Problem solving Decide

Kepner-Tregoe

Kepner-Tregoe is a sequence for gathering information, comparing alternatives, and examining risk before a decision.

The questionWhat do we know, what is the cause or the best choice, and what could go wrong with it?

Also called KT, the KT method, or the Rational Process.

Process · Charles Kepner and Benjamin Tregoe, 1965

1
Situation appraisal What needs attention, and who owns it
2
Problem analysis Is and is-not facts, then test causes
3
Decision analysis Musts and weighted wants for each option
4
Potential problem analysis What could go wrong with the plan

Schematic of the four Kepner-Tregoe processes, in the order a team usually runs them.

When to use it

Something has gone wrong with no obvious cause, such as a defect, a yield drop, or an outage, and the obvious explanations have been tried.
A choice between alternatives is being argued from preference, and the group needs agreed criteria before it compares them.
There are many concerns at once and nobody has sorted which ones need which kind of thinking.
A decision has been made and nobody has tested the plan for what could go wrong.

When to avoid it

Cause and effect cannot be separated from outside the situation, as in the Complex domain of Cynefin. Analysis produces a better model of the past, and small probes work better.
The decision is small and reversible. The full process costs more than the decision, and a rough ranking will do.
Nothing has changed because the thing never worked, as with a new design. Problem analysis looks for a deviation from performance that used to be normal. First principles or design thinking fits better.
There is an emergency and time is the main constraint. Stabilize first, then analyze, or use a loop such as OODA.

Overview

Kepner-Tregoe is a set of four structured processes for thinking through a business situation. They sort out what needs attention, find the cause of a problem, choose among alternatives, and check the chosen course for what could go wrong. It supports decisions where an unexamined first answer is costly and a group would otherwise argue from preference.

The distinctive technique is the is and is-not specification in problem analysis. You describe what the problem is and what it could have been but is not, then look for what distinguishes the two. Candidate causes have to explain every fact, not most of them. The method fits deviations from performance that used to be normal. It is less use for new designs or questions without agreed facts.

The four processes

Situation appraisal

Answers what is going on. List the concerns, separate them, and rank each by seriousness, urgency, and trend. Decide which process each one needs and who owns it. It also shows which concerns are several problems and decisions bundled together.

Problem analysis

For a deviation from expected performance with an unknown cause. Describe what the problem is and what it is not, across what, where, when, and how much. Look for what distinguishes the two sets and what changed. Test each candidate cause against every fact.

Decision analysis

Set the objectives and split them into musts, which are limits an option has to meet, and wants, which are weighted. Generate alternatives and screen them by the musts. Score the rest on the wants, then look at the risks of the leaders.

Potential problem analysis

Run it before acting. List what could go wrong with the chosen plan, and rate the likelihood and seriousness of each. Write preventive actions to reduce the likelihood, and contingent actions with triggers for the ones that still might happen. Later materials add potential opportunities.

What it is not

A decision matrix. Weighted scoring is one step in decision analysis. A matrix alone skips the objectives, the musts, and the risk check.
A root-cause tool and nothing else. Problem analysis finds the cause, but it is one of four processes.
A template. The discipline is in writing the facts of the is and is-not specification and testing causes against them. A filled-in form with opinions in it is a table.

The steps

Appraise the situation

List every concern, sort them by seriousness, urgency, and trend, and assign each to a process and an owner. The output is a short ordered list, so people stop treating everything as the same kind of problem.
State the problem or the decision precisely

For a problem, name the object and the deviation, such as line 2 underfills 500 ml bottles. For a decision, state its purpose in a verb and an object. Quality is down is not a statement, because it has no edges.
Run the matching analysis

For a problem, write the is and is-not specification, list distinctions and changes, and test candidate causes against every fact. For a decision, set the musts and weighted wants, screen the alternatives, and score them. Keep the facts visible to the whole group.
Verify, then check the risks

Test the most likely cause with one more observation before you fix anything. Then run potential problem analysis on the fix or the chosen option, listing what could go wrong, with preventive and contingent actions, triggers, and owners.
Record and follow up

Write down the facts, the reasoning, and what was decided. Schedule a check that the fix held. The output is a record the next team can read.

In practice

A beverage bottler had underfilled bottles coming off one of its three lines. The quality manager wanted to start with candidate causes. The facilitator started with the specification instead. What: 500 ml bottles were underfilled, and 1-liter bottles on the same line were not. Where: line 2 only. When: first seen on the night shift of the 12th, on every shift since, and not before.

Three causes came up. A new batch of valve seals failed the test, because it did not explain why the 1-liter bottles were fine. Operator turnover on the night shift failed too, because the same problem appeared on days. A recipe file update on the 12th, which changed only the 500 ml setpoint and loaded on line 2 because it runs the newer controller software, explained every fact. The team checked the file before changing anything on the floor, found the setpoint error, and fixed it.

What you leave with

A ranked list of concerns, each with its process and an owner.
A written is and is-not specification, with the distinctions and changes found.
A cause that survived testing against every fact, or a list of what remains unexplained.
Decision objectives, with musts and weighted wants, agreed before the options were scored.
A risk list for the chosen action, with preventive and contingent actions and triggers.

Failure points

Starting with solutions

The group skips specification and goes straight to candidate causes. The first plausible explanation gets adopted, and the specification is filled in afterward to match it.

A vague problem statement

Quality is down names no object and no deviation. The is and is-not columns have no edges, so every explanation fits.

Weights set to reach a conclusion

The group already prefers one option, so wants are weighted until it wins. The scoring looks rigorous and records the preference.

Skipping potential problem analysis

The decision feels finished once an option is chosen. The risks then appear during implementation, when preventing them would have been cheaper.

Ceremony in place of evidence

The forms are completed, but the distinctions are opinions and nobody checked the facts. A completed form stands in for the analysis, and a training certificate stands in for skill.

The failures above are usually Framing: the wrong problem was chosen before this framework was applied.

Origin

According to the company's own account, Kepner and Tregoe were social scientists who studied breakdowns in decision making at the US Air Force Strategic Air Command in the late 1950s. They concluded that officers who decided well did not differ from the rest in rank or career path. They differed in how they gathered, organized, and analyzed information before acting. They founded Kepner-Tregoe in 1958 and published The Rational Manager in 1965, and a revised edition, The New Rational Manager, followed.

The method has four processes: situation appraisal, problem analysis, decision analysis, and potential problem analysis, which later materials extend to potential opportunities. Names and details differ between editions. Kepner-Tregoe is also a company that trains and certifies people in its method, and that training and its materials are proprietary. This note describes the ideas in general terms, and the book is the public source.