Scenarios and sensitivity

A forecast answers "what do we expect?". Sensitivity analysis answers "what if one assumption is wrong?". Move one driver at a time, keep every other driver at the base case, and see how far the result moves. Use the same step for each driver, say 10%, so the answers can be compared.

For a per-lesson driver the shortcut is: change per lesson × lessons. Klasna's base case has 80,128 lessons and 2,251,968 UAH of operating profit.

• Price +10%: 65 UAH more per lesson, but the 2.5% payment fee takes its share, so 63.375 stays: 63.375 × 80,128 = +5,078,112 UAH.
• Tutor payout +10%: 41.50 UAH more cost per lesson. The fee is charged on the price, not on costs, so the full amount counts: −41.50 × 80,128 = −3,325,312 UAH.

One 10% miss on the payout wipes out the whole base profit and more. That's the kind of finding sensitivity analysis exists for: it shows which assumptions deserve the most care before anyone commits to the plan.

Put every driver's 10% move in one table and sort it by the size of the change. For Klasna's year (base 2,251,968 UAH):

Price and payout dwarf the rest because the margin is thin: of a 650 UAH lesson only 187.25 stays with Klasna. Ten percent of a big flow beats ten percent of a small leftover.

Learners and lessons per learner add about the same number of lessons, yet learners give less: more learners also bring more coordinators (two thresholds crossed, 76,000 UAH), more lessons from the same learners don't.

The ranking tells you where planning time should go: testing whether families accept the price and locking in tutor rates, not a week of arguing about learner counts.

Sensitivity analysis asks "what if the driver moves by 10%?". The reverse question is often more useful for the riskiest driver: how far can it move before the result hits zero, with everything else as planned?

For the price, spread the year's fixed costs over the lessons, add what each lesson costs, and divide by the share of the price left after the fee:

Break-even price = (fixed costs ÷ lessons + payout + video and AI) ÷ (1 − fee)

Klasna: 12,752,000 ÷ 80,128 = 159.15 of fixed costs per lesson. Add the 415 payout and 31.50 for video: 605.65. Divide by 0.975: 621.17 UAH. The plan is 650, so the price can fall only 4.43% before the year earns nothing, and 621 is roughly last year's average price.

That's a red line you can watch every month. If the realized average price after the September rise drifts toward 621, the year is heading to zero long before the annual report says so.

The distance between the planned value of a driver and its break-even value is the plan's margin of safety on that driver. Klasna plans an average price of 650 UAH and breaks even at 621.17, so the margin is only 4.43%.

A thin margin of safety means a small, ordinary miss is enough to turn the year. Last year's average price was about 620 UAH. If families don't accept the September rise and the price simply stays where it was, every lesson leaves 30 × 0.975 = 29.25 UAH less. Over 80,128 lessons that is 2,343,744 UAH, more than the whole base profit of 2,251,968, so the year ends at −91,776 UAH.

The check is quick: change per lesson after the fee × lessons, then subtract from the base profit. And the conclusion is a management one: a forecast whose profit depends on one price rise needs early evidence that the rise holds, not hope.

Both tools shake the same forecast, but they answer different questions.

Sensitivity analysis moves one driver at a time by the same step and keeps the rest at the base case. Because only one thing changes, the change in profit belongs to that driver alone, and you can rank the drivers. It answers: "Which assumption matters most?"

Scenario analysis moves several drivers together in a consistent story. In real life drivers rarely move alone: in a weak year fewer families come, more of them wait for discounts and the hryvnia slides. Klasna's weak case is learners −10%, average price 630 and the dollar at 45; the strong case is learners +10% and the dollar at 41. Each runs through the same sheet month by month. It answers: "What range of outcomes should we be ready for?"

Use sensitivity to decide where to look hard, and scenarios to decide what to prepare for. A scenario on its own can't tell you which driver did the damage.

In a weak scenario the damage comes from several small moves that multiply, not from one big one.

Klasna's weak case: learners −10%, average price 630 instead of 650, the dollar at 45 instead of 42. Volume falls by about 10%, to 72,116 lessons, which alone would cost less than the base profit. But every remaining lesson also earns less:

• price −20 UAH, of which 19.50 reaches profit after the 2.5% fee;
• video + AI: 0.75 × (45 − 42) = 2.25 UAH more per lesson.

That is about 22 UAH less per lesson, from a lesson that used to leave 187.25. Meanwhile fixed costs (team, office, most coordinators, marketing) barely move when learners fall by a tenth. Fewer lessons × thinner margin − almost the same fixed costs = −740,802 UAH instead of +2,251,968.

This is operating leverage at work: with high fixed costs, profit falls much faster than volume. A year can still look busy, with more than 72,000 lessons, and lose money.

A range like "from −0.74 million to +3.74 million" is honest, but on its own it only worries the board. The real use of a weak scenario is a plan agreed in calm weather:

• Trigger: an early, observable signal tied to a driver. Klasna's: October's learners come in 10% under plan.
• Response: actions decided and costed in advance, each with an owner. Halve marketing for June to August (saves 295,000 UAH) and postpone one of two hires (45,000 a month, 540,000 over twelve months).
• Result: the plan's effect computed for the weak case. Savings on fixed costs reach operating profit one to one: −740,802 + 835,000 = +94,198 UAH. Not great, but not a loss.

Agreeing all three before the year starts means nobody improvises in a panic in July, when most of the year is already spent. It also changes the board conversation: "we might lose money" becomes "here is the signal we watch and what we will do".

Many companies write a pessimistic scenario, look at the loss and file it. A plan that actually helps in a bad year has three parts, and each can be checked:

1. A trigger tied to a driver, visible early. "Learners 10% under plan in October" can be seen in November; "a bad year" can only be seen in August.
2. Actions agreed and costed in advance, each with an owner. Halving summer marketing saves 295,000 UAH; postponing one hire saves 540,000. Everyone knows who acts and when.
3. The plan's result computed for the weak case, not only the loss without it. Klasna's weak year moves from −740,802 to +94,198 UAH.

Two common substitutes don't work. Waiting for the year-end results is too late: summer is already lost. Cutting every line by the same 10% feels fair but ignores which costs move the result and may cut what earns money. A good plan is specific: which signal, which costs, how much.