| Info |
|---|
Labs Feature (Early Access) |
Process behavior charts are for characterizing a process as predictable or unpredictable. It does this by identifying:
points that represent exceptional variation (aka signals),
the amount of routine variation (aka noise) to expect from a predictable process in the future.
There are 2 sub-charts within this Process Behavior Chart, both reflecting information about your Cycle Time data:
Individuals Chart - a running record of the Cycle Times of individual work items
Moving Range Chart - a running record of the variation in Cycle Times
Here’s what we cover in this document
| Table of Contents | ||||||
|---|---|---|---|---|---|---|
|
The Individuals Chart (Finding Process Signals)
Here you can see the Cycle Times of individual items plotted over time
The horizontal axis represents a cumulative number of items ordered by the date they are finished. Items finished on the same day listed consecutively, ordered by item ID, alphanumerically. The vertical axis represents the Cycle Time of the items.
Process Signals Chart Control
The Signal Detectors controls allow you to separate signal from noise and find patterns within your noise to get early indicators that process changes are afoot (for better or worse). These detectors are listed in order of signal strength and should be used in that order.
Process Signal | What it shows you | How it is defined |
|---|---|---|
Large change | Highlights individual items that are out-of-control. These out-of-control items are the only true signals in your data. The rest is noise. | Any single point outside of the natural process limits. |
Moderate change | Highlights patterns inside the noise of your data that indicate moderate process changes. | A run of 3 items where 2 out of any 3 consecutive points within the process limits are above the 2-sigma line. |
Moderate, sustained shift | Highlights patterns inside the noise of your data that indicate sustained, moderate process changes. | A run of 5 items where 4 out of any 5 consecutive points within the process limits are above the 1-sigma line. |
Small, sustained shifts | Highlights patterns inside the noise of your data that indicate small, sustained shifts in your process. | A run of at least 8 successive values within the process limits on the same side of the average line (above or below) |
In practice, most people who use process behavior charts effectively find that they have plenty of signals to tackle. However, if greater sensitivity is required, you can use the other signal detectors.
The Moving Range Chart (Validating your data)
| Info |
|---|
The moving range chart is also sometimes called the mR chart or the XmR chart. |
While the Cycle Time Run chart shows the actual cycle time values for the work items, the moving range chart shows the differences between successive Cycle Time values. In other words, it is a running record of the Cycle Time variation generated from your process.
The XmR chart is your first, best, and probably last chart for characterizing the variation in your system. Nowhere else are you going to find such a clear, concise visualization of the probable noise and possible signal contained within your data. – Dan Vacanti
The vertical axis represents the variation between the two Cycle Times being compared.
The horizontal axis represents a cumulative number of comparisons made. There will always be one less point on the Moving Range chart than there is on the Cycle Time Run Chart. That’s because it takes two Cycle Time Run Chart points to plot one Moving Range chart point.
Chart Controls
Summary Statistics
These controls allow you to toggle a statistics bar on and off for each chart. This chart makes it easier to see the values for the basis lines used for each chart
Process Signals
These signal detectors are explained in detail in the Individuals chart documentation above.
These signals are listed in order of strongest to weakest. Most will find plenty to do by focusing on the first signal before moving on to weaker signals.
Basis Lines
These are lines that are used to analyze your data in the charts. The first 4 are in the Individuals Chart (top) and the last 2 are in the Moving Range chart (bottom).
The Process Signal detectors in the Process Signal chart control use these lines to find signals. When you read the description of each signal you can see that the calculations correlate to a specific basis line.
If you are curious about how these are calculated, you can find more information here.
| Panel | ||||||||
|---|---|---|---|---|---|---|---|---|
| ||||||||
Note: the sigma lines have nothing to do with standard deviation. They just unfortunately use the same nomenclature |
Layout
The layout determines which large sections of the chart you want to see. Use the checkboxes to toggle them on and off.
The Date control allows you to select a subset of the data to zoom into in the Cycle Time and Moving range chart. You do this by clicking and dragging your mouse.
Item Filter
You can filter down the dots shown on this chart by choosing one or more available filters.
If you want to clear your filters so that all dots show up again, you click the Reset button.
Workflow Stages
As this chart deals with Cycle Times, you need at least 2 workflow stages selected.
The top checked stage will signal to start the Cycle Time clock for a work item when it enters that stage. The last checked stage will signal to stop the Cycle Time clock when the item enters.
This allows you to look at the behavior of your entire process or just a specific portion of it.
FAQ
| Expand | ||
|---|---|---|
| ||
A process with data that falls completely within the Natural Process Limits is stable and predictable. |
| Expand | ||
|---|---|---|
| ||
If your process is largely stable then your data should be reasonably reliable for forecasting. You can tell if your system is stable by seeing how often items fall outside of the Natural Process Limits. You can choose the Large Change option from the Process Signals chart control to identify items that are “out of control”. |
| Expand | ||
|---|---|---|
| ||
Items that are above the Natural Process Limits are signals. The rest of the data points are noise. To identify your signals, use the Process Signals Chart control and select the first option. This will highlight every item outside of your natural process limits. In order to use this chart effectively, investigate each out-of-limits point in search of a cause. This will allow you to address them and bring your system back into control. |
| Expand | ||
|---|---|---|
| ||
When we don’t have a clear differentiation between signal and noise we can overreact to data points and cause more problems than we fix. When we can correctly identify true signals, we can take timely and appropriate actions. In cases where the data signals improvement, you can endeavor to sustain it. In cases where the data signals a turn for the worse, you know to look for causes so you can work to improve the situation. |
| Expand | ||
|---|---|---|
| ||
Please note that, in practice, most users of this chart will find plenty of signals using just the Large Change Process Signal detector However, you can use the more sensitive signal detection rules to find early indicators within your data’s noise. These rules should be investigated in order as you investigate smaller and smaller shifts in your process.
|
| Expand | ||
|---|---|---|
| ||
First, congratulations, you have a predictable system! Now, you might have a system that is technically predictable but with too wide of a range of Cycle Time data. In other words, you could be predictably unpredictable. How can you tell if this is the case? Look at the distance between the Natural Process Limits lines on the chart. (If no lower line is shown then use 0 as the lower limit.) If this is a problem, work on reducing work item age and see if signals start appearing. If they do, work to mitigate the causes as you continue to reduce your Cycle Time range through Work Item Age management |
| Expand | ||
|---|---|---|
| ||
The goal of these detection rules is to eliminate over 99% of noise. According to Dr. Wheeler, the use of three-sigma limits strikes a balance between the economic consequences of the dual mistakes of missing signals and getting false alarms. Using two-sigma limits on its own has a false alarm risk more than 4 times larger than with our signal detectors. So, if you’re seeing a signal for any of these rules then the cause is not likely due to chance. The cause of these signals can reasonably be interpreted as an assignable cause. |
Where to learn more
Donald J Wheeler - Contra Two Sigma: The consequences of using the wrong limits, Quality Digest Daily, May 1, 2013 http://www.spcpress.com/pdf/DJW255.pdf