Issue Trees 2026: Building One From the Client Decision

What the top of an issue tree has to be

The top node is the decision in front of the client, written as a question with a yes or a no in it. "Losses" is a topic. "Which losses should we spend next year's capital budget on" is a decision. The second one generates branches that compete with each other for money, and competition is what makes a tree useful.

Four tests, and a top node that fails any one of them will produce branches nobody can act on.

1. Is there a decision in it? Somebody has to be able to answer yes, no, or this one rather than that one.
2. Is the scope bounded? One asset, one region, one budget cycle, one product line.
3. Is there a metric attached? Megaliters, dollars, points of margin, accounts.
4. Is there a constraint? A budget, a deadline, a regulatory limit. Without one every branch looks affordable.

Oliver Wyman puts the same move at the front of its published method. Its six steps begin with pinpointing the issue and breaking it down, in that order, and the firm describes the exercise as "an interactive exercise in analytical thinking that allows us to assess how you approach and evaluate a problem", checked 30 September 2026 at oliverwyman.com. On its Dairy Farm case the criterion printed under the opening question is the ability to identify the key question and the pertinent information, which comes before any mention of structure.

The non-revenue water issue tree, branch by branch

A city water utility treats and pumps 100 megaliters a day into its network. It bills customers for 72. The missing 28 megaliters a day are what the industry calls non-revenue water, and the board has released 12 million dollars of capital to reduce it.

The tree has three first-level branches, because water leaves the system in exactly three ways.

• Physical loss. Distribution leaks, main bursts, reservoir overflow. The water never reaches a customer.
• Commercial loss. Meters that under-register, accounts that exist on the network but not in the billing system, unauthorized connections. The water reaches a customer and no invoice follows.
• Unbilled authorized use. Firefighting, mains flushing, standpipes on construction sites. The water is delivered on purpose and nobody is charged.

Now the measurements, which is where the branches get their sizes. Minimum night flow analysis across 41 district metered areas attributes 19 megaliters a day to physical loss. A meter test program on a stratified sample attributes 4. Operations logs account for 1 megaliter of unbilled authorized use. That leaves 4 megaliters unexplained, which belongs in the commercial branch as apparent loss until somebody proves otherwise.

Add the four: 19 plus 4 plus 1 plus 4 is 28, which closes on the gap. A tree whose branches do not sum to the thing at the top is a list wearing a diagram.

Pricing each branch before choosing one

Here is the step the guides skip, and it reverses the answer. A megaliter recovered from the commercial branch and a megaliter recovered from the physical branch are worth different amounts, because they recover at different prices.

Water recovered from a leak is water the utility no longer has to produce. It saves the marginal cost of abstraction, treatment and pumping, which is 35 cents per kiloliter. Water recovered from the commercial branch was already produced and delivered; recovering it means invoicing it, at the full tariff of 1.40 dollars per kiloliter. One megaliter is 1,000 kiloliters.

The arithmetic, in the order it was done. Physical: 19 megaliters a day is 19,000 kiloliters, times 365 days is 6,935,000 kiloliters a year, at 35 cents is 2,427,250 dollars, which rounds to about 2.43 million. Commercial: the 4 megaliters of meter error plus the 4 unexplained is 8, so 8,000 kiloliters a day times 365 is 2,920,000 kiloliters, at 1.40 dollars is 4,088,000 dollars, or about 4.09 million.

So the commercial branch carries 8 of the 28 megaliters, which is 29 percent of the volume, and 4.09 of the 6.52 million dollars, which is 63 percent of the value. It is worth 1.7 times the physical branch on 42 percent of its volume. A tree sized in megaliters says open the leak branch. A tree priced in dollars says open the metering branch. Only one of those is a recommendation.

The test that tells an issue tree from a list

Take any two sibling branches and swap them. If the children of one would sit just as comfortably under the other, the parent split is not real and the tree has stopped dividing. Under physical loss you expect pipe age, pressure, soil movement and repair response time. Under commercial loss you expect meter class, meter age, billing data quality and connection audits. Swap them and both halves read as nonsense, which is the result you want.

The second test is depth. A branch is finished when it names something you can measure or ask for. "Meter under-registration" is not finished. "Under-registration by meter class, from the 1,200 meter sample already tested" is finished, because somebody can hand you that file this afternoon. On our own structuring rubric that is the difference between a Coverage mark and a Depth mark, and the reasoning behind each weight sits on how an answer is assessed.

The third test is the one candidates skip. Read the whole tree back and ask what it refuses to contain. This one contains no branch for tariff policy, because the tariff is set by the regulator and the board cannot move it inside a capital budget. Saying that out loud is worth more than adding a branch you cannot act on. The MECE principle covers what happens when branches double count instead.

How firms describe the same move

None of the firms that publish a method names a framework, and three of them describe decomposition in their own words. Deloitte publishes a five-step approach whose first step is "Understand the issue", and it adds a warning against leaning on clarifying questions: "Ask questions, but don't use them as a crutch. While it's important to ask clarifying questions, avoid relying on them too heavily. We're interested in how you connect the dots when solving the problem", checked 30 September 2026 at deloitte.com.

Charles River Associates writes that what it watches is "how you react to situations with limited information; how you apply logic and reasoning to the question; what data you draw on; and how you communicate to bring the interviewer along with your thinking", checked 30 September 2026 at crai.com. Three of those four are properties of a tree. What data you draw on is the branch-pricing step above.

Read together, the three published methods say the same thing in three vocabularies: split the question, attach data to the splits, and bring somebody with you through the split. None of them says which splits. That part is the job.

Where an issue tree loses marks

Our own grader reports a structure on four lines, and no consulting firm publishes a marking scheme of its own. The weights below are ours, read from the rubric code, each out of one hundred.

Drilling a tree against the clock

Rebuild the example with one number changed and see whether your ordering survives. Set the tariff at 90 cents per kiloliter instead of 1.40. The commercial branch becomes 2,920,000 kiloliters at 0.90, which is 2,628,000 dollars, against 2,427,250 for the physical branch. The two branches are now within 8 percent of each other, and the ordering that looked obvious is a coin flip. That is worth knowing before a partner asks.

Then do it again with the marginal production cost at 55 cents, which is what a drought year does. Physical becomes 6,935,000 times 0.55, or 3,814,250 dollars, and it overtakes the commercial branch at the 90 cent tariff. Two sensitivities, two reversals, and you now know the decision hinges on the ratio between tariff and marginal cost rather than on volume at all.

Practice the build itself in the structuring drill, which names the dimension to fix first, and work through worked shapes in issue tree examples. Both are free and need no account. Neither knows anything about water utilities, and neither will tell you which branch the client can afford to open, which is the judgment being bought. Choosing which branch to open when the budget is interview minutes, on figures a firm publishes itself, is worked through in prioritization frameworks.