Marketing Cases 2026: Blended CAC Hides the Real Cost

What a marketing case interview prompt is really deciding

Three shapes cover most of these prompts. Allocate a fixed budget across channels, decide whether to enter a new channel, or explain why acquisition has become more expensive. The third is the most common in interviews and the least well handled, because the answer usually lives in a term the candidate never named. A launch prompt sits outside those three, and each of the four Ps it has to decide can be priced before you rank them.

Altman Solon publishes the dimensions it scores on a business case. Its interview page lists "Structured thinking and problem structuring", "Quantitative and qualitative analysis", "Business judgement", "Clear communication" and "Creativity and practical recommendations", checked 29 September 2026 at altmansolon.com. Read that list against a marketing prompt and the middle two are where candidates lose. Quantitative analysis means producing a cost per customer, and business judgment means saying which of the inputs you would refuse to trust.

The market this argument happens in is measurable. The Census Bureau reports that "E-commerce sales in the second quarter of 2026 accounted for 17.1 percent of total sales", growing 12.2 percent year over year against 6.7 percent for total retail, checked 29 September 2026 at census.gov. So the online channel is roughly a sixth of retail and pulling away. A channel shift case is about a real and quantified shift, and quoting the figure beats asserting the trend.

One clarifying question opens every one of these cases well. Ask whether the client measures acquisition cost against new customers or against all orders. Those two denominators differ by the repeat rate, and a business with three orders per customer per year reports a cost per order that is a third of its cost per customer. Both numbers get called CAC in the room.

The marketing case chain from spend to a paying customer

Write the chain as a product and every branch names itself. Spend divided by cost per thousand impressions gives impressions. Impressions times click rate gives clicks. Clicks times conversion rate gives buyers. Spend divided by buyers gives the cost of a customer.

Run it once with numbers. A brand puts $600,000 into paid search over a year at a $12 cost per thousand impressions, so $600,000 divided by $12 buys 50 million impressions. A 1 percent click rate on 50 million impressions is 500,000 clicks, which puts the cost of a click at $1.20. Then 4 percent of 500,000 clicks is 20,000 buyers, so the cost of a customer is $600,000 divided by 20,000, which is $30.

That chain has four terms, and the useful property is what it does to errors. Any single term falling by a fifth raises the cost of a customer by a quarter, because one divided by 0.8 is 1.25. A click rate that slides from 1.0 to 0.8 percent does the same damage as a conversion rate sliding from 4.0 to 3.2 percent. Neither looks alarming written down, and either one turns a $30 customer into a $37.50 customer.

This is why an auction based channel gets expensive in a way that feels sudden. Competitors bid the cost per thousand up while the click rate drifts down, and the two effects multiply rather than add. A 15 percent rise in media cost alongside a 15 percent fall in click rate takes cost per customer up by 1.15 divided by 0.85, which is 35.3 percent.

Brand effects belong in this structure, and they belong in one place. Brand moves the click rate and the conversion rate, because that is the only channel available to it inside the equation. Saying that is more useful than asserting a brand benefit the arithmetic cannot carry. A branch with no term attached to it will not survive the first follow up question.

Why a blended acquisition cost flatters every channel

Now put three channels next to each other. The brand spends $600,000 on paid search and acquires 20,000 customers, $900,000 on paid social and acquires 18,000, and nothing at all on organic search and referral, which still delivers 22,000.

The blended figure is $1,500,000 divided by 60,000, which is $25.00. The paid figure is $1,500,000 divided by 38,000, which is $39.47. The paid figure runs 57.9 percent above the blended one, and the blended number is the one that appears in most board packs.

The reason it matters is that the blended figure moves for reasons nothing to do with marketing. A public relations story lifts organic customers by 5,000, blended cost falls to $1,500,000 divided by 65,000, which is $23.08, and that is 7.7 percent below the old blended figure without a single bid changing. Stated as efficiency rather than cost the same move reads as an 8.3 percent gain, because one divided by 23.08 is 8.3 percent above one divided by 25.00, and the two numbers describe one event. Running the budget off that figure means increasing spend into channels that just got no better.

So the decision rule is to price each channel on its own denominator and keep the blended figure as a reporting line. The follow up question an interviewer will ask is what you would do with the organic customers, and the honest answer is that they set the size of the prize rather than the price of the next customer.

Incrementality, and the buyers who were coming anyway

Every channel figure above assumes that a customer attributed to a channel was created by it. That assumption is false in a knowable direction, and the correction is a single division.

Suppose one paid search buyer in five would have bought this year regardless, arriving through organic search if the ad had not been there. The channel created 20,000 times 0.8, which is 16,000 customers. Its true cost is $600,000 divided by 16,000, which is $37.50. That is exactly 25 percent above the reported $30, and the factor is always one divided by the incremental share.

Do the same to paid social. At the same one in five overlap it created 14,400 customers, so its true cost is $900,000 divided by 14,400, which is $62.50. The gap between the two channels widens in absolute terms, from $20 to $25, because the same proportional correction applies to a larger number.

Two consequences follow, and both are worth saying in a recommendation.

1. Test by withholding. Turning a channel off in one region and measuring what happens to total customers gives the incremental figure directly. An attribution model allocates credit among channels, which is a different job, and it cannot produce this number.
2. Overlap is highest where intent is highest. A brand search ad captures people who typed the brand name, so its overlap with organic is large. That makes the cheapest reported channel the one most likely to be mismeasured.
3. Size the correction before you price it. If the withheld test is not available, state the overlap you assumed and show the answer at two values. Here, 10 percent overlap gives paid search a cost of $33.33 and 30 percent gives $42.86.

The retention number a marketing case turns on

None of the costs above can be judged without knowing what a customer is worth, and that requires one number nobody in the room has measured.

ZS publishes a usable anchor. Its consumer goods case gives "Profit per product: $24.99 minus $9.99 equals $15" and then works from "Assuming one purchase a year", checked 30 September 2026 at zs.com. At one purchase a year and $15 of contribution, a first year acquisition that costs more than $15 loses money in that year, full stop. That is a published ceiling rather than a rule of thumb.

Take our brand at $15 of contribution per purchase and 1.8 purchases a year, so annual contribution per customer is $15 times 1.8, which is $27. Paid search at a true cost of $37.50 pays back in $37.50 divided by $27, which is about 1.4 years. Paid social at $62.50 pays back in about 2.3 years.

Whether those are acceptable depends entirely on retention. At 50 percent annual retention the expected customer life is one divided by 0.5, so two years, and lifetime contribution is $54. At 70 percent retention the life is one divided by 0.3, which is about 3.3 years, and lifetime contribution is roughly $90.

Put the four combinations together and the decision becomes visible. At 50 percent retention, paid search returns $54 against $37.50, a ratio of 1.44, while paid social returns $54 against $62.50 and destroys value. At 70 percent retention, paid search reaches 2.40 and paid social reaches 1.44. The same channel is fatal or fine depending on a rate that sits in the product team's hands.

That is the recommendation this case wants. Fund paid search now, hold paid social at test level, and make the paid social decision conditional on retention clearing a measured threshold. Oliver Wyman's own pricing case reaches for the same lever from the other side, recommending a subscription because it "provides consistent, recurring revenue, helping to cover fixed costs over time", checked 30 September 2026 at oliverwyman.com. Changing the payment model is a way of buying retention.

Where a channel answer loses marks

No firm publishes a marking scheme. The weights below are ours, applied by our own grader, read from the rubric code in src/lib/grading. The reasoning behind each sits on how an answer is assessed.

What we corrected on this page

Each entry carries the date of the change.

7 October 2026. A percentage pointed the wrong way, twice. This page said the blended acquisition cost of $25.00 was "57.9 percent lower" than the paid cost of $39.47, and the summary at the top repeated the same claim as a gap of 57.9 percent. The arithmetic behind the number was right and the direction was wrong. The difference between the two figures is $14.47. Divided by the blended cost it gives 57.9 percent, which is how far the paid figure sits above the blended one. Divided by the paid cost it gives 36.7 percent, which is how far the blended figure sits below the paid one. Those are two different statements about the same pair of numbers, and this page made the second one using the first one's value.

It matters here more than it would elsewhere, because the section exists to teach a candidate that a blended figure flatters a channel mix. A reader who redid the division and got 36.7 percent would have concluded they had made the mistake. They would have been right and the page wrong, which is the failure this site cannot afford on a page about arithmetic. Both sentences now name what the 57.9 percent measures. No other figure in the section changed, and the table was correct throughout.