Gains:
- Ability to transform an ambiguous customer question into a clear, solvable problem statement and establish a MECE hypothesis tree (issue tree) with artificial intelligence
- Ability to use artificial intelligence to generate hypotheses and detect blind spots, and decide with one's own judgment which hypothesis will be tested first.
- Ability to recognize and avoid common pitfalls in problem structuring (conflicting branches, assuming the solution in advance, wrong question).
The most expensive mistake of a consulting project is solving the wrong question. The customer asks "our sales are decreasing, what should we do?" comes: but the real question is "in which segment, why and how much of it is under our control?" it could be. An experienced consultant starts by defining the problem correctly, not by looking for answers. Artificial intelligence is the perfect thinking partner at this stage: it generates dozens of possible causes, sub-questions and hypotheses within minutes. But deciding which question is the right question takes judgment, and that judgment is yours.
In this unit, we will combine two tools that are the backbone of consultancy with artificial intelligence: the hypothesis tree (issue tree, a tree diagram that branches a problem into sub-questions) and the MECE principle (Mutually Exclusive, Collectively Exhaustive — the branches do not overlap each other and together cover the issue completely).
From vague question to solvable problem
A good problem statement includes four things: who (decision maker), what (decision), criteria (what success is measured by), boundary (scope and duration). “Sales are falling” is a complaint; "Which are the 3 most effective levers that will increase turnover in the B2B segment by 10% in the next 12 months, with the current budget?" is a problem statement. You can make this transformation happen to artificial intelligence, but first you must explain the raw situation clearly.
Your role: problem structuring coach for the management consultant. I'll give you the client's raw complaint. Your task:1) Transform this into a one-sentence, measurable problem statement (must include decision maker, decision, success criteria, scope, duration).2) Reveal the 3 implicit assumptions on which this statement is based.3) Suggest 2 alternative problem frameworks to counter the risk of "solving the wrong question." Raw complaint: "[in the client's words...]"
Tip: Read the problem statement back to the customer. “So success for you is a 10% increase in B2B turnover in 12 months, correct?” This single sentence resolves the direction of the project and the subsequent "we didn't want this" debate from the start.
Building the hypothesis tree with artificial intelligence
Once the problem is clear, you branch it out. The root of the tree is the main question; branches are possible response axes; leaves are testable hypotheses. For example, "Why isn't B2B turnover increasing?" The question can be divided into two main branches: (1) new customer acquisition is weak, (2) existing customer revenue is decreasing. Each branch is divided into its own sub-branches. A good tree is MECE.
Your role: MECE hypothesis tree expert.Main question: "[problem statement]"Task: Branch this question in no more than 2 levels.Rules:- First level branches must be MECE (no overlap, no gaps).- Each leaf must be a data-testable hypothesis sentence (of the form "... because ...").- Next to each hypothesis, write the single data source needed to test it.Format: indented bullet list.Probe: At risk of overlap/gap under the heading "MECE audit" mark the branches.
AI gives you a tree with 15 hypotheses in 3 seconds. But your job doesn't end there, it begins: you choose which branches are truly MECE, which hypothesis is the most likely and easiest to test.
Prioritization: which hypothesis to test first
There is no time to test all hypotheses. Use two axes: impact (how big it contributes to the solution if it's true) and ease (how fast/cheap it is to test). Those with high impact + high convenience come first. This is called a "quick win sweep."
hypothesis
Effect (1-5)
Ease of testing (1-5)
priority
Pricing is behind competitors
5
4
Before
Sales team is inadequate in B2B
4
3
then
There is a product feature gap
4
2
research
Brand awareness is low
3
2
hold
You can have artificial intelligence fill this matrix; but review the impact and convenience scores with your own industry knowledge. The model often overestimates ease of testing.
Your role: prioritization analyst. Place the following list of hypotheses into an impact-ease matrix. For each hypothesis: impact (1-5), ease of testing (1-5), single data source needed, recommendation (before/after/investigate/hold). Rules: give ease score optimistically, write rationale. Highlight the 2 hypotheses with the highest “impact × ease” multiplier. Hypotheses: [list]
three mini cases
Case 1 — Wrong question corrected. A retail chain asked, "We have a high employee turnover rate, should we increase salaries?" he asks. The consultant opens the problem frame with artificial intelligence; The model illustrates three implicit assumptions: that pay is the problem, turnover is the same across all stores, and pay is the solution. The data is drawn: 70% of the turnover rate is in 3 stores and always in the same two managers' team. The real question is manager behavior. A salary increase of 14 million TL would be the wrong solution.
Case 2 — Non-MECE tree. A consultant asks the model "why are profits falling?" He wants the tree. The model gives branches "income is low", "cost is high", "competition has increased". The first two branches are MECE (profit = revenue − cost), but “competition increased” is a cause and revenue/cost is an effect; Branch levels are mixed. The consultant fixes the tree: level one is revenue and cost only; Competition causes a sub-branch of income.
Case 3 — Prioritization paid off. In a B2B software company, 12 hypotheses are generated. The impact-convenience matrix puts the "onboarding process is long" hypothesis in the high impact-high convenience box. Confirmed by a week's data analysis: The cancellation rate of customers who are not activated in 30 days is 3 times higher. A little intervention (guided installation) reduces cancellation by 22%. Testing the correct hypothesis first shortens the project.
Weak prompt / Strong prompt
Weak prompt:
Make me a list of why this company's profits are falling.
The model enumerates random, overlapping, and unsubstantiated causes; It does not establish a thinking framework.
Powerful prompt:
Your role: MECE problem structuring specialist. Main question: "Company just build.
Check questions of a good problem structure
When you finish your tree, ask these questions: You can also ask artificial intelligence, but you make the final decision:
- Is the root question really the question the customer wants to decide? - Do the first-level branches overlap? (Mutually exclusive?)- Is anything left out? (Is it complete together?) - Can each leaf be tested with data, or is it opinion? - Have I assumed the solution in advance? (For example, "salary is the solution")
Common mistakes
- Embed the solution in the question. "Should we lower the price?" The question presupposes the solution and makes the other levers invisible. Ask "why" first.
- Overlapping branches. If "corporate customers" and "large customers" are separate branches, the same customer will be counted in two places; MECE is broken.
- Mixing level. Putting the cause (competition) on the same level as the effect (revenue decline) makes the tree irrational.
- Untestable leaves. "Poor management" is a judgment, not a hypothesis. "Decision time is 40 days on average, 12 days in the industry" is the testable hypothesis.
- Accepting the tree of the model as it is. AI gives a quick outline; MECE auditing and prioritization is your job.
Caution: AI presents a hypothesis tree in a very convincing way. Persuasion is not accuracy. Always manually inspect the tree for MECE; Drawing branches with a pen and paper often shows overlap that the model missed.
In summary
In consulting, the value is in the right question before the right answer. AI is powerful at turning a vague complaint into a measurable problem statement, uncovering implicit assumptions, and quickly sketching out a hypothesis tree. But it is up to you to decide whether the tree is MECE, the testability of the hypotheses, and which one to test first. The model produces a draft; It is the consultant's job to own, control and prioritize the problem.
Application task
Choose a real or imagined customer complaint (one sentence). First, transform this into a measurable problem statement and three implicit assumptions with a strong problem-framing prompt. Then build a two-level tree with the MECE hypothesis tree prompt. Manually inspect the tree to find and fix at least one conflict or gap. Finally, fill in the effect-ease matrix and select two hypotheses to test first.
checklist
- [ ] I transformed the raw complaint into a problem statement with who/what/criteria/scope/duration.
- [ ] I uncovered implicit assumptions.
- [ ] I inspected the first level of my hypothesis tree as MECE.
- [ ] I verified that each leaf is a testable hypothesis.
- [ ] I made sure I didn't bury the solution in the question.
- [ ] I prioritized with the impact-ease matrix.
- [ ] I did not blindly accept the model's tree and corrected it manually.