The situation

The complaint says: “Bearings are noisy.” That is all. Noisy which ones? Since when? On which line? The team begins to look for causes anyway, because looking feels better than asking.

Three days later, they have three theories and no description of the problem. Is / Is Not analysis fixes the order: describe first, then search.

What it is for

Is / Is Not analysis defines the edges of a problem. For each question you write what the problem is, and what it could be but is not. The difference between the two columns points to the cause.

You ask in four areas:

  • What: which part, which defect.
  • Where: which place on the part, which machine, which station.
  • When: first seen, which shift, how often.
  • Extent: how many, how much, and the trend.

Terms used here:

  • Distinction: something that is different about the “is” compared to the “is not”.
  • Change: something that altered in, on, or around that distinction recently.
  • Deviation: the gap between what should happen and what does happen.

A short history

The method comes from Charles Kepner and Benjamin Tregoe. Their problem analysis approach was described in their 1965 book The Rational Manager. Sources also link their early work to research for the US Air Force. We could not verify the details of that early work, so we do not rely on them here.

Where it applies, and where it does not

Use it when:

  • a complaint is vague, such as “parts do not fit”;
  • the defect appears on some machines, shifts, or cavities but not others;
  • you are writing the problem description in D2 of an 8D;
  • you need the current condition of an A3.

Do not use it when:

  • everything is broken. If the whole site lost power, there is no unaffected case to compare;
  • the cause is visible, such as a forklift that snapped a pipe;
  • the topic is opinion or strategy, where there is nothing to measure.

How to do it

The example is fictional. Baltic Precision Bearings makes bearing BB-6204-2RS. Dariusz is the quality engineer. Lucja supervises the grinding line.

  1. State the deviation. Write the facts, with no cause in it. Chatter marks on the outer raceway, with high vibration. Reject rate is 6.8%.
  2. What. Is: BB-6204-2RS raceway chatter. Is not: the shielded variant BB-6204-ZZ. Distinction: the 2RS has an extra seal groove grinding pass.
  3. Where. Is: Grinder 2. Is not: Grinders 1 and 3. Distinction: Grinder 2 has a different spindle model.
  4. When. Is: from Tuesday 14:00, second shift. Is not: before that, and not on nights. Distinction: the spindle drive belt tension was readjusted at lunch on Tuesday.
  5. Extent. Is: 140 of 2,000 parts. Is not: the other 1,860. Distinction: the bad parts came in the first 45 minutes after tool dressing.
  6. Look at the distinctions and ask what changed. The belt tension was adjusted at lunch, on the one grinder with that spindle. After dressing, the grinder vibrates until it settles.
  7. Test each idea against both columns. A good cause explains why the problem is in the “is” and absent from the “is not”. Here the belt tension, set too tight, explains Grinder 2, the time, and the post-dressing pattern. Dariusz and Lucja then confirm it at the machine with a vibration reading and the maintenance record.

Note what happened: the team did not need to guess. The table asked the questions, and the answer fell out.

Three ways it goes wrong

1. “Is not” that is too wide. Writing “all other parts” gives you nothing. Pick the nearest relative. If cavity 3 flashes, the “is not” is cavities 1, 2, and 4 in the same mould. Wide comparisons show no distinctions.

2. A theory in the description. “Operator torqued the flange wrongly” is a theory, not a fact. Write what you can see: “flange bolt is loose by 2.5 Nm”. A theory in the description hides every other cause. “Operator error” is not a root cause, and it is not a description either.

3. Skipping the changes. The grid is neat, the distinctions are listed, and nobody asks what changed. You end up with a pretty record of the problem and no trigger. The question “what changed here recently?” is where the grid pays off.

Our recommendations

  • Fill it in at the gemba, or with data from it. An office table is a guess.
  • Choose a close “is not”. Same part family, same line, same week.
  • Mark each answer. Observed, or still assumed.
  • Do it before the fishbone. It narrows the search, so the fishbone has less to cover.
  • Keep the sentence neutral. No names, no blame, no theory.
  • Remember containment. While you analyse, protect the customer. Containment is not correction.

Try it

Our Is / Is Not tool is coming soon. It will run in your browser, with no account, and your data will stay on your device. Meanwhile, see the tool page: /tools/is-is-not.

From the book

This guide supports Chapter 10, the 8D chapter, of the field manual, where it is used to describe the problem. See /book for the book.