by Commoncog

The Xmrit User Manual

Table of Contents

Getting Started with Xmrit’s Charting Tool

What does the Xmrit tool do?

Xmrit is the simplest way to create beautiful XmR charts directly in your web browser. Xmrit generates two charts from your data: an X-Chart (so-named after the metric you’re interested in, aka ‘the X variable’) and an MR-Chart (which stands for ‘Moving Range’).

You may read a more gentle introduction to XmR charts here.

All the data you input into Xmrit is 100% confidential, no data is ever sent to our servers. If you decide to share your XmR Charts with others the data is stored within the link, not at Xmrit.

What are the coloured dots?

XmR charts help you separate signal from noise in your data. It does this via a handful of detection rules. Historically, there have been many detection rules. Xmrit uses a simplified set of three rules, as introduced by Donald Wheeler in his books Understanding Variation and Making Sense of Data.

Rule 1: Process Limit Rule

When a point is outside the limit lines on either the X chart or the MR chart, this is a strong source of exceptional variation and you should investigate.

An X chart displaying one point out of the limits and coloured red

Rule 2: Quartile Limit Rule

When three out of four consecutive points are nearer to a limit than to the centre line, this is a moderate source of special variation and you should investigate. In the chart below, the dotted line is the mid-point between the centre line (in red) and the limit lines (in blue). Three out of four of the yellow points are below that dotted line, and are therefore nearer to the lower limit line than the centre line.

Rule 3: Runs of Eight

When eight points lie on one side of the central line (in red), this is a weak source of special variation and you should investigate

How do I create my first graph?

To create your first graph you need to enter your own data into the Xmrit spreadsheet, found on the left hand side of the tool. The spreadsheet works like Excel and Google Sheets, and you can directly copy data from those programs into Xmrit, or directly edit the spreadsheet in Xmrit.

It is usually easier to paste in your data from a spreadsheet program, and make small edits in the Xmrit spreadsheet.

Below is a GIF of how to copy and paste data directly from Google Sheets into Xmrit, and the XmR Charts being rendered automatically.

Note that dates should be in the format YYYY-MM-DD, e.g. 1991-01-25.

What data can I put into Xmrit’s spreadsheet?

In Xmrit’s date column the dates should be entered in the format YYYY-MM-DD e.g. 1991-01-25, however it will attempt to interpret non correctly formatted dates.

The value column will accept integer and decimal values only.

How can I label my charts?

You can label your graphs by either clicking on the X-Chart axes in your XmR Charts or on the table headers in Xmrit. Below is a GIF of those two methods.

Advanced Functions of Xmrit

What are dividers and how do I use them on my data?

Dividers allow you to segment your XmR charts, creating multiple analyses within a single XmR chart. This is useful when you are trying to see if a change you made actually made a difference to your process. For example if you are analysing the number of calls a week your sales team is making you may want to divide your data on the date you introduced a new training plan.

Dividers are added and moved around on the X-Chart. When a divider is added to the X-Chart it creates a new segment, and the process limit lines are recalculated using only the points present within that segment. You can add a maximum of 3 dividers onto one chart, creating 4 segments.

This kind of segmentation is almost impossible to do within Excel and Google sheets, and is cumbersome to do even within expensive statistical programs such as Minitab. Xmrit makes it as easy as clicking on a single button and dragging it to the position you wish to analyse separately.

Below is a GIF of adding, moving, and removing dividers on an Xmrit dataset.

What is the “locking limits” button for and how do I use it?

The control limits on a standard Xmrit XMR chart have their limit lines calculated from the data on the chart. If the chart has no dividers then the limit lines are calculated from all the data on the charts, if there are dividers then the limit lines for each segment are calculated by the data within each segment. Limits calculated in the standard way are indicated by dashed limit lines.

Sometimes you may wish to calculate or set limits that don’t directly correspond to the data currently on the chart. This is known as “locking limits”. Xmrit charts with locked limits will be visually signalled with solid limit lines, rather than the normal dashed limit lines.

Common reasons to lock limits include:

  1. Preventing Continual Updates of Limits: You might want to “lock” the limits after a certain number of points, to prevent continual updating of the dataset. Standard practice is to take 15-20 points of a stable process to calculate locked limits.
  2. Outliers: Removing outlier data points on the charts from the calculation of limits. Say for instance you want to remove one or two data points from consideration when calculating limits — which is common when you have an outlier event that you know won’t happen again.
  3. Limits That Exceed Process Behaviour: To prevent the limits going beyond a natural limit of the process e.g. The automatically calculated limit gives a value of -10, but the process can never go below 0.  

To lock your limits, click on the Lock Limits button. You will then be presented with a popup that gives you two ways to lock your limits:

  1. [Recommended] Locking limits from a dataset using the popup spreadsheet.

After clicking the Lock Limits button you will be presented with a popup spreadsheet. This will be prefilled with the data in the main Xmrit Spreadsheet. You can paste in new data in the popup spreadsheet or remove specific data points already present.

Once you have finished altering the popup spreadsheet dataset click the lock limit button to save the locked limits. Below is a GIF of the limits being updated using the popup spreadsheet.

  1. [Advanced] Manually entering data to lock limits to specific values

It is also possible to manually alter the limits by directly entering values into the cells at the top of the popup spreadsheet. This is not the recommended approach for most situations as it is possible to set incorrect limits that are too sensitive, or not sensitive enough.

WARNING: If you set non symmetrical process limits Xmrit will turn off the quartile rule for the process limit that you have changed, e.g. if you set non symmetrical process limits by changing the upper limit, then the upper quartile limit will be turned off. If you set non symmetrical process limits by changing the process average value then the quartile rule will be turned off for both the upper and lower quartile limit.

Below is a GIF limits being updated by manually inputting values.

When should I lock my limit lines?

The most common reason to lock your limits is when you are continuously analysing a process, and don’t want every additional point you add to change your process limits. Having your process limits continuously change on you is bad — it’s akin to constantly shifting your goalposts!

The rule of thumb is once you have 15-20 data points from a predictable process you can feel comfortable locking your process limits.

There are two other more advanced reasons you may wish to lock your limits:

  1. Removing Outliers: Let’s say that you have one or two points of exceptional variation in your dataset, causing your limits to be wider than they should be. Let’s also say that you’ve already investigated these points and discovered the root causes, and now want to estimate process limits without these out-of-limit points. This would be a good situation to exclude the points from your limit calculation.

  2. Limits That Exceed Process Behaviour: You may have a process that logically cannot go above or below a certain value. For instance, your metric is ‘On-Time Percentage’, where the maximum logical limit is 100%. Or ‘Number of Service Outages’, where the logical limit can never be less than 0. If you find the upper or lower process limit to be above 100 or below 0 (which happens!), you may manually shift the process limit to that logical limit, so you don’t confuse your audience. In Xmrit, you may fix limits using the manual entry feature.

When should I recalculate my locked limit lines?

The simplest answer is: when you see eight points in a row on one side of the average line. A run of eight is strong evidence that your process behaviour has changed and that you should recalculate your limit lines. You may place a divider right before the eight points, just to see what the new limits would look like.

Here is an example. Before placing a divider, we see multiple XmR chart rules being triggered:

Unpredictable process

It is very clear that there is a run of eight points above the centre line (the blue points are overridden with orange and red because Xmrit emphasises those colours when multiple rules are triggered at the same time).

Inserting a divider makes the process shift clearer (and reveals a sequence of exceptional variation that we should investigate):

Example of a permanent process shift in an XmR chart

However, there is room for qualitative judgement here. Sometimes you will observe (or initiate!) a process change that results in differences so large that you don’t have to wait for eight points to know that process behaviour has permanently changed. With Xmrit, you may quickly insert a divider at the point of your change, just to get a feel for the new limit lines.

With experience, you will find that your use of dividers and your decision to recalculate limits will become more confident.

How do I download images of my XmR charts?

You can download the images of your XmR charts by clicking the download image button at the bottom of the X-Chart and MR-Chart.

To share a link to your XmR Charts click on the share link at the top right hand side of Xmrit. The link will be copied to your clipboard, and can be shared with anyone you wish. All dividers and locked limits will be copied across in the link.


How many new data points to trust in the limit lines?

Let’s say that you’ve made a change. How many new data points must you wait before you can trust the new limits?

  1. 6 data points: this is the absolute minimum.
  2. Between 6 to 12 data points: the limit lines begin to gel.
  3. Between 12 to 20 data points: the limit lines begin to harden.
  4. More than 20 data points: there is typically marginal benefit to waiting for more data.

How many data points should you use when calculating your limit lines? Well, this depends on your use case:

  • If you are estimating limit lines for a stable process — use 15 to 20 data points. (Read the documentation for locking limit lines in Xmrit here). This is a rule of thumb. Many textbooks recommend between 25 to 30 data points for the best possible limits. However, in business, you often do not want to look at data that is too old. 15 to 20 is a good rule of thumb; use your judgment.
  • If you are introducing a process change and want to see if it has taken, 6 to 8 new data points may be enough! In Xmrit, place a divider to see if the process has changed. Use your judgment: do you think the change you have made in your business justifies waiting for more data? Or less?
  • Are you looking at daily data, weekly data, or monthly data? Bear in mind that anything more than 18 months ago may be irrelevant for your current use case. Business environments and businesses themselves can change over the course of 12-18 months. As always, you are an expert in your business, so use your judgment.

What is a predictable (or unpredictable) process?

A predictable process is a process that does not show any of the three rules (explained at the top of this User Manual). That is, it only shows routine variation.

An unpredictable process shows one or more of the three rules. This means that it displays routine and exceptional variation.

How should I improve an unpredictable process?

In classical Continuous Improvement, we are told that when we are presented with an unpredictable process, we should investigate the source of exceptional variation … in order to remove it!

This makes sense in most manufacturing contexts: exceptional variation is bad, and you want to get rid of it. In general, we can only improve a process if it is predictable — which means removing sources of exceptional variation.

In practice — and especially outside of manufacturing — some sources of exceptional variation may be positive. In such instances, your job is to run experiments to see if you can reliably reproduce such effects.

How should I improve a predictable process?

In classical Continuous Improvement we are told that when we are presented with a predictable process, we need to completely rethink the process in order to improve it.

A predictable process shows only routine variation, which means it’s performing as well as it possibly can. Your job is to fundamentally rethink how the process is done: perhaps you should introduce new training, or enact some new policies, or buy some new technoology?

When doing this, you have two goals:

  1. Either shrink the limit lines (reduce the variation)
  2. Or shift the entire range of variation upwards (or downwards, depending on which is better).

Classical continuous improvement literature will tell you to do (1) before you do (2), because a well behaved process will be easier to improve. More specifically, the classical literature will tell you that there are two benefits: first, a process with smaller variation is ‘cheaper’ to run (less variance means it is easier to plan around); second, with tight variation, it is easier to see even small shifts in performance.

That said, you should use your judgment for your specific business scenario — it may not be possible to tighten the variation for certain metrics. So it may make sense for you to go after (1) or (2) directly.

Something unexpected has happened with my chart, how can I reset?

Like in Excel or Google Sheets you can undo an action (Mac: Command + Z, Windows: Control + Z).

You can also click the Refresh Charts button to have Xmrit recalculate the charts.


How do I download the data I have inputted into Xmrit?

To download a CSV files of the data you have inputted into the Xmrit spreadsheet click the download data button at the top of Xmrit.

Which browser should I use?

Xmrit is designed to be used on all modern browsers, but has been tested primarily on Chromium browsers e.g. Google Chrome and Microsoft Edge. It may suffer performance issues on other browsers.

Xmrit doesn’t function well on mobile browsers, unfortunately.

Last Updated: 8 Mar 2024