Case: double marketing before exam season?

Leo wants to double marketing for exam season: 600,000 UAH a month in February–April 2027 instead of 300,000. "Twice the spend, twice the families." Last spring's tests show how each channel actually responds.

The current plan: paid social 150,000, search 90,000, school partnerships 60,000. From the tests that brings 50 + 34 + 26 = 110 families a month. Average CAC: 300,000 ÷ 110 = 2,727.27 UAH.

Double every channel and read the tests at the new levels: social at 300,000 brings 72, search at 180,000 brings 47, partnerships at 120,000 bring 47.

72 + 47 + 47 = 166 families, not 220.

Doubling the spend adds 56 families, about half again. Channels saturate: social shows the ad to the same parents again, and search runs out of people searching that week.

The slip is to scale the current result linearly (110 × 2 = 220). Always read the response at the new spend level from data, never extrapolate it.

A budget question is almost always a question about the next hryvnia: should we spend 300,000 more? The first 300,000 is already decided and keeps buying families at its old price. So the extra money has to be judged on its own.

• Average CAC = total spend ÷ total customers. Doubling the budget moves it from 2,727.27 to 3,614.46 UAH: up by a third, which sounds acceptable.
• Marginal CAC = extra spend ÷ extra customers = 300,000 ÷ 56 = 5,357.14 UAH: almost twice the old average.

The average looks calm because the cheap first half dilutes the expensive second half. If you approve the increase on the average, you are really paying 5,357 for each new family without seeing it.

The same logic works in reverse: when cutting a budget, the families you lose are the most expensive ones, so a cut saves more per lost family than the average suggests. The slip is to compare the new average with a target and stop there.

Marginal CAC is the cost of one more customer: extra spend ÷ extra customers it brings.

For the whole plan: (600,000 − 300,000) ÷ (166 − 110) = 300,000 ÷ 56 = 5,357.14 UAH per extra family. The average only rises to 600,000 ÷ 166 = 3,614.46, because the first 300,000 still buys families at 2,727. The average hides the expensive margin.

Channels saturate at different speeds, so split it:

• search, 90,000 → 180,000: 34 → 47 families, 13 more for 90,000. Marginal CAC 90,000 ÷ 13 = 6,923.08;
• paid social, 150,000 → 300,000: 50 → 72, 22 more for 150,000: 6,818.18;
• partnerships, 60,000 → 120,000: 26 → 47, 21 more for 60,000: 2,857.14, barely above their average.

One channel still buys cheap families; two have hit the wall. The plan's 5,357 is a blend of both.

The slip is to divide the new total spend by the new total customers for a channel (180,000 ÷ 47 = 3,830 for search): that is still an average.

A plan-level number blends channels that behave very differently. Doubling Klasna's three channels:

The plan's 5,357.14 is a weighted mix. Partnerships still have room: their extra families cost little more than their average. Social and search are saturated: the extra families cost more than twice the old average.

Why channels differ: paid social keeps reaching the same parents, search depends on how many people search for tutoring that week, while each new school partnership opens a fresh group of families.

So never approve or reject a budget increase as a whole. Compute marginal CAC per channel, compare each with the value of the customers it brings, and keep only the increases that pay. The slip is to judge every channel by the plan's blended figure.

For a seasonal customer, value isn't a fixed LTV: it depends on how much season is left when they join.

Klasna's exam families study until the exams at the end of May, about seven maths lessons a month, which leaves 1,400 UAH of contribution a month. Then most stop. Some come back: about 30% return in September and leave around 2,000 each over the autumn.

Value of one family = months of study left × 1,400 + expected return.

The expected return is a probability times a value: 0.3 × 2,000 = 600 for every spring family, whether or not that particular family returns.

• February start: 4 months (February–May) × 1,400 = 5,600 + 600 = 6,200;
• March start: 3 × 1,400 = 4,200 + 600 = 4,800;
• April start: 2 × 1,400 = 2,800 + 600 = 3,400.

An April family is worth barely half of a February one. The slips: counting the start month out (March has three months to the end of May, not two), or adding the full 2,000 instead of the expected 600.

Is 5,357.14 UAH per extra family too much? That depends on what an extra spring family leaves Klasna, and that depends on when it starts.

Exam families study about seven maths lessons a month until the exams at the end of May, leaving about 1,400 UAH of contribution a month; then most stop. About 30% come back in September and leave around 2,000 each, which adds 0.3 × 2,000 = 600 to every spring family.

• February start: 4 × 1,400 + 600 = 6,200;
• March: 3 × 1,400 + 600 = 4,800;
• April: 2 × 1,400 + 600 = 3,400.

Against a marginal CAC of 5,357.14, only a February family pays back. March and April families cost more than they leave. Against the 2,727.27 average every month looks fine, which is exactly why the average misleads.

The rule: put the marginal cost next to the value of the marginal customer, the one the extra money actually buys. For seasonal products, that value shrinks every month closer to the season's end.

On value the whole plan is slightly negative: 900,000 UAH of extra ads for 56 × (6,200 + 4,800 + 3,400) = 806,400 of value, −93,600. The bank account sees it in a different order.

The extra 300,000 leaves every month in full. Each extra family brings 1,400 a month while it studies, and the families pile up: 56 in February, 112 in March, 168 in April.

• February: 56 × 1,400 − 300,000 = −221,600;
• March: 112 × 1,400 − 300,000 = −143,200, running total −364,800;
• April: 168 × 1,400 − 300,000 = −64,800, total −429,600, the low point;
• May: no ads, 168 × 1,400 = +235,200, total −194,400;
• June–August: nothing;
• September: 30% of 168 return, 0.3 × 168 × 2,000 = +100,800, total −93,600.

The low point lands at the end of April, right before Klasna's summer dip, and September never closes the gap. The slip is to count only the ad spend as cash out and the families' full value as cash in on day one.

With marginal CAC by channel and the value of a spring family by start month, each channel's increase gets its own verdict.

An extra family a month across February–April is worth 6,200 + 4,800 + 3,400 = 14,400 UAH. For each channel over the three months:

• partnerships: 21 extra families a month × 14,400 = 302,400 of value for 3 × 60,000 = 180,000 extra spend: +122,400;
• paid social: 22 × 14,400 = 316,800 for 450,000: −133,200;
• search: 13 × 14,400 = 187,200 for 270,000: −82,800.

Together: 122,400 − 133,200 − 82,800 = −93,600, the whole plan's result. One good channel was hidden under two bad ones.

Partnerships add a family for 2,857, below even an April family's 3,400, so they pay in every month. The data covers the higher level (the test went up to 120,000), but partnerships need time to set up: the school talks would have to start in January.

The slip is to reject the whole plan because its total is negative, and lose the one part that earns money.