So far the graphs itself are nothing new, but a nice way to display the data. The blue bars at the top are called a “label plot,” showing you when changes (e.g., switching of magazines, batches, product type, etc.) happen.
If there is a LOT of data, such a plot would be overwhelming, but the application then automatically switches to a high-speed mode, omitting some data points through sampling.
Ridgeline Plot (RLP)
The ridgeline plot (RLP) (formerly known as a Joyplot) is based on a histogram, but you get a graph showing you the histogram for each time interval as a density curve in chronological order. It looks like a mountain ridge, hence the name. This way you can visualize changes in distribution over time that would not be visible in a simple graph. You can see the fluctuations in the process over time. While the full point plots above could be done in Excel, doing a ridgeline plot would be quite tough using standard spreadsheet software. The screenshot below is roughly half of the graphs in the sample data.
The dotted line to the side of the ridgeline plot shows increasing or decreasing trends in the distribution. Is the overall decreasing or increasing for this particular point in the plot below. A straight line of dits means there is not much change over time, but a moving line of dots suggest that something is going on, and you have a fluctuation.
Calendar Heat Map (CHM)
The calendar heat map (CHM) displays time-series data in the familiar format of a monthly or yearly calendar, using color intensity to represent the magnitude of a particular value for each day, week, or month. This is a great tool to analyze long-term data and to visualize cyclical patterns where certain days, weeks, or hours are different from others. For example, it is possible to visualize defect rates and production changes over a long period of time. You can see anomalies based on the time of the event, regardless if it is hourly, daily, weekly, monthly, or any other time range. A common result is that Monday mornings are bad for quality, as are shift changes. The Analysis Platform allows plots on a daily (0:00–24:00h) or weekly (Sunday–Saturday) basis (for now).
Multi-Scatter Plot (MSP)
A multi-scatter plot (MSP) is just what it says, a scatter plot showing data with three variables on two axes. In such a scatter plot, you can easily see the correlation between two variables, with multiple plots showing you all correlations between the selected variables. It helps you to understand correlations and clusters quickly, especially for a large number of variables and their correlation. Different from a normal scatter plot, all combinations of variables are shown. In this tool, up to seven variables can be compared with each other at the same time. If there is a lot of data, normal scatter plots will be overwhelmed, but in this case the Analysis Platform will automatically switch to a contour plot as shown below. Such contour plots with over one million data points have been used, visualizing outliers and odd chunks.
Parallel Coordinate Plot (PCP)
A parallel coordinate plot (PCP) connects multiple variables in parallel and allows you to see the correlation between these variables with respect to the desired outcome. It is easy to see what kind of value combinations are common and where the outliers are. It shows you the relationships between different values and a desired measurement(s). The color indicates the level of correlation, highlighting you where to look at in particular. This tool can show you which variables are likely to be relevant for the quality issue(s) you are looking at.
Sankey Diagram (SkD)
Everybody should know the Sankey diagram. This is a type of flow diagram used to visualize the flow of resources, such as energy, money, material, or people, through a system or process. In Japan, they also see this as a variant of the fish bone diagram. The width of the arms (links) is proportional to the quantity of the flow they represent, and thicker arms are usually more relevant. This tool helps you to narrow down the important factors from a huge number of possible options.
Co-Occurrence graph (COG)
The co-occurrence graph (COG) analyzes events that happen on the same time interval (e.g., how often an alarm happens, its relationship to events, and other things that happen at the same time). It focuses on the frequency with which one phenomenon and another appear at the same time. For a better understanding, this also includes a Pareto chart next to the co-occurrence graph.
…at least that is what it is supposed to do, but unfortunately, for me this module glitched. Sorry. According to the support website, this requires a particular data format, and a standard CSV file may need to be converted into a special CSV file.
In my next post I will look at some additional tools beyond the digital native seven quality control tools, which may also be useful. But now, go out, dig through your mountains of data to find the needle in the haystack that will make all the difference, and organize your industry!
PS: Many thanks to Yasuaki Matsunaga from Denso for showing these tools at the APMS Keynote 2025 in Kamakura, Japan!
Discover more from AllAboutLean.com
Subscribe to get the latest posts sent to your email.