When To Use a Framework in 2026: The Switching Signal

L.E.K. publishes both instructions and no tiebreak

L.E.K.'s interview preparation page lists seven numbered steps for working a case. Step three is titled sticking to your approach, and its gloss reads: "Outline how you are thinking and move methodically to address each issue, point-by-point". Step six is titled coming up with an answer, and its gloss reads: "Commit yourself to a course of action, but show flexibility and a willingness to rethink in the face of new data", checked 30 September 2026 at lek.com.

Both steps are still on the page. Neither says which wins, and neither names the data that should trigger the rethink. A candidate reading the list in order gets told to be methodical and then, three items later, to be willing to abandon the method.

Kearney poses the same tension without resolving it either. Under its communication tip it writes: "Be coachable, soliciting feedback and integrating it quickly. At the same time, be firm about the things you believe in", checked 30 September 2026 at kearney.com. Two firms, two published tensions, no published rule.

So the rule has to come from somewhere else, and it can be arithmetic instead of temperamental. The question to answer is narrow: can the branch you are standing on still be large enough to contain the whole answer. That question has a number attached to it, and the number is usually three data points away.

Whether to use a framework turns on whether the branch holds the gap

Every case has a gap: a number that moved, a target that is missed, a decision that is open. A branch is worth continuing while it could still account for the whole gap. It stops being worth continuing the moment the arithmetic says it cannot.

That test is available early and it is cheap. Ask for the two or three numbers that size your branch, compute its maximum possible contribution, and compare that with the gap. If the branch caps out at a fifth of the gap, four fifths of the answer is somewhere else, and continuing is a choice to be wrong methodically.

Deloitte's published five-step approach implies the same discipline in its second step, which is identifying the underlying assumptions. It also warns against a common substitute for the test: "Ask questions, but don't use them as a crutch. While it's important to ask clarifying questions, avoid relying on them too heavily", checked 30 September 2026 at deloitte.com. Asking for the two numbers that size your branch is not a crutch. Asking eleven exploratory questions instead of sizing anything is.

The retailer, decomposed two ways

A specialty retailer's revenue fell from 214 million dollars to 196 million over two years, so the gap is 18 million dollars, which is 8.4 percent. The candidate opens with a traffic hypothesis, which is the reasonable first guess in specialty retail.

Then the data arrives. Transactions went from 6.1 million to 6.05 million, so down 50,000, which is 0.8 percent. Average basket was 214 million divided by 6.1 million, or 35.08 dollars, and is now 196 million divided by 6.05 million, or 32.40 dollars, so down 7.7 percent.

Decompose the gap into the two effects, holding one factor at a time.

The two effects sum to 18.0 million dollars, which closes on the gap. Transactions explain 1.75 of 18, which is 9.7 percent. The traffic branch has a ceiling, and the ceiling is a tenth of the problem.

That is the switching signal, in one comparison. It arrived from three numbers, it took forty seconds of arithmetic, and it is not a matter of judgment or confidence.

What the switch buys, and what it costs

Switch and the next question is already waiting, because basket decomposes further. Units per transaction went from 4.2 to 4.05, which is down 3.6 percent. Price per unit went from 35.08 divided by 4.2, or 8.35 dollars, to 32.40 divided by 4.05, or 8.00 dollars, which is down 4.2 percent.

Check that the pieces multiply back: a 3.6 percent fall in units and a 4.2 percent fall in price per unit compound to 7.7 percent, which is the basket fall. Both halves are real and price per unit is the larger.

So a candidate who switched now has a chain that reaches a decision: 90 percent of the gap is basket, and slightly more than half of basket is price per unit rather than units. That points at discounting and mix, and it can be tested against the promotional calendar in the next question.

The cost of the switch is the sunk time. If the traffic branch consumed eight minutes of a thirty-minute case, switching writes off eight minutes and buys 90.3 percent of the answer. Holding the branch protects the eight minutes and spends the remaining twenty-two on 9.7 percent of the problem. Stated as a trade it is not close, and stating it as a trade is what makes it a decision rather than a loss of nerve.

Four signals in a case interview that say hold

Not every surprise is a switching signal, and switching on every new fact is its own failure mode. Hold the structure when any of these is true.

1. The new fact sits inside a branch you already have. A competitor's price cut belongs under price. Nothing needs to move.
2. The new fact is smaller than the gap by an order of magnitude. It is a detail for the risk section rather than a new spine.
3. Your branch has not been sized yet. Abandoning an unsized branch replaces one guess with another.
4. The interviewer is testing the branch rather than redirecting you. A follow-up question inside your branch is usually a request for depth.

Signal three is the one that catches most people. A branch that feels wrong and has never been quantified is a feeling. Size it first, then decide, and the sizing takes less time than the agonizing.

Four signals in a case interview that say switch

Switch when the arithmetic or the objective has moved, and say out loud that you are switching.

1. Your branch caps out below the gap. The retailer arithmetic above, and it is the cleanest of the four.
2. The objective changed. A prompt that turns from growth to cash flow has changed what the branches have to add up to.
3. A branch you do not have arrives with a number attached. New data on an unlisted factor is a coverage failure showing itself.
4. A unit has changed under you. Per store, per square foot and per transaction are different trees, and the wrong one cannot be repaired by adding branches.

Kearney gives explicit permission for all four: "If things aren't going well, don't try to tough it out. Acknowledge that you are stuck, seek more information, or pursue a different logical path." The firm treats the switch as a professional move rather than a concession.

The sixty-second switch script

Say it in this order and the switch reads as control rather than collapse. First the arithmetic: transactions can only explain 1.75 million of the 18 million. Second the conclusion: so the gap is basket. Third the new structure: basket splits into units per transaction and price per unit, and I would like to size both. Fourth the concession, once and briefly: the traffic branch is capped, so I am leaving it.

Simon-Kucher lists what it watches in a case, and two of its named factors are logical structuring of information and the ability to interpret and communicate results, checked 30 September 2026 at simon-kucher.com. A narrated switch demonstrates both at once, because the interpretation is the reason for the switch.

The weights below are ours, taken from the rubric code our own grader runs. No consulting firm publishes a marking scheme, so none of this is L.E.K.'s reading or Kearney's.