Finance Cases in Consulting Interviews: The But-For Number

Three firms publish finance cases, and not one discounts a cash flow

We read every published case with figures we could find at firms whose work is financial. Here is what the finance content turned out to be.

Not one of them discounts a cash flow. Every one of them asks a candidate to construct a baseline that did not happen and measure a real number against it. That is worth knowing before you spend an evening on terminal value formulas, and one caveat belongs with it. The absence of a published discounted cash flow tells you what these firms chose to publish. It says nothing about what an individual interviewer may ask.

Discounting and time-based criteria do come up, and they have their own arithmetic. NPV in cases covers discount factors and the rate that changes the sign; return and payback covers the undiscounted pair and a firm case that rejects a profitable venture on a time criterion.

The but-for price, defined by the firms whose finance cases use it

Both economic consulting firms define the concept in their own words, and the definitions are almost identical, which makes them worth quoting rather than paraphrasing.

Bates White writes this in its battery case. "Damages are the difference between the actual price that TalkTek paid for batteries and the 'but-for' price they would have paid but-for the alleged price-fixing."

Cornerstone Research writes that in its securities case, damages "boil down to the difference between Zilo's observed stock price during the class period and the 'but-for' price, or what the price would have been absent news of the alleged misconduct."

One number is observed and the other is constructed. The observed number takes ten seconds. The constructed one takes the whole interview, and every point of disagreement in the case lives inside it.

Cornerstone also states the failure mode directly. The flaw it identifies in the plaintiffs' work is that their "approach effectively attributed all stock price movements during the class period to the alleged misconduct. However, there could also have been other news unrelated to the alleged misconduct that would affect Zilo's stock price." The intellectual property case is framed the same way, as a mandate "to determine the extent to which ACME's alleged lost profits resulted from unauthorized use of its manufacturing techniques." The words that carry the analysis are "the extent to which."

How much a trend assumption carries in a finance case

Bates White's battery case constructs its baseline by extrapolating a trend, and publishes the result. "If you assume that, instead, prices should have continued decreasing at a rate of $0.50 a year, then TalkTek paid $3.50 more per battery than it otherwise would have in 2019 and 2020."

The price table itself is published as an image we could not read, so the annual prices are not reproduced here. What can be tested is how much weight the trend assumption carries, and the extension below is ours rather than the firm's.

Suppose the last unaffected year is 2018. Under a $0.50 annual decline, the but-for prices are $0.50 below 2018 in the first affected year and $1.00 below in the second. Now assume the decline was slowing, at $0.25 a year. The but-for prices become $0.25 and $0.50 below 2018, so they are higher, and the overpayment shrinks by $0.25 in the first year and $0.50 in the second.

Half a dollar against a $3.50 headline is 14 percent of the claim, from one judgment about whether a downward trend was linear. Multiply across millions of batteries and that single assumption is the largest number in the case. The candidate who says so has found the analysis; the candidate who extrapolates and moves on has done the arithmetic.

The number we will not compute

Cornerstone's Zilo case opens with a company that "completed its initial public offering (IPO) at $20 per share" and where, in the firm's words, "In less than a year, the stock was trading at over $60 per share." It ends with the firm's conclusion. "By eliminating stock price movements unrelated to the allegations in the case, the team concluded that only $3 of the price drop might relate to the allegations."

The obvious next step is to express $3 as a share of the total decline. We do not, because the page never publishes the total decline. It gives an offering price and a peak, and says nothing about where the stock landed. Dividing $3 by a fall we invented would produce a confident percentage resting on nothing, which is the exact failure the case is about.

What can be said without a denominator is the scale of the stake. Damages are $3 multiplied by the affected share count, whatever it is. Had the plaintiffs' method survived, the multiplier would have been the entire decline instead. On a hypothetical 50 million shares, purely to show the arithmetic, $3 is $150 million and a $10 attribution would be $500 million. The construction of the baseline is worth more than every other line of the analysis combined.

In a finance case, a valuation multiple is a counterfactual too

If a case does hand you a valuation, the same discipline applies, and this worked example shows why. The figures are invented.

A business reports $12 million of EBITDA and the owner asks $96 million. That is 96 over 12, so 8.0 times EBITDA. Comparable businesses trade at 10 times. At 10 times, $12 million is worth $120 million, so the asset looks $24 million cheap.

Then a line item appears. Three million dollars of that EBITDA comes from a single contract that expires in eighteen months and is not expected to renew. Adjusted EBITDA is $12 million minus $3 million, so $9 million.

The implied multiple on the sustainable figure is 96 over 9, which is 10.67 times, against the 8.0 times the headline suggested. A line worth a quarter of EBITDA, since 3 over 12 is 25 percent, moved the implied multiple by 2.67 turns and swung the conclusion by $30 million on a $96 million price, which is roughly 31 percent of the asking price.

The counterfactual here is the same shape as the damages one. What would this business earn absent the thing that is about to stop. A multiple applied to a reported figure answers a question nobody asked.

Profit equals revenue less costs, and the firms mean it literally

Cornerstone's hint for its intellectual property case reduces the exercise to profit being revenue less costs. Read next to Bates White's correction, that plainness is the point.

Bates White's third answer nets a cost move against a price move: "overpayment per unit was $1.50 instead of $3.50, since the actual price per battery of $3.50 above the but-for price is offset in part by the costs for 2019 and 2020 being $2.00 above the trend." Check the subtraction: $3.50 minus $2.00 is $1.50, and the $2.00 removes 57 percent of the headline number, because 2.00 over 3.50 is 0.571.

The lesson is that lost profit is never a revenue figure. A plaintiff deprived of sales did not incur the costs of making them, and a buyer overcharged on inputs may have passed some of it on. Charles River Associates puts the working requirement bluntly on its interview tips page: candidates should "be prepared to explain how you came to your answer and show your math." On a finance case the shown math is where the netting becomes visible.

How we score a finance answer, and whose scale this is

No firm publishes a marking scheme. Cornerstone publishes model answers with hints; Bates White publishes model answers with video walkthroughs; neither publishes a weighting, and Charles River Associates publishes only tips. The hundred points below are ours, written by us and read out of src/lib/grading/rubrics.ts. They are what our own grader measures on numerical work, and no firm has seen them.

Four questions that locate the counterfactual in a finance case

1. What is the observed number, and what is the constructed one? Write them in two columns before doing anything. Only one of them is a fact.
2. What builds the baseline? A trend, a comparable market, a comparable firm, a period before the event, or a model. Each has a different weakness and the case usually hands you one of the five.
3. What else moved in the same period? This is the netting question, and Bates White's own correction is the model answer to it.
4. How sensitive is the result to the baseline? Change the assumption by a plausible amount and quote the range. This is the sentence the analysis exists to produce.

Question three is the one that separates a first answer from a defensible one. Question four is the one that almost nobody volunteers, and it is worth the full Interpretation weight on its own.

The arithmetic underneath all of it is borrowed rather than invented, and every relation a case uses is recomputed in the consulting math reference.

Drilling the finance case counterfactual

Practice on any case with a before-and-after number, which is most profitability cases. Force yourself to write the baseline explicitly rather than assuming last year, then name one other thing that changed in the same window and net it off.

Then do the sensitivity, out loud, as a range rather than a point. "One fifty per unit on the stated cost adjustment, and somewhere between one and two if the pre-period trend was flattening." That sentence is what the firms above are testing for, and it takes one extra subtraction.

Where the finance content is a discount rate rather than a baseline, the arithmetic and the failure modes are different, and NPV in cases handles them. Where it is a cost structure question, scale against profitability has the formula.

The weights above and the reasoning behind them sit on how an answer is assessed. The math drill reports the four dimensions separately and names the first to fix. Every figure on this page is either quoted from the three firm pages linked in the text or computed here in full, and nothing sits behind a sign-up.