From problem statement to issue tree: a consulting project toolkit

The owner of a 30-person industrial catering firm (invented) says: "Revenue keeps rising, but I keep less and less of it." The new consultant has six weeks, ten Excel tabs and no question yet. This post chains four tools from the consulting-method literature into one working order and runs them on that invented example. There is no real company here, and every number is built so that the arithmetic can be checked.

How do the four tools fit together?

Each tool answers one question, and the output of one feeds the next. The scoping sheet fixes which question is being solved. The issue tree lists the branches that could explain it and which to test first. The Canvas shows which block of the business model carries the gap. The Gantt chart lays out who does what, and when. The order is my arrangement. The sheet, the tree and the work plan sit inside the seven-step process described by (Umbrex, n.d.) and (Conn and McLean, 2019, p. 3); the Canvas comes from a different tradition (Osterwalder and Pigneur, 2010), and I have added it as a check.

The scoping sheet: what goes in when the client only says "margin is falling"?

Turn the symptom into a question whose answer is the solution, then fill in the other five cells with the client. Umbrex asks for a question that is specific, measurable, action-oriented, relevant and time-bound, and warns that the way the problem is stated limits the solutions that can follow (Umbrex, n.d.).

The invented firm has 30 staff and cooks for several factories. Over the last 12 months revenue rose about 4 per cent to 25.0 billion dong, while gross margin fell from 17 per cent to 11 per cent.

The "no price rise" constraint, like the 15 per cent target, is one I assign to the invented owner. It deserves attention because it removes a whole branch of solutions before any analysis starts. Conn and McLean remark that clear boundaries on scope make problem solving faster and more accurate (Conn and McLean, 2019, p. 13). The price: if the owner sets a constraint out of habit, the consultant must record it as a constraint, and must not let it quietly turn into a "finding" that price is not the problem.

The issue tree: where does MECE break, and how do you repair it?

MECE breaks when branches overlap or leave a gap, and a tree built from a calculation repairs both. Umbrex describes two variants: a hypothesis tree when the team knows enough, and an issue tree, written as open questions, when it does not. The aim is pieces that do not overlap and that together cover the whole, which is MECE (Umbrex, n.d.).

The draft I built for the example is the kind of tree that comes out when you list causes off the top of your head, and it fails at the first level:

The repair comes from Conn and McLean: a profit tree, which they call mathematically complete and therefore usable for modelling different assumptions (Conn and McLean, 2019, p. 23). Here, gross profit is the number of meals times (selling price minus cost per meal). Every level is a calculation, so in my reasoning nothing overlaps and nothing is missing.

Check it: 833,000 × 3,300 dong = 2.75 billion; 800,000 × 5,100 dong = 4.08 billion; the difference of −1.33 billion is +0.17 billion from extra meals and −1.50 billion from cost per meal. This split is a profit bridge. It also shows that the "customers squeeze price" hypothesis has no room at the level of the average, leaving one open question: does the 30,000 dong average hide a change in contract mix?

Two limits. This is MECE for the arithmetic, not for the causes: "price rose" and "quantity rose" at the leaves are still hypotheses that need data. And I have read nothing showing that an arithmetic tree beats a thematic one; it only fits when there is a total to divide. To choose which branch to test first, Umbrex says teams commonly use a two-by-two of impact and ease (Umbrex, n.d.). Ingredients carry 72 per cent of the cost increase and the invoices already exist, so they go first.

What does the Canvas check that the tree has not seen?

It shows which block carries the gap and which two blocks are out of step. Osterwalder and Pigneur describe a business model through nine blocks in four areas: customers, offer, infrastructure and financial viability. They also write that costs can be calculated fairly easily once key resources, key activities and key partnerships are defined (Osterwalder and Pigneur, 2010).

Read across the last two rows and a hypothesis appears: a selling price fixed for the year, ingredient purchases floating with the market. The tree's leaf only asks "did purchase prices rise?"; the Canvas forces the second question, "does the selling price move at the same pace?" This is my inference from invented numbers, and weeks 2 to 4 would test it against invoices and contracts.

To be plain, the Canvas here finds nothing that a tree question, "are purchases fixed or floating?", would not find. It describes and does not calculate, and has no competitors or timeline. Its value is that it makes you look sideways across blocks.

What goes into a six-week Gantt plan?

Tasks, weeks, milestones and waiting dependencies, arranged so that nobody is overloaded. The table is a six-week plan for the invented example, read row by row.

Umbrex says a work plan should record, for each hypothesis, the analyses, end products, sources, timing and owner, with more detail for the near weeks than the far ones (Umbrex, n.d.). Conn and McLean use the idea of a one-day answer: the team always holds its best current summary of the problem and the way to solve it (Conn and McLean, 2019, p. 4). So the plan has a milestone at the end of week 1: a draft answer, even if wrong. The "Waits for" column is common-sense reasoning: interviews come after the numbers, so you ask in the right places.

Where do the tools overlap, and when should you drop one?

They overlap in three places. The sheet's context and the Canvas both describe the firm. The sheet's scope decides which branches of the tree exist. The tree's priority order is almost the list of rows in the Gantt chart.

One general limit: I have read no study showing that the four-tool chain beats each tool used alone. The chain is my arrangement.

Going deeper: what did I read, and what did I not?

Umbrex: I read the whole guide. It says it is largely informed by McKinsey Staff Paper 66, which I have not read. The page has no page numbers, so I paraphrase.

Conn and McLean (2019): only chapter 1 and the contents list from the publisher's free Wiley excerpt, pages 1 to 29. I have not read later chapters.

Osterwalder and Pigneur: the free preview of the self-published 2009 edition (the nine blocks and the Canvas), not the 2010 Wiley edition the library lists. I assume the text is the same but did not compare, so I give no page numbers.

The catering firm and every figure about it are my invention; they illustrate use and do not show that the tools work. See also Porter for small firms and feasibility checks before a plan.