Prioritization Frameworks 2026: Rank Branches by the Minute

What a prioritization framework has to receive before it ranks

A ranking device produces nothing on its own. It takes a closed list and returns an order, and three inputs decide whether that order means anything.

The first is the list itself, finished. The second is a unit on the value axis, which is dollars a year on most case problems. The third is a unit on the cost axis, which can be dollars, weeks, headcount or minutes. Miss either unit and the framework returns high and low, which are adjectives, and adjectives cannot be divided by each other. That failure is worked through in detail in the impact effort matrix.

The order matters, and Oliver Wyman publishes it. Its second step is to break the problem down: "Are there smaller problems inside the big one? Put your initial thoughts on paper." Its third step is the choice: "Pick an approach. Show your client why it's better than the alternatives. State your assumptions", checked 7 October 2026 at oliverwyman.com. Decomposition first, then selection, then a justification of the selection.

The same page says what the firm trains for, in its own words: "we learn to quickly pinpoint priority issues." Speed at choosing is named as a skill. Deloitte asks for the same thing in its third step, which is to "clearly outline the key issues and your findings based on the information provided", checked 7 October 2026 at deloitte.com. The word carrying the work in both is the word that selects.

Six named prioritization frameworks and the input each one needs

The query collects six devices that all do the same job with different inputs. Sorted by whether a case interview can actually feed them.

Four of the six were built for software delivery, and that is where they belong. RICE needs a user count per quarter for each candidate item. MoSCoW sorts a scope against a date. Neither has any meaning for the question of which branch of a profit tree to open at minute eleven.

Weighted scoring deserves a separate warning, because it looks the most rigorous of the six. It is only as good as its weights, and no consulting firm publishes a weighting for anything. Any weighted score you present in a case carries weights you invented thirty seconds earlier, so say so out loud and say what they rest on. We publish our own weights lower down, and they are ours rather than any firm's.

That leaves two devices that a case can feed, and the second is the one this page works. It needs no inputs the room does not already contain.

A published case that promises ten branches and opens three

Oliver Wyman publishes a dairy farm case at medium difficulty. A private equity client is deciding whether to buy one of two dairy farms in a county of Illinois. The firm prints a sample framework, and the framework is the whole problem this page is about.

It has three first level branches and ten leaves beneath them.

Four leaves plus three plus three is ten. Checked 7 October 2026 at oliverwyman.com.

Then the case asks four questions. One on competitive positioning, one on profitability, one on execution risks, and one asking for the recommendation. The fourth comes with a published cap: "keep it concise (30-60 seconds)". So three of the ten leaves get opened and seven do not, and the firm's own case is the proof that a ten leaf framework is a promise nobody keeps.

Which is the real exercise. Building the ten leaves takes two minutes and earns the structure mark. Choosing three of them is what the remaining twenty minutes consist of, and no published method says how. Our own pages build trees and price branches in dollars; issue trees does exactly that, and it never divides by the time a branch costs to test.

Recomputing the firm's own numbers, and the share that does not hold

Every figure below comes from the case and was redone by hand before anything was built on it.

Farm one has 10,000 cows at 100 units of milk each, which is 1,000,000 units. Farm two has 5,000 cows at 80 units each, which is 400,000 units. Together that is 1,400,000 units, so farm one holds 1,000,000 of 1,400,000, which is 71.4 percent rounded.

Now the profit lines. Standard milk sells to the single county processor at 20 dollars a unit, organic at 30, both set by the government. Farm one sells standard, so 1,000,000 units at 20 dollars is 20,000,000 dollars of revenue. Its feed runs 10 dollars a unit, so 1,000,000 at 10 is 10,000,000 dollars. Fixed costs are 5,000,000. So 20,000,000 less 10,000,000 less 5,000,000 is 5,000,000 dollars of profit, and 5,000,000 over 20,000,000 is a 25 percent margin.

Farm two sells organic. 400,000 units at 30 dollars is 12,000,000 dollars. Feed at 15 dollars is 6,000,000. Fixed costs are 3,000,000. So 12,000,000 less 6,000,000 less 3,000,000 is 3,000,000 dollars, and 3,000,000 over 12,000,000 is also 25 percent. Identical margins, and 5,000,000 less 3,000,000 is 2,000,000 dollars of absolute gap. The published answer says both of those things, and both check.

One published claim does not check. The case concludes that "farm one has more than 70% of the market share in terms of the number of cows and milk production volume." On volume that is right at 71.4 percent. On cows, 10,000 of 15,000 is 66.7 percent rounded, which is under 70. The sentence holds on one of the two counts it names. We publish both readings rather than deciding for the firm, which is the rule whenever a published exercise and its own arithmetic disagree.

Pricing six branches in dollars per interview minute

Take six testable branches drawn from the ten leaves and give each one two numbers. The first is the swing, meaning the dollars a year the answer can move. The second is the minutes it takes to get the answer in the room. Both the swing bands and the durations below are ours, and the assumptions are stated.

The arithmetic, in the order it was done. Feed is 10 dollars on 1,000,000 units, so the line is 10,000,000 dollars and a ten percent band is 1,000,000. Fixed costs are 5,000,000, so a ten percent band is 500,000. The acquisition price is the one figure the case withholds, so put it at ten times profit, which is 50,000,000 dollars, and one turn of that multiple is 5,000,000.

Two branches price at zero, and both do so for a published reason. The milk price is set by the government at 20 dollars, so there is no band to test. And the county has one processor which "is supply constrained and purchases all the dairy from both dairy farms in the area", so demand for everything the farm can make is already committed. Testing regional growth costs five minutes and moves nothing inside the holding period.

Then the divisions. 5,000,000 over 4 minutes is 1,250,000 a minute. 1,000,000 over 3 is 333,333, which rounds to about 333,000. 500,000 over 2 is 250,000. The spread between the best and the worst priced branch is a factor of five, on two numbers the case hands you.

The twelve minute budget, and the branch it forces out

Now the constraint, because an unconstrained ranking is a list. Say twenty minutes remain once the framework is on paper. The published case spends part of that for you: a risk brainstorm, the profitability arithmetic, and a recommendation capped at 30 to 60 seconds. Allow four minutes for the brainstorm, three to read and set up the math, and one for the recommendation. Twelve minutes are left to spend as you choose, and that allocation is ours.

The six tests above cost 4 plus 3 plus 2 plus 2 plus 5 plus 2 minutes, which is 18. Twelve minutes therefore buy two thirds of the available work, and something gets dropped.

By hand. The first set is 4 plus 3 plus 2, which is 9 minutes, and 5,000,000 plus 1,000,000 plus 500,000, which is 6,500,000 dollars. Three minutes remain, and the capability branch costs two, so the second set runs to 11 minutes. The third set is 5 plus 4 plus 3, which is 12 minutes, and it buys 5,000,000 plus 1,000,000, which is 6,000,000 dollars.

So the set that includes regional demand consumes the entire budget and returns 500,000 dollars less than a set that finishes with a minute to spare. The five minute branch is what pushes it out, and that cost has nothing to do with how interesting the question is. Say the trade out loud, because a stated drop is worth more than a silent one. The same reasoning under a capital limit sits in the impact effort matrix, and the branch-level version of the value axis is in driver trees.

Two money rankings that agree, and what both of them miss

Dollars a minute is one way to rank. Points of entry yield is another, and it is the one a private equity client actually cares about. Both were computed on the 50,000,000 dollar price posed above.

The arithmetic. 5,000,000 over 50,000,000 is 10.0 percent. Feed up ten percent takes the line from 10,000,000 to 11,000,000, so profit falls to 4,000,000, and 4,000,000 over 50,000,000 is 8.0 percent. Fixed costs up ten percent takes 5,000,000 to 5,500,000, profit falls to 4,500,000, and that is 9.0 percent. Paying fifteen times instead of ten puts the price at 75,000,000, and 5,000,000 over 75,000,000 is 6.666 percent, which rounds to about 6.7.

The two rankings agree. Price first at 3.3 points and 1,250,000 dollars a minute, feed second, fixed costs third, and the two zero branches last. Agreement between two independent orderings is usually reassuring. Here it is the warning, because both of them rank the capability branch last or refuse to rank it, and the capability branch is the one the published answer uses.

The branch with no number is the branch that decides

Oliver Wyman publishes two recommendations for this case and says both are defensible. The first buys farm one, on share, margin and a 25 percent productivity edge per cow. The second declines, and here is its whole reasoning: "while both farms exhibit strong profits, the client has no expertise in the dairy industry. They may struggle to promote further growth and profitability, hurting their return on investment", checked 7 October 2026.

Read what that does to the tables above. The branch that flips the answer is the one with no swing, no band and no dollars a minute. It sits in the third group of leaves, under compatibility with the client's portfolio, which is the group a money ranking reaches last. A candidate who spent all twelve minutes on the two priced branches arrives at the recommendation with excellent arithmetic and nothing to say about the thing the firm decided on.

The repair is a reservation rather than a different ranking. Hold a fixed slice of the budget for the branch that cannot be priced, before the ranking runs. Two minutes of twelve is about 17 percent, and it is the cheapest insurance in the exercise.

Two tests tell you which branch deserves that slice. Ask which branch, answered either way, changes the verb in your recommendation from buy to decline. And ask which branch nobody in the room can answer with a number, because those are the two properties that travel together. On this case both tests point at the same leaf, and neither of the money rankings can see it. Structured thinking covers the habit of reading a structure back to find what it refuses to contain.

Where prioritization frameworks earn or lose our marks

Our own grader reports a brainstorm on four lines. No consulting firm publishes a marking scheme, so the weights below are ours, read from the rubric code in src/lib/grading, each out of one hundred.

Drilling the choice against a clock

Four moves, and each one is a repetition rather than a reading.

1. Rebuild the six branch table with one number changed. Put the multiple at fifteen times rather than ten. The price branch now moves 2,500,000 dollars per turn, so 2,500,000 over 4 minutes is 625,000 a minute, and it still ranks first. The ordering survives that change, which is worth knowing before a partner tests it.
2. Then change the band on feed. At a five percent band the feed branch moves 500,000 dollars, so 500,000 over 3 minutes is about 167,000 a minute, and it falls below fixed costs at 250,000. One assumption reverses second and third place, which is why the band gets stated.
3. Time yourself on the structure. Charles River Associates publishes the plainest instruction on this: "the assessments are usually timed; have a watch ready to take note of the start and finish time so you can pace your progress", checked 7 October 2026 at crai.com. The same discipline applies to the ten leaves you just drew.
4. Close every list by naming the first move. Say the branch, the number behind it, and the branch you are choosing to leave shut.

Our brainstorming drill scores breadth, buckets, specificity and prioritization separately, and the free drills run on the weights above. How an answer is assessed sets out where those weights come from and who wrote them. None of them knows what a dairy farm trades for, which is the assumption the whole ranking on this page rests on, and that assumption is ours. The step after the choice is saying it in one breath, which case interview synthesis covers, and the rule that keeps branches from double counting is in the MECE principle.