Skip to content
LeanSuite - Lean Manufacturing Software

Book summary · Quality and Six Sigma · Intermediate

Understanding Variation: summary and review

The Key to Managing Chaos

By Donald J. Wheeler. First published 1993.

The short answer

Understanding Variation teaches managers to read their routine numbers as a sequence over time, not as this month against last month. Wheeler's central claim is that every process produces ordinary noise, and a simple chart with limits computed from successive differences shows when a value is a real signal. The book is written for managers and is opinionated. Read it if you review daily or weekly metrics. Skip it if you need the full SPC toolkit for capability studies and sampling plans.

Read it if

  • You run a daily or weekly meeting around KPIs and spend the time explaining why a number moved.
  • You manage supervisors who are asked to explain every miss.
  • You are an engineer or quality lead who wants the plain-language case for individual-values charts.

Skip it if

  • You need construction details for subgroup charts, sampling plans or capability indices. Use a statistics text or SPC manual.
  • Your quality group has standardized on p and u charts for count data. Wheeler argues against them, so know that debate first.

When it pays off: When your daily meeting reacts to every red number and before you add more KPIs. Read it before you build boards, not after.

The core ideas, in plain language

  1. 1

    Two kinds of variation

    Routine variation is present when nothing special happens. Exceptional variation signals that something changed. Mark Graban's takeaway: do not waste time chasing noise; improve the system when variation is routine, investigate when it is exceptional.

  2. 2

    The process behavior chart

    Plot values in time order, draw the average, and set limits from the average moving range, the gap between consecutive values. For individual values the limits sit at 2.66 average moving ranges either side of the average, per Baudin. Wheeler's writing warns against overall standard deviation, which assumes the data are homogeneous, the thing the chart tests.

  3. 3

    Why three sigma

    In his Quality Digest columns, Wheeler argues that Shewhart chose three-sigma limits as economic action limits, not probability limits, and that they work without normally distributed data. That claim is disputed (see below).

  4. 4

    Context and time order

    Two-point comparisons, dense tables, bar charts and trend lines mislead. Graban's lessons from the book include using line charts for time series and refusing conclusions from two points.

  5. 5

    Three responses to pressure

    A reviewer summarizes Wheeler's point that pushed to hit a number, an organization can improve the system, distort the system or distort the data. Many default to the last two.

  6. 6

    Voice of the customer versus voice of the process

    Specification limits say what the customer wants. Natural process limits say what the process delivers. Mixing them produces impossible tolerances or false comfort.

How to use it on Monday morning

  1. 1

    Pull 20 to 30 values of one metric

    Choose a number discussed daily, such as scrap rate per shift or downtime minutes per day. Pull the last 20 to 30 values in time order from the board or system, one per period, unaveraged.

  2. 2

    Compute limits by hand or in a spreadsheet

    Average the values. Take the absolute difference between each consecutive pair and average those. Add and subtract 2.66 times that second average from the first. A spreadsheet does it in minutes.

  3. 3

    Circle only the points outside the limits

    Ask for causes only on circled points. Tell supervisors that points inside the limits will not be explained one by one, and replace the bar chart on the board with this line chart.

  4. 4

    Compare limits with the customer limit

    If the process's natural limits reach past the specification, daily reaction cannot fix it and the system has to change. Open a problem-solving item at that level, not a daily explanation.

What to be careful about

  • The statistics are contested

    Michel Baudin derives the 2.66 and 3.27 coefficients from Gaussian assumptions and shows the moving range chart signals at 1 percent versus 0.3 percent for the individuals chart in stable normal data, roughly three times the false alarms. Wheeler's position is that normality is not required.

  • The p and u chart dispute

    Wheeler promotes XmR charts as a general tool. Lloyd Provost and the Healthcare Data Guide favor p and u charts for count data, per Jay Arthur. Arthur's advice: pick one method, apply it consistently.

  • A signal says look, not why

    A chart point outside the limits shows that a cause exists. It does not name the cause. You still need root cause work.

  • Data collection has moved on

    Baudin notes that one of Wheeler's reasons for plotting the moving range chart is spotting chunky, coarsely rounded data. He argues automatic capture from sensors has nearly removed that problem.

Understanding Variation compared with Out of the Crisis

Deming gives the management argument for why variation matters, in a long and loosely organized book. Wheeler gives the working method. Pick Wheeler if you own a metrics meeting and want something to do on Monday. Pick Deming if you first need the case against targets and ranking.

Read the summary of Out of the Crisis.

Book details

Bibliographic details of Understanding Variation
TitleUnderstanding Variation: The Key to Managing Chaos
AuthorDonald J. Wheeler
First published1993. First edition 1993, SPC Press (Knoxville, Tennessee); second edition 2000 (Open Library dates its record November 1999).
Edition shownSecond edition, hardcover, SPC Press, 2000
ISBN-13978-0945320531

Details checked against public records on 7 October 2026. Other editions have other ISBNs. We carry no affiliate links.

FAQ

Questions about Understanding Variation

Rather see it on your own problem?

Bring the problem you are stuck on, and we will show you on a call how LeanSuite handles it.

Pass it on

We want LeanSuite to be the best place on the internet for lean help. If this was useful, send it to someone on your team or in your network who needs it.